Method outline

A schematic of the overall synthesis process of sticky-ended DNA molecules (SE-DNA) is presented in Fig 1A. It begins by selecting a DNA template with a known sequence (e.g. from a commercially-sourced bacteriophage genome) such that PCR primers can be designed to amplify a specific internal subregion of this sequence. After amplification, the product is digested with restriction endonucleases to create short (e.g. 4 nt) single-stranded overhangs on one or both ends of a central fragment. These ssDNA overhangs are targeted in a subsequent ligation step as the attachment sites of small double-stranded “linker” molecules.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Method outline.

a) Overview of the basic SE-DNA synthesis protocol. The length of the final product is determined primarily by the distance between the two restriction sites (which can be adjusted as desired during the PCR step, see main text) while the number of sticky ends and their sequences are flexibly determined by the choice of linker molecule(s). b) Agarose gel (0.5%, 0.5× TAE, pre-stained with Gel Red, 70 V) of 6.2 kbp SE-DNA produced by this method, here using λ-DNA template, KpnI and SacI restriction enzymes, and a single linker. The second lane from the left (“50°C”) is the result of incubating the product for 1 hr at 50°C (~1 nM DNA in 10 mM MgCl₂) and shows two main bands: the linear product (“L”) migrating between the 5 kbp and 7 kbp DNA ladder bands, and a slower-migrating circular product (“C”) resulting from the annealing of the two complementary ends of a linear molecule. The next lane over (“70°C”) is the result of taking an aliquot of the same “50°C” mixture and heating it above the melting temperature of the 12-nt sticky ends (10 min at 70°C)–here the circular band is greatly reduced in favour of the linear band as the annealed ends of the circles are reversibly melted.

https://doi.org/10.1371/journal.pone.0290559.g001

The specific design of the linkers used in this step will define both the presence and the sequence of the sticky ends of the final product. In the example presented in Fig 1A, the PCR amplicon has been digested by two restriction enzymes to produce a different short ssDNA overhang on each end. The linker in this example then consists of two oligonucleotides, each of which features a complementary region to one of the two short overhang sequences (red and green segments in Fig 1A) followed by a region that is complementary to its analogue on the other linker strand, allowing for their hybridization into a single molecule. In this case, the sticky end sequences on the final product will be the same as those holding the two linker strands together, and will thus be cohesive with each other. However, by choosing to use one or two different linker molecules (two or four linker oligonucleotides) and by varying the lengths and specific sequences of their central regions, an essentially arbitrary ssDNA extension could be independently placed on one or both ends of the final product.

The final step in the process consists of melting off one of the linker strands on each end to expose the desired ssDNA extensions comprising the sticky ends. This relies on the fact that the linker molecules are not 5′-phosphorylated (as can be specified when ordering commercial chemically-synthesized DNA oligos) and so during the ligation step, only their 3′ ends are covalently attached to the digested amplicon by the ligase. This leaves nicks placed near the ends of the products that lead to relatively unstable terminal stretches of annealed dsDNA (~10 bases long, exact length determined by the linker sequences, see Fig 1A) that can be selectively melted by heat or chemical denaturation, leaving the stable central region that composes the bulk of the molecule (~1000+ bases long) intact. S1 and S2 Files in the Supporting Information contain the oligo sequences used in this work (primers and linkers), as well as detailed experimental protocols, respectively.

Referring back to the example of 1A we see that the length of the final sticky-ended product in this case is determined primarily by the distance between the recognition sites of the two restriction enzymes in its blunt precursor (PCR amplicon). An obvious method of modifying this length to meet the needs of an application could therefore be to target alternative recognition sites in the template sequence or even to use different restriction enzymes. There are, however, several downsides to such an approach. For one, the exact product lengths achievable this way are limited to only those that correspond to the locations of existing recognition sites (which may or may not be beneficially distributed in a given DNA template). Additionally, using different restriction enzymes on a template typically produces different cleaved ends, varying in overhang length (including 0 nt–blunt ends), polarity (5′ or 3′), and/or sequence. This necessitates careful consideration that a targeted cleavage site is compatible with the construction of a particular product, or, at the very least, requires the re-design and purchase of additional linker molecules for each new cut type.

A more flexible approach to setting the product length might instead start with the selection of a single set of restriction enzymes whose recognition sites are rare or non-existent in the native sequence of the DNA template. New recognition sites for these enzymes can then be introduced at practically arbitrary locations within the template (thus defining the product length) by modification of the primers used in the PCR stage of synthesis–either by point mismatches between primer and template to create cleavage sites from closely-related sequences or by simple addition of the recognition sequence (typically ~6 bp) to the 5′ end of the primer [26]. S3 File in the Supporting Information presents an illustrative example involving the combination of KpnI and SacI restriction enzymes with a λ-DNA template, where the “forward” PCR primer is fixed near the location of a KpnI site and the “reverse” primer is placed somewhere within the region directly downstream (~17 kbp wide), which is devoid of SacI sites in native λ-DNA. Even when restricting the placement of the reverse primer to sequences within this region that differ by one base at most from a SacI site (and so could be converted to an active site with a single base substitution in the primer), a full range of product lengths from 0–17 kbp is achieved, with an average gap between adjacent lengths of ~350 bp (see S3 File). Creating a wide range of different SE-DNA product sizes in this case then involves modifying a base synthesis protocol by only a single, low-cost PCR primer at a time.

Finally, we note the importance of removing any “extra” DNA fragments with single-stranded overhangs that appear at various stages of the synthesis protocol. These include the “caps” of the PCR amplicon that are cleaved off of the central fragment during the restriction digest, as well as any excess linker molecules that have not been incorporated onto this fragment during the ligation stage. If these spurious fragments are not removed, their presence in the reaction mixtures of subsequent stages could create interference by annealing with complementary regions on the main product and thus blocking interactions of these regions with their intended targets (see Fig 1A). So long as these extra fragments were relatively short (≲ 100 bp), their separation from the larger (1000+ kbp) main product was found to be easily achieved by incorporating a standard DNA purification step–e.g. by silica membrane-based spin column or alcohol precipitation–in between synthesis stages (see S4 File in the Supporting Information). Crucially however, it was observed that any purification attempts that involved extracting DNA from agarose gels (as might be necessary for separating larger DNA fragments by size) created problems for the downstream nanopore sensing of these molecules. We found that samples that had interacted with agarose systematically led to increased pore clogging (both reversible and irreversible with voltage inversion or solution flushing) and additional, low-amplitude blockages in the current signal (as if from contaminants), when compared to control samples that lacked agarose purification (see S5 File). It is therefore important to design any intermediate regions that will not make it into the final product (such as between a restriction site and the end of a PCR amplicon) to be relatively short and thus easily separable.

Fig 1B shows a gel image of a 6216-bp product created with the above protocol (using a λ-DNA template and KpnI/SacI restriction enzymes) that features two (5′, 12-nt long) cohesive ends, as in the example of Fig 1A. The first output lane (labelled “50°C”) shows the result of incubating a sample of this product at intermediate temperatures (50°C), below the melting point of the sticky ends, for 1 hr (~1 nM DNA in 10 mM MgCl₂ + 10 mM Tris, pH 8). Two main populations are visible–a faster band migrating between the 5- and 7-kbp bands of the ladder that is attributed to the linear 6.2 kbp product, as well as a slightly slower-migrating band that is attributed to the two cohesive ends of the product annealing to form a circular molecule. A third, slower band is faintly visible above these two (between the 10- and 20-kbp bands of the ladder) and may represent rare dimers that were formed during the ligation step. In the adjacent output lane (“70°C”), an aliquot of this same sample mixture was briefly heated at temperatures above the melting point of the 12-nt ends (10 min at 70°C) and then immediately submerged in an ice bath. In this case, the “circular” band is greatly reduced in favour of the “linear” band, confirming that the sample contained at least a substantial proportion of products with two active sticky ends–ends that were stably annealed to each other at intermediate temperatures and that could be reversibly dissociated (not locked-in during the ligation stage) with simple temperature cycling, as desired.

Validation

To test the viability of sticky-ended DNA produced by these methods to act as molecular carriers in a nanopore experiment, two product lengths were constructed (1842 bp and 6216 bp) that shared a λ-DNA template, KpnI and SacI restriction enzymes, and a single linker molecule, but that differed in their PCR primers, as outlined in the previous section. The particular linker used here resulted in two complementary ssDNA extensions (5′, 12 nt long) being created on either end of each product–see S1 File in the Supporting Information for the exact oligo sequences (PCR primers and linkers) used.

As proof-of-concept carrier targets, DNA multi-way junctions were utilized in this work. These DNA nanostructures have previously been demonstrated to generate recognizable translocation signals on a nanopore [14, 27, 28]. Here, 12-way junctions (“12-arm stars”) were designed such that each junction arm was 24-bases long, with one of the arms featuring a 12-nt single-stranded extension that is complementary in sequence to a sticky end on the carrier. Incubating a sample of carriers with an excess (≳ 30×) of complementary DNA stars then resulted in the majority of carrier molecules having a target star attached to one end (see S6 File for an example gel shift assay of this attachment step).

A schematic of the nanopore set-up used to detect these carrier-target constructs is illustrated in Fig 2A. Under this scheme, a DNA sample is first diluted in a high-ionic strength buffer (3.6 M LiCl, pH 8) and injected on one side of the pore-containing membrane. A voltage applied across the membrane is then used to transport the negatively-charged molecules though the pore by electrophoresis. As individual molecules approach and translocate though the pore, they block portions of a background flow of ions (Li⁺ and Cl^- in this case) that are also being electrophoretically driven across the membrane. Molecular translocations can thus be detected with high-bandwidth amplifiers through the transient current fluctuations (“events”) they create, signalling the presence of the target molecules in a sample and allowing for their characterization through the details of their current signatures [3, 29].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Validation.

a) Schematic of the experimental set-up for nanopore detection of a DNA nanostructure attached to multiple SE-DNA lengths. 1.8 kbp and 6.2 kbp SE-DNA molecules were first combined with 12-arm DNA stars via the hybridization of complementary single-stranded extensions from each piece. Individual (negatively-charged) hybrid DNA structures were then electrophoretically driven through a nanopore, generating transient fluctuations (“events”) in a background ionic current while they obstruct the pore. b) Sample translocation event (current blockage vs. time) illustrating calculations of the event statistics “max deviation” and “truncated ECD” (TrECD). c) Scatter plot of max deviation vs. log[TrECD] for events from a single nanopore experiment (~13 nm pore diameter, 3.6 M LiCl pH 8 buffer, 150 mV transmembrane potential). Two main peaks are visible in the histograms (along the margins of the scatter plot axes) of both statistics, resulting in four total event populations corresponding to: i) 1.8 kbp SE-DNA on its own, ii) 6.2 kbp SE-DNA on its own, iii) 1.8 kbp SE-DNA attached to a 12-arm star, iv) 6.2 kbp SE-DNA attached to a 12-arm star. Representative current traces of events from each of the four populations are presented in the panels on the right.

https://doi.org/10.1371/journal.pone.0290559.g002

Fig 2B shows a sample current trace of such a translocation event (here resulting from a 12-arm star attached to a 1.8-kbp carrier and passing through a ~13-nm pore), where the amount of ionic current being blocked over time is displayed. Overlaid on top of the raw blockage values (in grey) are a set of fitted step functions (“sublevels”) that relate to different substructures of the molecule passing sequentially through the pore [30, 31]. In this case, the sequence consists of a baseline (open pore) sublevel, followed by two relatively shallow and long-lived sublevels (corresponding to the extended DNA carrier), then two relatively deep and short-lived sublevels (corresponding to the bulky DNA star), and finally a return to baseline. This suggests that this particular structured molecule was initially captured by its carrier end (specifically in a hairpin configuration given the blockage depth of the first level being ~double that of the second [6, 32]) which then threaded through the pore until the bulkier DNA star was reached, resulting in elevated blockage levels above those of the carrier.

Two analysis metrics were employed here to characterize the nanopore signals generated by translocating carrier/DNA star molecules–“max deviation” and “truncated ECD”. Max deviation corresponds to the single point of deepest blockage in the signal and is calculated as the difference (in current) between this point and the averaged baseline values (see Fig 2B). As discussed, for hybridized carrier-target molecules, the point of deepest blockage is expected to occur during the translocation of the bulky star attachment and so max deviation indirectly characterizes its size through how much current is blocked [33]. Truncated ECD (for “equivalent charge deficit”), on the other hand, is the integrated area underneath particular fitted sublevels of the event (in between these levels and the baseline) [21]. Here we specifically use the levels attributed to blockages of the carrier within the overall event (see Fig 2B), in order to characterize the length of these carrier molecules through how much total charge they block. Under nanopore conditions where DNA folding is permitted, ECD has been previously been demonstrated to be a more reliable measure of DNA length when compared to overall passage time [17, 24, 25], since folded sections of the polymer are more compact and so can pass through the pore in shorter times than their fully-extended analogues (in the absence of pore wall interactions).

Fig 2C shows a scatter plot of max deviation vs. truncated ECD (log-scaled) for all carrier translocation events of the same nanopore experiment as the example of Fig 2B (1.8 kbp and 6.2 kbp carriers incubated with 12-arm stars, run though a ~13 nm pore at 150 mV in 3.6 M LiCl pH 8). Excessively short events in duration (< 40 μs) resulting from unsuccessful translocations (“collisions”) or the passage of free DNA stars have been removed prior to plotting to reduce clutter (see S7 File). From either the scatter plot itself or from 1D histograms of each metric plotted on the axis margins, at least four main event populations can be identified that are easily associated with the particular configurations of translocating molecules that produced them. The two peaks in truncated ECD, at fC and fC, correspond to the two different carrier lengths of 1.8 kbp and 6.2 kbp present in the sample. Meanwhile, there is a broad peak (2.9 ± 1.0 nA) at higher values of max deviation and a narrow peak (1.1 ± 0.1 nA) at lower max deviations corresponding to carriers with and without attached stars, respectively. The large spread in max deviation for signals with a star subevent is likely a combined effect of the many possible configurations available to the 12 arms of a translocating star as well as of the variable attenuation of these fast subevents due to the limited bandwidth of the system [27, 33]. Together, the event distributions in max deviation and truncated ECD combine to create four main populations in the scatter plot corresponding to: i) free 1.8 kbp carriers, ii) free 6.2 kbp carriers, iii) star-attached 1.8 kbp carriers, and iv) star-attached 6.2 kbp carriers. A representative current trace from each population is presented in the rightmost panels of Fig 2C.

By examining the relative numbers of events in each of the populations (separated using statistic thresholds, see S7 and S8 Files in the Supporting Information), a few observations can be made. First of all, there are similar numbers of events resulting from 1.8 kbp (3209 events, populations i & iii) and 6.2 kbp (3061 events, populations ii & iv) carriers. This reflects the roughly equal concentrations of both carrier types (~0.4 nM each, as estimated by gel electrophoresis) present in the sample mix introduced to the nanopore. Moreover, within a single carrier type, a majority of events have max deviation values consistent with the attachment of a target star, although the exact ratios differed in the two cases– 58% target-attached (iii) / 42% free carriers (i) for 1.8 kbp and 79% target-attached (iv) / 21% free carriers (ii) for 6.2 kbp. This discrepancy between carrier lengths could result from the particular (and arbitrary) choice of sticky ends used here, specifically each molecule containing a set of cohesive ends that can anneal to each other to form a closed circle. The shorter the contour length of a carrier molecule, the more often these two ends will encounter each other in the random conformations they adopt before a target star can be attached, leading to an elevated proportion of shorter length carriers that are bound in a circular state [34]. Indeed, looking at the current traces of individual events from the free carrier populations (e.g. see panels i and ii on the right of Fig 2C), a large fraction contain only a single blockage level at twice the depth of unfolded DNA, consistent with circularized molecules which can only pass through the pore in an even number of dsDNA fragments (see also S8 File for histograms of average blockage for each population). We note however, that whether this circularization effect is viewed as beneficial or detrimental depends on the specific design and goals of a particular nanopore experiment involving these sticky-ended molecules, and that carriers can easily be constructed with only one sticky end or with two ends of non-complementary sequences using the synthesis methods presented here if circularization is to be avoided.

DNA synthesis

DNA carriers were synthesized as described in the main text (detailed protocols are presented in S2 File). Summarizing, a subsection of a DNA template (λ-DNA, New England Biolabs) was amplified by PCR (LongAmp polymerase, NEB) using specific primer oligonucleotides (Integrated DNA Technologies) to define the amplicon length. The PCR output was then digested by restriction enzymes (KpnI and SacI, NEB) to produce a central fragment with single-stranded overhangs. Finally, DNA linker oligos (IDT) were ligated (T4 ligase, Thermo Scientific) to the ends of this central fragment via its overhangs to create new single-stranded extensions defined by the linker sequences.

For the DNA multiway junctions (“stars”) used as targets for the carriers, these were assembled from individual oligonucleotide strands (IDT), as described previously [27, 28]. Briefly, a number of strands matching the number of arms in the final structure were mixed in equimolar ratios, heated to 95°C, and slowly cooled back to room temperature. The sequences of each strand (48 nt long) are designed to hybridize under these conditions to those of neighbouring strands across two adjacent dsDNA arms (24 bp each). One of the assembled arms is also designed to have a single-stranded extension through which it can be annealed to a complementary extension of the carrier. After assembly, the structures were imaged by polyacrylamide gel electrophoresis, their main gel bands excised, and the purified products re-suspended in solution by electroelution (D-Tube Dialyzer Maxi, EMD Millipore).

Nanopore sensing

Nanopores used in this work were fabricated by the controlled breakdown method (CBD), using previously published protocols [38, 39]. In short, a silicon nitride membrane (#NBPX5004Z-HR, Norcada) is immersed in an electrolyte solution (1 M KCl, pH 8) and a relatively large time-varying voltage (~10 V) is applied across it via Ag/AgCl electrodes, using custom instruments and flow cells similar to products from Northern Nanopore Instruments. A computer-controlled amplifier circuit is used to monitor the current through the membrane and to shut off the applied voltage when a sharp increase in this current is detected, signalling the breakdown of the membrane in a single nm-scale hole, as described in Waugh et al. [39].

Prior to a nanopore sensing experiment, a sample of DNA carriers is incubated with an excess of targets (e.g. ~1.5 nM carriers to ~50 nM stars) to anneal the two pieces together (see S6 File). This annealed mixture is then equilibrated with a high-conductivity sensing solution (3.6 M LiCl, pH 8) and 40 μL injected (at a final carrier concentration of ~0.4 nM) on one side of a nanopore-containing membrane (embedded in a flow cell with freshly-flushed 3.6 M LiCl), where an application of moderate transmembrane potentials (~0.1 V) electrophoretically drives the negatively-charged carrier-target molecules through the pore. A high bandwidth current amplifier (VC100, Chimera Instruments), operating at 1 MHz instrument bandwidth (4.17 MHz sampling rate), is used to detect these molecular translocations as fluctuations in the ionic pore current.

Analysis of the current signals (“events”) generated by translocating molecules is carried out in the Nanolyzer software package (Northern Nanopore Instruments), which applies a digital low-pass filter to the current trace (400 kHz in this work), locates the events within the timecourse, and fits their signals to step functions that characterize the event substructure (see Fig 2B) [30, 31]. Statistics of the analyzed events (e.g. max deviation, average blockage, dwell time, truncated ECD) were calculated either in Nanolyzer itself, or with custom scripts written in Python using the ‘pandas’ library.