|Home | About | Journals | Submit | Contact Us | Français|
Single-stranded oligonucleotides are important as research tools as probes for diagnostics and gene therapy. Today, production of oligonucleotides is done via solid-phase synthesis. However, the capabilities of current polymer chemistry are limited in comparison to what can be produced in biological systems. The errors in synthetic DNA increases with oligonucleotide length, and sequence diversity can often be a problem. Here, we present the Monoclonal Stoichiometric (MOSIC) method for enzymatic DNA oligonucleotide production. Using this method, we amplify oligonucleotides from clonal templates followed by digestion of a cutter-hairpin, resulting in pools of monoclonal oligonucleotides with precisely controlled relative stoichiometric ratios. We present data where MOSIC oligonucleotides, 14–378 nt long, were prepared either by in vitro rolling-circle amplification, or by amplification in Escherichia coli in the form of phagemid DNA. The formation of a DNA crystal and folding of DNA nanostructures confirmed the scalability, purity and stoichiometry of the produced oligonucleotides.
Since the solid-phase synthesis technique was first described by Merrifield in 1963,1 it has been the method of choice for the rapid synthesis of DNA oligomers. Many recent advances in genetics and molecular biology, including for example the recent massive sequencing efforts,2 would not have been possible without DNA oligonucleotides [or oligodeoxyribonucleotides (ODNs)] provided by the solid-phase method.
Although impressive, the capabilities of current polymer chemistry dwarfs in comparison to the sequence controlled polymers that nature can produce.3 The number of errors in synthetic DNA increases with the number of nucleotides and purities of only 70% are common in commercial ODNs comprising 51 bases.4 Post-synthesis purification should in principle help but low purities are reported even for high performance liquid chromatography (HPLC) purified ODNs.5 When very monodisperse ODNs are needed for experiments, the standard method is usually polyacrylamide gel electrophoresis (PAGE). The method is labor intensive in nature, leading to high costs for such ODNs. Moreover, even in experiments where PAGE purified ODNs are used, negative effects attributable to ODN impurities are often observed,6–8 indicating that PAGE purification, at best, provides only an enrichment of the desired product.
Given the quality problems with synthetic DNA oligonucleotides, people have tried alternative paths of production. An attractive path is to use enzymes to produce ODNs. Enzymes have low error rates and can in general copy any given DNA sequence with a high degree of reliability.
PCR has been used to produce DNA oligomers.9–12 One problem is that PCR relies on synthetic ODNs for the primer sequence and every PCR product includes a synthetic ODN at its 5'-end. The primer also needs to be of a certain length, usually above 20 nucleotides, and that extra sequence tends to create purification problems, especially when the ODNs that are being produced are of similar lengths to the primers. In addition, the end product of any PCR reaction is double stranded. Because most applications require single stranded ODNs, primers must be chemically labeled for extraction and to eliminate the unwanted strands, extra steps are needed. The combination of primer removal and complementary strand removal adds costs and complexity and can increase the possibility of errors.
To overcome some of these difficulties, methods using rolling circle amplification (RCA) strategies have been tested.10,13 These methods rely on the ligation of small ODN circles that subsequently act as templates for amplification by a strand displacing phi29 polymerase. The result is a long single strand containing the complement to the circle sequence repeated a large number of times, a so-called tandem repeat. This tandem repeat sequence can be cut into monomers by restriction enzymes to release the desired ODN. Both the PCR and RCA methods outlined above present ways to replicate synthetic ODNs. Errors in the original pool of ODNs will, however, be propagated to the final product and no significant enrichment of sequence purity is achieved.
Here we present a method that is able to produce single-stranded DNA (ssDNA) ODNs directly from clonal templates derived from single colonies of bacteria where the DNA correctness has been verified by sequencing.
In the present Monoclonal Stoichiometric (MOSIC) oligonucleotide production method, the starting point is a list of ODN sequences required by the experiment. A Python script (Supplementary Fig. S1), furnishes an in silico DNA construct, which contains all the desired sequences on a line, separated by digestion hairpins. If the ODNs are required to have a specific molar ratio with respect to each other, this ratio is hard-coded into the construct by repeated insertion of identical ODN sequences in the construct. For example as outlined in Fig. 1a, three ODNs A, B and C could be produced in a 3:3:1 stoichiometric ratio by encoding sequence A three times, sequence B three times and sequence C once. This computed MOSIC pseudogene is subsequently produced by sequence verified gene-synthesis, i.e. assembly PCR14 followed by cloning into E. coli and subsequent repeated iterations of sequencing and mutagenesis until the correct sequence has been obtained as monoclonal plasmid DNA. In the work presented here, we acquired most of the constructs from commercial gene-synthesis service providers (see Methods).
After plasmid DNA containing the MOSIC pseudogene has been obtained, the DNA can be amplified in single-stranded form in vitro by excising the pseudogene followed by re-circularization and nicking to create templates for rolling circle amplification (RCA) (Fig. 1b). Alternatively, the DNA can be amplified by cloning into a phagemid vector and subsequent ssDNA production via helper phage rescue (Fig. 1c). Both processes, (further outlined in Supplementary Fig. S2) furnish ssDNA, which after a digestion step using a type IIs restriction enzyme, releases the MOSIC ODNs.
Using the restriction enzymes BtsCI or BseGI and the hairpin architecture depicted in Fig. 1d, the sequences of the consecutive ODNs can be designed independent from each-other as no base-pairing between the MOSIC ODNs are required for the digestion.
As an initial proof of concept of this technology we decided to hard-code the oligonucleotides used to form a rationally designed, tensegrity triangle, DNA crystal recently published.15 In this work a triangular structure comprising the unit cell was designed to fold from three different sequences in a ratio of 3:3:1, with seven oligonucleotides in total. They form three helices, tailed with short, single-stranded cohesive segments that lead to polymerization and ultimately a 3D DNA crystal. This model was chosen because crystallization generally requires very pure DNA oligonucleotides in a precisely controlled relative stoichiometry. A commercial gene synthesis service provided our designed pseudogene, similar to that schematically depicted in Fig. 1a and it contains the required oligonucleotides with hairpins in between [hereafter referred to as the Crystal Pseudogene (CP)]. The complete sequence can be found in Supplementary Fig. S3.
We amplified single-stranded DNA from the clonal template in two ways according to the method outline above: a) In vitro amplification of the pseudogene using RCA: Here, we produced double stranded plasmid DNA in bacterial culture and purified using a plasmid miniprep kit. Subsequently, we performed four different enzymatic reactions for in vitro amplification: (i) linearization of the CP from the vector by using a non-palindromic restriction enzyme that recognizes a sequence at the ends of the pseudogene and provides sticky ends, (ii) ligation to circularize the CP, (iii) cleavage by a nicking endonuclease that cuts one of the strands providing a free 3´OH end, (iv) rolling circle amplification (RCA) by using the processive and strand displacing phi29 DNA polymerase. b) Growth in E. coli cultures and extraction of helper-phage particles: In this case, large amounts of single stranded DNA can be produced with little effort, however the vector DNA is also amplified along with the pseudogene (Supplementary Fig. S4). In all cases, the resulting ssDNA was digested using BseGI (Fermentas) (Fig. 2). We also tested BtsCI (New England Biolabs) with similar results.
When run on a polyacrylamide gel, the long, ssDNA RCA product, made of the CP repeated in tandem, got stuck into the well (Fig. 2a, Supplementary Fig. S5). As shown in Fig. 2a, the digested pool of ODNs contained the expected length sequences and the hairpins by-product. By measuring the intensity of the gel bands, we verified the 3:3:1 stoichiometric ratio of the ODNs (Supplementary Fig. S6). Moreover, no detectable undigested product seemed to be present. The digested phagemid ssDNA (Fig. 2b) displayed the same product bands as the in vitro MOSIC ODNs but large amounts of higher molecular weight bands were also visible as is to be expected because of the presence of the entire vector ssDNA. Also in this case we verified the desired 3:3:1 stoichiometry (Supplementary Fig. S7). For the crystallization experiment, we were interested to see whether crystals could be formed without any further purification and it was therefore decided to focus on the in vitro variant.
The unpurified, in vitro produced, MOSIC ODNs for the tensegrity crystal were desalted on a Sep Pak C18 column and subsequently lyophilized. A quality comparison with HPLC purified, synthetic, oligonucleotides from a large commercial vendor was also performed by running an analytical HPLC analysis (Fig. 3a, Supplementary Fig. S8) and a denaturing PAGE (Fig. 3b). The HPLC peaks of unpurified MOSIC ODNs, and the corresponding bands in a polyacrylamide gel, appeared sharper than a control of commercial HPLC-purified synthetic oligonucleotides of the same sequence. The apparent polydispersity of the commercial HPLC-purified ODNs was in accordance with previous reports.5
Using a similar protocol to the one described in ref15 we prepared samples for crystallization. After cooling from 60°C to room temperature in a sitting drop plate, we observed rhombohedral-shaped crystals (Fig. 4, Supplementary Fig. S9). We acquired X-ray diffraction data and the structures determined by molecular replacement closely matched (Fig. 4d) with the DNA crystals structures previously published.12,15 Similarly to ref.12, because of the enzymatic production, our data show 5’-end phosphates (Supplementary Fig. S10), lacking in the original published structure derived from synthetic ODNs.15 The resolution obtained in our study using conventional beamlines is slightly lower compared to refs.12 and15. In the latest published work,12 the authors find that 5’-end phosphates improve the resolution of these type of crystals. Our findings do not contradict this, rather, our data is not comprehensive enough to make conclusions whether 5’-end phosphates improves resolution or not.
As a further proof of concept, we proceeded to produce MOSIC ODNs for folding of DNA origami nanostructures. In contrast to the MOSIC ODNs discussed above where in vitro amplification was used, the phagemid production strategy (Fig. 1c) was used in this case.
One of the advantages of using this technology is the possibility to scale up the production of a large number of high quality ODNs. The development of technologies which are based on self-assembled DNA nanostructures and in particular the introduction of DNA origami method involves the use of a large number of oligonucleotides.16–19 In particular certain applications, such as the production of alignment media for protein structure determination20 or the folding of complex DNA nanostructures used for biological applications21–23 require a large number of the same oligonucleotide sequences. With these considerations in mind, the folding of a DNA origami structure was selected as an example to demonstrate the capabilities of phagemid produced MOSIC ODNs.
We designed a honeycomb 10 helix bundle using 180 staple strands. Each of the staple ODNs comprised 42 nucleotides. 72 of the ODNs were encoded on 4 different DNA pseudogene constructs as described above. Each pseudogene contains 18 staples, hereafter referred to as a staple set. We cloned the staple sets into phagemid vectors (pBluescriptIISK(−) or pGEM-Teasy), followed by amplification in E. coli and helper phage rescue (Supplementary Fig. S11). We then digested the resulting ssDNA with BseGI (Fig. 5a) and purified the DNA by PAGE to remove by-products from the vector (Fig. 5b). We subsequently used the obtained staple strands in DNA origami folding reactions.
The 72 strands selected for MOSIC production lie along the length of the tube structure on one side (Supplementary Figs. S12 and S13). Because of this design, structures folded without these 72 ODNs should lack rigidity on one side. Indeed, in transmission electron microscope (TEM) imaging we observed formation of structures that appeared to curl up when excluding these 72 staple ODNs (see Fig. 5d). In contrast, when we added the missing strands, using ODNs produced by MOSIC, the straight original design (Fig. 5c), re-appeared (Fig. 5e, Supplementary Fig. S14), indicating that the staple sets produced by the MOSIC method successfully can replace synthetically produced oligonucleotides also in DNA origami.
The production of long ODNs comprising several hundred nucleotides is currently very challenging via synthetic methods. In the MOSIC method however, very long oligonucleotides can be readily extracted by simply increasing the distance between the cutter hairpins in the clonal substrate. To demonstrate the capability of the MOSIC method in terms of producing long ODNs, we encoded a 378 nt long strand on a clonal template flanked by the same cutter hairpins as described above. Large amounts of 378 nt ODNs could be extracted (Fig. 6a, Supplementary Fig. S15) and compared to the few available sources of synthetic DNA of that length, the method yields ssDNA ODNs with a considerably lower cost (see Supplementary Fig. S15 and discussion below). We demonstrated the long MOSIC ODN functionally by using the 378 nt strands, without further purification, as a scaffold for folding a range of ultra-small DNA-origami structures (Fig. 6b–e and Supplementary Fig. S16–S17). We tested three constructs of varying width, where the 378 nt ODN folded up into short 17-, 9- or 6- helix-bundles. Such small, scaffolded origami constructs have not been demonstrated before and the MOSIC method opens a path to construction of DNA objects with ODNs much longer than the standard oligonucleotides (< 40–60 nt long) normally used for these types of applications. Each ultra-small DNA origami is expected to form a small brick, approximately 10 × 9 × 8 nm in size in the case of the 17 helix bundle. To facilitate TEM imaging, the monomeric structures were designed to polymerize head-to-tail using protruding staple sticky-ends. Indeed, TEM imaging of the folded structures displayed long polymeric nanotubes, Fig. 6c–e. We implemented the 17HB design with a DNA twist of 11 base-pairs per turn, previously reported to increase polymerization efficiency,24 leading to a global right-handed twist of the polymers. The structures display the expected width and internal structure. Additional TEM data can be found in Supplementary Fig. S18. As we demonstrate with this long ODN, the method opens a path to designs that are a hybrid between the recently reported single stranded tile25 and scaffolded-origami16 methods.
DNA oligonucleotides play a crucial role as a molecular tool for a number of important research areas, be it as probes for cell biology and diagnostics, as primers for molecular genetics or as components for DNA nanotechnology and DNA computing. In addition to their use in research, ODNs are also being used to develop therapeutics, with one ODN based drug already approved.26,27 Given the demand for high quality single-stranded DNA oligomers, the reported method solves several problems associated with synthetic ODNs.
Firstly, the MOSIC ODNs are monoclonal, with highly monodisperse products as verified both by our gel results (Figs. 2a, b and 5a, b) and by the analytical HPLC results (Fig. 3). The correct folding of a DNA nanostructure (Fig. 5) and in particular, the crystal structure data (Fig. 4), is a clear evidence of the sequence correctness of the MOSIC ODNs. Here it should be noted that in contrast to the original work15 where two consecutive HPLC purification steps were used to prepare the synthetic ODNs, we were able to get crystallization directly from our pseudogene digest without any further purification except desalting and enzyme removal (see methods). Per-base error rates in MOSIC oligos are estimated to be equivalent to the replication error rates for phi29 polymerase (about 10−5 to 10−6),28 or to the error rates of the bacterial replication machinery (about 10−7 to 10−8),29 for the in-vitro and phagemid variants respectively.
Secondly, MOSIC ODNs can be produced in a precisely controlled relative stoichiometry easily implemented by hard coding the ratios on the pseudogenes. The crystallization, directly from the digest of the pseudogene (Fig. 4), where an exact 3:3:1 ratio is necessary for assembly of the structure,15 is indicative of the capabilities of the MOSIC method in this regard. The stoichiometry control could prove important for small therapeutic DNA structures, such as a recently proposed method for tumorigenic gene silencing.30
As shown above, the method provides the flexibility of choosing whether to amplify in vitro or in E. coli. Although we did not investigate this, it is reasonable to assume that the phagemid variant would result in methylated ODNs when non methylation-deficient bacterial strains are used. Depending on the application, this could be useful or viewed as a disadvantage. The in vitro variant gives a completely clean digest (see Fig. 2) but is slightly more labor intensive and requires the purchase of more enzymes than the phagemid variant.
Synthetic methods for ODN production are severely limited when it comes to the production of very long products. Here, we have presented the production of DNA oligomers ranging from 14 – 378 nucleotides in length, demonstrating that the method is easily applicable for production of ODNs comprising several hundreds of bases. In addition, in the case of long oligonucleotides, the MOSIC method produces ODNs for a significantly lower cost than what is available commercially for synthetic ODNs of the same length, about 15–30 times cheaper for a 378nt ODN (excluding initial cost for gene synthesis of templates, Supplementary Fig. S15), this holds true despite the fact that our calculations are based on small-scale list-prices for all involved enzymes.
Moreover, in contrast to enzymatic production proposed before9–11,13, the MOSIC method does not rely on addition of any synthetic primers, neither for amplification, nor for the digestion. In fact, regardless of whether amplification is done in vitro or in vivo by E. coli/phage, all the components necessary for mass production of MOSIC ODNs, can be readily grown in bacterial cultures. Taken together, the MOSIC method will therefore provide a route for cheap abundant production of oligonucleotides for any application that requires high-quality monoclonal ODNs in a controlled stoichiometry.
We ordered the Crystal Pseudogene (CP) from gene synthesis (MrGene) which provided it directly as a double stranded plasmid construct in a pMK-RQ vector. The CP is enclosed by two restriction sites, CGTCTC, recognized by the restriction enzyme BsmBI, and the circularized form is characterized by a nicking site, GCAATG, recognized by Nb.BsrDI (Supplementary Fig. S19a). The seven oligonucleotides are separated by eight hairpins characterized by a three base loop, a four base GCGC stem, the BseGI restriction site GGATG, and two variable overhanging bases which are complementary to the last two bases of the preceding encoded oligonucleotide (see Fig. 1d).
We made the 378 nt long ODN pseudogene in our laboratory by performing a PCR where pBluscriptSKII(+) as template and two partially homologous primers (Forward: 5’-GGTCTCACATTGCATAATTAACATCCGCGGAACGCGGATGTTCCGGCGTCAATACGGGATAA-3’ Reverse: 5’-GGTCTCCAATGCAGAAGAGCTGCATCCGCGTTCCGCGGATGCACTTTTCGGGGAAATGTGC-3’) were used. To perform the in vitro amplification the pseudogene was enclosed by two restriction sites, GGTCTC, recognized by BsaI, and the circularized form is characterized by two nicking sites, GCAATG and GCTCTTC, recognized respectively by Nb.BsrDI and Nt.BspQI which cut at one base distance on the same strand (Supplementary Fig. S19b). The 378 nt long pseudogene is also characterized by two hairpins like those ones used for the CP.
We designed the 10 helix bundle nanotube using the software caDNAno32 in accordance with the design rules for 3D DNA origami on a honeycomb lattice reported previously.17 The staple strands were sorted into pools of 18 strands resulting in 10 staple sets (Supplementary Fig. S12). Staple sets 1 to 4 were assembled by gene synthesis according to a similar scheme used for the crystal ODNs as described above. The staple set pseudogenes were provided directly from commercial gene synthesis services as a double stranded plasmid construct in a pBluescript SKII(−) vector (Genewiz, staple set 1 and 2) or a pGEM-T easy vector (Bioneer, staple set 3 and 4). The sequences corresponding to the staple strand ODNs were separated by 19 hairpin sequences characterized by a three base loop (GAA), a 3 base CGC stem, the BseGI restriction site GGATG, and two variable overhanging bases which were complementary to the last two bases of the preceding encoded oligonucleotide. The pseudogenes for staple set 1 and 2 were enclosed by the restriction sites G'GATCC, recognized by the restriction enzyme BamHI and T'CTAGA, recognized by the restriction enzyme XbaI.
We inserted the linear CP DNA into a pBluescript SK II(−) for the phage cloning strategy and we used the pMK-RQ containing the CP for the in vitro ampliflication. We cloned the 378 nt long ODN pseudogene into a pDrive cloning vector (QIAGEN, PCR cloning kit). SCS110 or XL10 gold competent cells (Agilent technologies) were thawed on ice. For each reaction we gently mixed and incubated 100 µl of cells with 1.7 µl of β-mercaptoethanol (1.42 M) for 10 min, swirling gently every 2 min. Then we mixed 50 ng of the vector containing the pseudogene or pUC18 plasmid as control with competent cells, incubating for 30 min on ice. Afterwards we heat-pulsed the reactions in a 42°C water bath for 45 s and incubated on ice again for 2 minutes. We then added 0.9 ml of preheated (42 °C) NZY+ broth and we incubated the reactions at 37 °C for 1 h at 250 rpm.
We plated 100 µl of the transformation mixture on LB agar plates containing kanamycin (50 µg/ml) for pMK-RQ transformed cells and ampicillin (100 µg/ml) for the pBluescript SK II(−), pdrive cloning vector and pUC18 plasmid. We incubated the plates at 37°C overnight. The day after we screened single colonies, from pMK-RQ and pBluescript SK II(−) transformed cells, for the CP insert by BsmBI digestion, and by BsaI-HF digestion for the 378 nt long ODN pseudogene. Complete sequence of the 378 nt pseudogene, verified by sequencing, can be found in Supplementary Fig. S20.
We performed transformation of E. coli (SCS110) with staple pseudogenes as described above, plating the transformed cell on LB agar plates containing ampicillin (100 µg/ml) and screening single colonies DNA for the pseudogenes. We used XL10 gold in all cases except when producing phagemid DNA, then SCS110 was used.
We inoculated 5 ml LB containing kanamycin (50 µg/ml) with a single colony of E. coli transformed with the pMK-RQ vector containing the CP pseudogene and incubated overnight at 37 °C, shaking at 250 rpm. We isolated the plasmid DNA from the saturated cell culture by a plasmid mini-prep kit (Omega Bio-Tek), and digested the pMK-RQ vector containing the CP (20 ng/µl) by BsmBI (0.25 U/µl), New England Biolabs), in 1× NEB2 buffer reaction at 55 °C for 2 h, followed by heat inactivation at 80 °C for 20 minutes. We loaded the digestion products on a 1% agarose gel containing ethidium bromide (1 µg/ml, Sigma Aldrich), and purified the linear pseudogene (329 base pairs) by gel extraction (kit from Omega bio-tek) and eluted in 30 µl 10mM Tris-HCl (pH 8.5). We applied the same procedure to get the linear 378 nt long ODN pseudogene but using ampicillin (100 µg/ml) in the 5 ml LB culture and BsaI-HF as restriction enzyme (New England Biolabs). We ligated the linear pseudogene (5 ng/µl) by T4 ligase (0.25 U/µl,Fermentas) in 1× Rapid ligation buffer at 22 °C for 10 min, followed by an inactivation step at 65 °C for 10 min. We nicked the resulting circular pseudogene (1ng/µl) by Nb.BsrDI (0.5 U/µl, New England Biolabs) (with addition of Nt.BstQI at the same concentration for the circular 378 nt long ODN pseudogene) in 1× NEB2 buffer reaction at 65 °C for 2 h; we stopped the reaction by heating at 80 °C for 25 min. We amplified the nicked circular pseudogene (0.25 ng/µl) overnight at 30 °C by rolling circle amplification (RCA) in a single stranded form by phi29 DNA polymerase (0.5 U/µl, Fermentas) in a 1× phi29 reaction buffer reaction (33 mM Tris-acetate, 10 mM Mg-acetate, 66 mM K-acetate, 0.1% (v/v) Tween 20, 1 mM DTT, Fermentas) containing dNTPs mix (1 mM, Fermentas) and T4 gene 32 protein (0.1µg/µl, New England Biolabs). All these reactions (see Supplementary Fig. S2) do not require any desalting steps in between, instead we diluted the solutions from one step to the other adding the new buffer reaction.
In a typical experiment we ligated 100 ng of linear pseudogene in 20 µl, and performed the nicking reaction in 100 µl and the RCA in 400 µl. In the final BseGI digestion step we further diluted the slimy RCA solution into 1.6 ml, producing MOSIC ODNs typically around 12 ng/µl (see “Oligonucleotide production” for BseGI digestion details).
We grew phagemid transformed single colonies in 25 ml LB agar with ampicillin (100 µg/ml) at 37 °C overnight. We then inoculated 4× 300 ml 2xYT media (containing 100 µg/ml ampicillin) with 3 ml of the saturated overnight culture, and added 1.5 ml sterile 1M MgCl2. We incubated the culture at 37 °C and 250 rpm and monitored the OD600 during the cell growth. We added VCSM13 helper phage at OD600 0.45 corresponding to around 2.5 × 108 cells/ml with a multiplicity of infection (MOI) of 20 (phage-to-cells ratio). After further 5 hours we discarded the bacteria by repeating twice a 5000 rpm centrifugation for 20 min at 4 °C. Afterwards we dissolved PEG8000 (40 g/l, Sigma Aldrich) and NaCl (30 g/l, Sigma Aldrich) in the recovered supernatant and incubated it for 1 h on ice. We spun down the phage in this cloudy solution at 10000 rpm for 30 minutes, and after discarding the supernatant we re-suspended the pellet in 6 ml Tris (10 mM, pH 8.5). We centrifuged the collected suspension at 5000 rpm for 5 minutes at 4°C to get rid of any bacterial residue and we transferred the supernatant to a fresh centrifuge bottles. To remove the phage proteins we added a double volume of (0.2 M NaOH, 1% SDS) to the solution, and after gently swirling we incubated at room temperature for 3 min. We added a one and a half volume of (3 M KOAc pH 5.5 titrated with glacial acetic acid), we gently mixed the solution by inversion and we incubated it in ice cold water for 10 min. After centrifuging at 16500 g for 30 min at 4 °C we decanted the supernatant to a fresh bottle and added at least two volumes of 100% ethanol. After 1 hour incubation in ice cold water, we performed another centrifugation at 16500 g for 30 min. The obtained single stranded DNA (pellet), was washed with 70% ethanol and re-suspended in 2 ml Tris (10 mM pH 8.5) and analyzed by agarose gel (Supplementary Figs. S4 and S11).
We digested the RCA product (up to 30 ng/µl) and the DNA from bacterial production (52 ng/µl) by BseGI (0.5 U/µl, Fermentas) in 1× Tango buffer by incubating at 55 °C for 24 h followed by heat inactivation at 80 °C for 20 min. We analyzed the digestion products by PAGE (20 % polyacrylamide, 20 % formamide, 8M urea mixed in 1× TBE which is made of 89 mM Tris-borate, 2 mM Na2EDTA dissolved in deionized water) at 180 V for 1 h (Fig. 2a–b) and for the 378 nt long ODN pseudogene by PAGE (5 % polyacrylamide, 20 % formamide, 8M urea mixed in 1× TBE), at 180 V for 35 min (Fig. 6a). We stained the gels with SYBR Gold (1X, Invitrogen) for 15 min and acquired images by UV trans-illumination (UVITEC) and analyzed by the software ImageJ. We also analyzed the MOSIC CP oligonucleotides by HPL-Chromatography (HPLC) and we compared them with the synthetic corresponding oligonucleotides. In detail, we performed analytical reverse phase separations of oligonucleotides on an Agilent 1100 system. We used a C18 column (X bridge, 3.0 × 50 mm) from Waters in a system of buffer A (10 mM NH4HCO3 in H2O, pH 10) and buffer B (MeCN) at a flow of 1 ml/min (buffer gradient from 1% till 11% buffer B in 4 min).
We desalted oligonucleotides from the RCA digest, after StrataClean Resin (Agilent) extraction to remove proteins, by Sep-Pak C18 cartridges (Waters) following the manufacturer’s instructions. We subsequently lyophilized them, and re-suspended them in crystallization buffer. Crystals were grown from 50 µl sitting drops in a thermally controlled incubator containing 250 ng/µl DNA, 30mM sodium cacodylate, 50 mM magnesium acetate, 50 mM ammonium sulfate, 5 mM magnesium chloride and 25 mM Tris (pH 8.5), equilibrated against a 1 ml reservoir of 1.7 M ammonium sulfate. Rhombohedral-shaped crystals with dimensions between 70 to 150 µm were obtained by slow annealing, in which the temperature was decreased from 60 °C to room temperature (~20 °C) with a cooling rate of 0.2 °C per hour over a period of 8 days, during which the volume of the drop diminished by about 90 %. Crystals were obtained at the end of the cooling step, and appeared full-sized within a day. We transferred the crystals to a cryosolvent of 30 % glycerol, 100 mM ammonium sulfate, 10 mM MgCl2 and 50 mM Tris and froze them by immersion into liquid nitrogen.
We collected a native dataset to 4.79 Å resolution from a single crystal at BESSY BL14-1 in Berlin33. (We performed a total of 3 crystallization droplet experiments, each one gave ample amounts of crystals. In total, we collected three datasets from 3 harvested crystals one from BESSY BL14-1 and two from Diamond I02, not used in this study.) Our MOSIC DNA oligonucleotides crystallized in space group H3,34 with cell parameters a = b = 106.4 Å and c = 95.15 Å (PDBid=4B8D) while the solid-phase produced oligonucleotides with identical DNA sequence used for the first tensegrity triangle crystal structure15 (PDBid=3GBI) also crystallized in H3 with almost identical cell parameters, a = b = 107.2 Å and c = 93.14 Å. The 4.79 Å dataset included 1980 reflections and the structure was re-determined by molecular replacement using 3GBI15 as search model and MOLREP integrated in the CCP4i software package35. Molecular replacement parameters in Supplementary Table 1. After initial refinement of the 3GBI model difference density for the 5-prime phosphate of our MOSIC made DNA oligonucleotides was visible in our 4.79 Å difference map (Supplementary Fig. 10) and the 5-prime phosphate was subsequently included in our model being the sole addition made to the 3GBI template model. During refinement in autoBUSTER36 the crystallographic B-factor modeling disorder was kept constant for all atoms of the model as the 4.79Å resolution data does not permit refinement of individual B-factors. The resulting DNA tensegrity model contains 866 nucleic acid atoms refined to crystallographic R/Rfree factors of 0.185/0.205 respectively. The asymmetric unit was kept the same as during 3GBI refinement and our structure does not differ significantly from the original 3GBI structure when superimposed (Fig. 3d). We generated all the molecular model figures using PyMOL37.
We purified staple sets 1 to 4 for folding of 10 helix bundle by PAGE (20 % polyacrylamide, 20 % formamide, 8M urea). The oligonucleotides were washed out of the gel with water, then desalted by Sep-Pak C18 cartidges (Waters), we lyophilized and re-dissolved them in water.
We prepared each sample by combining 5 nM scaffold (p8064), 25 nM of each staple oligonucleotide, buffer and salts including 5 mM Tris, 1 mM EDTA (pH 7.8 at 20°C), and 10 mM MgCl2. For the staple strands produced by MOSIC the concentration was raised to 75 nM. We carried out the folding by rapid heat denaturation followed by slow cooling (80 to 60 °C in 20 min, followed by cooling from 60 to 24 °C over 14 h).
We spotted 5 µl of the folding reaction on glow discharged, carbon-coated Formvar grids (Electron Microscopy Sciences), incubated them for 20 s, blotted off, and stained with 2% (W/V) aqueous uranyl formate solution. We performed EM analysis using a FEI Morgagni 268(D) transmission electron microscope at 80 kV with nominal magnifications between 28000 and 44000. We recorded images digitally using the Advanced Microscopy Techniques Image Capture Engine 5.42.
After StrataClean Resin (Agilent) extraction, we desalted the 378 nt long oligonucleotide from RCA digested product by Sep-Pak C18 cartridges (Waters) following the manufacturer’s instructions, we lyophilized, and re-suspended it in deionized water to be used as scaffold for ultra-small DNA origami bricks. We performed the folding reaction by combining 0.1 µM of the long strand with 0.5 µM staple oligonucleotides (including protruding ones for the polymerization of the origami bricks), buffer and salts including 5 mM Tris, 1 mM EDTA (pH 7.8 at 20°C), and 10 mM MgCl2. We carried out the folding by rapid heat denaturation followed by slow cooling (80 to 60 °C in 20 min, followed by cooling from 60 to 24°C over 14 h).
We spotted 3 µl of the folding reaction on glow discharged, carbon-coated Formvar grids (Electron Microscopy Sciences), incubated them for 20 s, blotted off, and stained with 2% (W/V) aqueous uranylformate solution. We performed EM analysis as described above.
The project was funded by the Swedish Research Council (Vetenskapsrådet) through a repatriation grant and a project grant to BH (Grants 2010-6296 and 2010-5060). BH is a recipient of an assistant professorship with startup funding funded by Carl Bennet AB, Karolinska Institutet and Vinnova. We thank Shawn Douglas for help with the 10HB design. Further, we thank the Laboratory of Chemical Biology at KI for HPLC support, Tobias Karlberg and the Protein Science Facility, psf.ki.se (Dept of Medical Biochemistry and Biophysics, Karolinska Institutet), for crystallography support, Monika Schultz and Camilla Sandén for help with pseudogene cloning and enzymatic reactions for the long ODN, and Prof. Ned Seeman for fruitful discussions.
Author ContributionsCD, CK and BH contributed to experiments. MM contributed to the crystallography experiments. BH conceived the method principle and BH and WMS contributed to the method design. CD and CK contributed to method development and implementation. All authors contributed to figure production and manuscript writing.
Competing financial interests
BH is the co-founder and CEO of a recently started company (Basestack Labs AB) that will commercialize some applications of the presented methods.