Silver-mediated base pairings: towards dynamic DNA nanostructures with enhanced chemical and thermal stability

The thermal and chemical fragility of DNA nanomaterials assembled by Watson–Crick (WC) pairing constrain the settings in which these materials can be used and how they can be functionalized. Here we investigate use of the silver cation, Ag+, as an agent for more robust, metal-mediated self-assembly, focusing on the simplest duplex building blocks that would be required for more elaborate Ag+–DNA nanostructures. Our studies of Ag+-induced assembly of non-complementary DNA oligomers employ strands of 2–24 bases, with varied base compositions, and use electrospray ionization mass spectrometry to determine product compositions. High yields of duplex products containing narrowly distributed numbers of Ag+ can be achieved by optimizing solution conditions. These Ag+-mediated duplexes are stable to at least 60 mM Mg2+, higher than is necessary for WC nanotechnology schemes such as tile assemblies and DNA origami, indicating that sequential stages of Ag+-mediated and WC-mediated assembly may be feasible. Circular dichroism spectroscopy suggests simple helical structures for Ag+-mediated duplexes with lengths to at least 20 base pairs, and further indicates that the structure of cytosine-rich duplexes is preserved at high urea concentrations. We therefore propose an approach towards dynamic DNA nanomaterials with enhanced thermal and chemical stability through designs that combine sturdy silver-mediated ‘frames’ with WC paired ‘pictures’.


Introduction
The thermal, chemical and mechanical fragility of DNA nanomaterials assembled by Watson-Crick (WC) pairing constrain how these structures can be functionalized and the settings in which they can be used. For example, WC-paired linker stability and charge repulsion limit the achievable packing densities of metal nanoparticles on DNA origami [1,2], and the delicacy of tile-assembled DNA nanotubes leads to tube-opening ruptures in AFM imaging [3,4]. Chemical compatibility is another limiting factor: denaturing conditions are inhospitable to WC paired structures, while dilution on repeated operation limits the performance of dynamic DNA nanostructures that employ strand displacement reactions on toehold designs [5,6]. New approaches to DNA self-assembly and dynamic manipulation of DNA structures are needed to overcome these types of limitations.
Due to the special interactions of silver (I) cations with the nucleobases [7][8][9][10][11], DNA assemblies that incorporate Ag + provide a possible route to more robust, self-assembled DNA materials that can survive in new settings. Silver is quite unusual amongst metal cations for its specific interaction with the nucleobases, rather than the sugar-phosphate backbone (Hg 2+ also associates exclusively with the bases, but its toxicity makes it undesirable for general self-assembly schemes) [7,10]. This special property of Ag + has already been exploited for metallization of DNA by chemical reduction after the addition of Ag + to DNA solutions [12,13]. As discussed in recent review articles [14,15] much of the current effort on metal-DNA assemblies is focused on fluorescent silver clusters whose colors are selected by the DNA oligomer template [16][17][18]. In such 'Ag N -DNA,' roughly half the silver content is neutral, presumably forming the cluster core, with the remainder as cations (Ag + ) that may serve to attach the cluster core to the bases [19,20]. Thus base-Ag + interactions appear to be a key feature of fluorescent Ag N -DNA and are likely to be generally important in other settings involving metallic silver in intimate contact with DNA.
Rather than as a step towards metallization, here we are primarily interested in the potential of Ag + for selfassembly of new, more robust types of DNA nanostructures. Ag + incorporation has already been used to increase the melting temperature of canonical duplexes with sparse insertions of C-C mismatches [21,22]. Ag + is also known to assemble homobase DNA duplexes via non-canonical, Ag + -bridged pairings of guanine (G) [23] or cytosine (C) bases [23][24][25]. Recent quantum chemical calculations found high binding energies for C-Ag + -C and G-Ag + -G pairs, roughly four times larger than for native C-G base pairs [23]. These prior findings suggest that Ag + -bridged pairings of the natural bases may provide a route to more robust DNA nanostructures, without the need for expensive artificial bases and/or backbone modifications. Important unresolved issues for using Ag + -mediated self-assembly in DNA nanotechnology include the types of base motifs suitable for such metal-mediated pairings and the length of Ag + -paired regions needed for stability. Currently, only two motifs that form Ag + -paired duplexes have been identified, specifically homobase runs of C bases and homobase runs of G bases [23,25]. If this were the complete set of base motifs capable of forming Ag + -assembled duplexes, it would not be possible to go beyond the simplest silver-assembled structures. Thus the usefulness of Ag + in future DNA nanotechnology hinges on the yet-to-be established existence of sufficiently diverse base motifs for Ag + -paired structural elements to enable low-error assembly at an interesting degree of structural complexity. Achieving sufficient yields of designed products will also require identification of solution conditions that realize planned Ag + strand pairings while suppressing undesired byproducts. Because combining Ag + -mediated and WC pairings would maximize design flexibility, it is important to identify solution conditions compatible with sequential stages of Ag + assembly and WC assembly. In addition, Ag + pairings may permit DNA scaffolds to have spatial resolutions finer than the 10-15 base pair 'sticky ends' currently used in dynamically responsive, WC paired DNA structures [26]. Thus it is important to determine whether Ag + -paired motifs can remain robust at significantly fewer paired bases than required for stable WC pairings.
Here we select simple base motifs that enable a first study of these issues (table 1). To explore the length of Ag + -paired regions needed for stability, we study short strands of all C and all G bases. To examine the issue of motif diversity, we employ length 11 oligomers with a central heterobase within an all C or all G strand context. At the outset, it is unclear whether such heterobase crossovers will be destabilizing, by inhibiting Ag + incorporation; or stabilizing, by enhancing Ag + incorporation. (This issue has not been explored previously in experiment or theory, and is not amenable to simple predictions due to the wide diversity of possible binding geometries [23,27].) To determine whether stable duplex formation can be preserved by short C and G repeat motifs, we study 12 base strands with a 4 base separation between two, non-adjacent base crossovers. To examine the effects of a higher crossover content, we study one exemplary strand with 5 heterobase crossovers. We use circular dichroism (CD) spectroscopy to probe for changes in Ag + -paired structure relative to the homobase context. As a testbed for developing solution conditions that select for formation of duplex structural motifs (in preference to strand monomer byproducts), we use 20-24 base strands that also enable tests of the structural stability of Ag + -bound duplexes in the presence of additional salts required by DNA nanotechnology schemes such as origami. This is important for determining whether future self-assembly schemes could combine sequential Ag + -mediated and WC mediated stages. These longer strands additionally serve to probe the stability of silver-mediated pairings in the presence of a chemical denaturant, urea, which has interesting potential for suppressing WC pairing during Ag + -mediated assembly stages for future strand sets designed to form structures that incorporate both types of pairings.
Here we find that widely varying C-and G-rich base motifs can be paired by Ag + . Under optimized solution conditions, Ag + -mediated assembly results in duplex products with narrowly distributed numbers of attached Ag + , indicating suitability as building blocks in more elaborate structures. CD spectroscopy suggests simple helical structures for both homobase and heterobase Ag + -mediated duplexes, although base mutations inserted between long homobase G runs may produce a different secondary structure. For C-rich strands, high stability is exhibited in urea and with added salts, while G-rich sequences display more varied behaviors. We propose simple prototype assemblies to indicate how Ag + could be used in dynamic DNA structures that preserve strand content by combining both WC and Ag + -mediated pairings.

Methods
All DNA oligonucleotides were synthesized by Integrated DNA Technologies with standard desalting and additionally desalted by spin filtration. Strand G 20 contains the longest series of guanine bases provided at this time by at Integrated DNA Technologies as a standard order. All mass spectra show 90+% purity. All solutions were prepared with nuclease free water (Integrated DNA Technologies) except for the HPLC running buffers which used 18 MΩ MilliQ water. The AgNO 3 was analytical grade, 99.999% purity (Sigma-Aldrich). For preparation of any solution containing DNA for instrumental analysis, the solution was annealed at 90°C for 5 min and cooled slowly to room temperature.
CD spectra were measured on an Aviv 202 circular dichrometer. Multiple sample spectra and multiple buffer blank spectra were separately averaged to enable accurate background subtraction. All spectra were measured on pH 7 solutions, buffered by the concentrations of ammonium acetate specified below.
High-performance liquid chromatography (HPLC) was performed on a Hitachi L-6200A pump and L-4200 UV/vis detector using a pH 7 water-methanol gradient from 15% to 40% methanol at 1% per minute, with 35 mM triethylamine acetate as the ion-pairing agent to the C18 column (50×4.6 mM Kinetex EVO with 2.6 μm particle size and 100 Å pore size).
Solutions for mass spectrometry were injected into a Waters QTOF2 mass spectrometer at 10 μl min −1 in negative ion mode with a 2 kV capillary voltage, 45 V cone voltage and 14 V collision energy. All solutions were injected with 10 mM or 50 mM ammonium acetate buffer as specified (pH 7). Prior to injection, samples collected from HPLC were solvent exchanged into 10 mM ammonium acetate by spin filtration.

Results and discussion
Predictable formation of all but the simplest Ag + -paired DNA structures will require diverse base motifs that produce well-defined Ag + -mediated pairings. Therefore the distribution of Ag + -bound products formed by multi-base motifs is a crucial issue. Here we use negative ion electrospray-ionization mass spectrometry (ESI-MS) [23] to identify the composition of the Ag + -DNA products that form on oligonucleotides with wideranging lengths and compositions (table 1).
In ESI, non-equilibrium ionization processes involving a series of Coulomb explosions ultimately yield the desolvated ion that is detected in MS. In principle this can result in product distributions that differ from those in solution. However due to the gentleness and ease of ionizing DNA in negative ion mode, ESI-MS has long been used to detect delicate DNA bindings, including non-covalent attachment of molecules to DNA [28] and for investigation of binding stoichiometries to DNA [29]. Gas phase ESI-MS data for binding stoichiometry studies has also been experimentally shown to agree with high fidelity to solution phase data [30], though some fragmentation of canonically paired DNA oligomers may occur depending on the instrumental conditions employed [31,32]. Due to the considerably higher strength of Ag + -mediated pairings than WC pairings [23], we expect minimal fragmentation for these non-canonical, metal-bridged products and gas-phase distributions that are good approximations to solution phase abundances. (The exceptions would be any oligonucleotides that produce products with high degrees of mass heterogeneity, for which widely varying ionization efficiencies can mask true abundances. In addition, sufficiently high mass Ag + -DNA aggregates may not support high enough ionization rates to be detected within the range of instrument sensitivity.) The high resolution of the ESI-MS instrument used here generally produces mass accuracies comparable to or better than a third of a proton, or ∼0.3% the mass of a silver atom, allowing facile differentiation of the number of Ag + bound to different Ag + -DNA products. (We note that mass accuracy differs slightly depending on charge state and quality of calibration over the relevant region of the m/z axis.) Figure 1(a) shows a typical mass spectrum of Ag + -DNA in this study. The overall peak width is set by the distribution of elemental isotope abundances in the DNA strands and silver as well as the charge state. These natural abundances are well characterized, enabling accurate prediction of the isotope peak envelope for any given product. In cases where the product counts are high enough, the individual isotope peak fingers are clearly resolved ( figure 1(b)).
Negative ion mode ESI produces deprotonated DNA gas phase ions. If no other ions are bound to the DNA, the negative charge state of the detected ion, z, is equal in magnitude to the number of deprotonated sites on the DNA. When there are additional positive charges embedded in the DNA, such as Ag + , more deprotonated sites are required to achieve the same value of z. Therefore, the high mass accuracy of the ESI-MS instrument enables determination of the net positive charge associated with the total silver content through fits to the isotope peak envelope, which shifts by 1.007/z for each proton removed [19,20,33]. (However if the heterogeneity of products is too high, fitting becomes infeasible due to overlapping signals.) Our previous study of Ag + -DNA solutions formed on homobase oligomers found that negative ion ESI left all of the silver content in cationic (Ag + ) form, a finding that carries over to the heterobase strands studied here [23].
Previous studies found that Ag + causes self-assembly of C and G homobase duplexes [23], but does not form A or T homobase duplexes. Because these studies did not investigate mixed base building blocks for Ag + assembly, it is currently unknown whether mixed base strands can form Ag + -paired duplexes or whether mixed base composition perturbs the Ag + -duplex structure to the extent that it is no longer a well-defined building block. The ability to specifically control where the Ag + -base pairs are formed is also an important aspect of designing more complex structures that can self-assemble in an organized manner. To investigate these issues, we explore the duplex products produced by Ag + -assembly of DNA oligonucleotides with mixed base Figure 1. (a) Isotope peak envelope in a typical mass spectrum (red curve) for an Ag + -DNA product at charge state z=−6, for a solution containing the strand CG 3 C 3 G 2 C 2 G at 1 Ag + /base. Black curve: the Gaussian fit to the calculated isotope peak envelope, which determines the experimental mass to be 9669.59 amu, within a fraction of a proton to the theoretical mass of 9669.94 amu for a product which contains two DNA strands and 22 silver cations. (b) With sufficient signal, individual isotope fingers are clearly resolved within the isotope peak envelope. compositions using ESI-MS, at the same 1 Ag + /base concentration ratio as in previous studies of homobase duplexes [23]. Strand monomer products are also detected (full mass spectra are shown in the SI), in most cases as minor products in comparison to duplexes. Because we show below that salt conditions in solution can be manipulated to eliminate these one-strand byproducts, here we focus on the 'duplex' products with Ag + bridging two strands.
We first examine the consequences of replacing the central bases in C 11 and G 11 strands with heterobase mutations. Due to the distinct silver binding sites and strengths on the different bases [23,27], this heterobase inclusion could be destabilizing for Ag + -mediated pairings (which we expect to reduce Ag + incorporation), or stabilizing (which we expect to increase Ag + incorporation). Figure 2 displays the distribution of the numbers of Ag + in duplex products for C 11 and G 11 , together with their daughter strands containing central single base mutations (full mass spectra are displayed in figures SI.1 and SI.2). We use the integrated area of Ag + -duplex product peaks in the mass spectrum as a semi-quantitative measure of the relative abundances of each Ag +duplex species. We find a narrow distribution of Ag + /duplex for the C 11 DNA strand (figure 2(a)), with an abundance-weighted Ag + stoichiometry of 11.0 Ag + /duplex. This binding stoichiometry corresponds to one Ag + atom bridging each C-Ag + -C base pair in the duplex, suggesting a simple helical structure, and agrees with previous studies [23]. Upon mutation of the central cytosine to A and G bases (figures 2(b) and (c)), the Figure 2. Distribution of the number of attached Ag + in duplexes formed by the strands (a) C 11 , (b) C 5 AC 5 , (c) C 5 GC 5 , (d) C 5 TC 5 , (e) G 11 , (f) G 5 AG 5 , (g) G 5 CG 5 and (h) G 5 TG 5 at 80 μm DNA and 1 Ag + /base in 50 mM ammonium acetate. The relative abundances represent the integrated area of the Ag + -duplex product divided by the total integrated area of all Ag + -duplex products in the mass spectrum. The abundance-weighted Ag + per duplex for (a) C 11 is 11.0 Ag + and (e) G 11 is 10.8, corresponding closely to 1.0 Ag + per base pair, as marked by the gray dashed line. (b)-(d) Single base mutations to the central base in C 11 produce somewhat wider distributions that are only slightly shifted. (f)-(g) Single base mutations to the central base in G 11 produce significantly higher incorporation of Ag + . distribution of Ag + -mediated duplex products becomes wider, but the abundance weighted Ag + stoichiometry is nearly unchanged (10.8 and 11.2 Ag + per duplex, respectively). The wider distribution of products with more than 11 Ag + /duplex likely corresponds to structural perturbations of the Ag + -mediated duplex near the mutation site. (The wider distribution could instead reflect formation of single-stranded overhangs in the presence of the central base mutations. However, we expect low yields of such products because quantum chemical calculations found 50-60 kcal mole −1 higher binding energies for Ag + that bridge between two bases, as G-Ag + -G and C-Ag + -C, than for Ag + bound to individual bases, as G-Ag + and C-Ag + [23,27].) CD spectroscopy at UV wavelengths is quite sensitive to relative base orientations, and thus serves to probe structural similarity amongst strand assemblies with similar base compositions. The CD spectra for C 5 AC 5 , C 5 GC 5 and C 5 TC 5 all closely resemble that for C 11 ( figure SI.3(a)). Because duplex products dominate the integrated intensities in the mass spectra ( figure SI.1), we expect the CD spectra to be dominated by the corresponding duplex structures. We conclude that duplexes are similarly structured across C 11 and all of the mutated C 11 strands. (Although the mass spectrum for C 5 GC 5 , shown in figure SI.1(c), exhibits a higher degree of strand monomer products, the CD is also similar to that for C 11 , with a reduced amplitude that is consistent with the same underlying duplex structure, diluted spectrally by the higher presence of strand monomer products.) We also note that the T mutation increases the abundance weighted Ag + stoichiometry to 11.8 Ag + / duplex ( figure 2(d)). This is significantly higher than the value of 10 Ag + /duplex we had naively expected based on the low affinity of thymine homobase strands for Ag + [23,34]. The incorporation of ∼2 additional Ag + and the narrower Ag distribution for the T mutation of C 11 , in comparison to the A and G mutations, suggests that the CTC motif is especially stabilized by additional Ag + incorporation, and may be particularly useful for lowerror assembly of Ag + -mediated DNA nanostructures.
The central single base mutations in G 11 strands introduce larger differences in Ag + distributions (figures 2(e)-(h)). For G 11 , Ag + -duplex products are narrowly distributed around an abundance-weighted Ag + per duplex of 10.8 Ag + , consistent with 1 Ag + bridging each base pair (figure 2(e)), just as for the case of C 11 ( figure 2(a)). However G 5 AG 5 , G 5 CG 5 and G 5 TG 5 exhibit significantly higher abundance-weighted Ag + per duplex (16.2, 18.1 and 18.1 in figures 2(f)-(h), respectively, corresponding to 1.47, 1.65 and 1.65 Ag + per paired base). Because the full mass spectra are heavily dominated by duplex products (figure SI.2), the CD spectra should reflect the duplex structure. The pronounced shift to higher Ag + content in strands with central base mutations (figures 2(f)-(h)) is accompanied by small shifts in wavelengths of CD spectral features but more significant shifts in CD amplitudes ( figure SI.3.(b)). Thus it is unclear whether the additional~7 Ag + introduced by the mutations to G 11 (figures 2(e)-(h)) are held near the mutation, retaining an overall similar duplex structure; or if the central base mutations produce a different type of structural motif.
To test whether shorter homobase C and G runs can be combined into larger motifs that preserve overall duplex structure, we used the sequences C 4 G 4 C 4 and G 4 C 4 G 4 . We selected these palindromic sequences because, although calculations suggest a parallel strand orientation for Ag + mediated pairings [23], this has not been definitively established by published experimental results. In particular, if Ag + were to assemble duplexes of antiparallel strands, then non-palindromic sequences would not be self-complementary for Ag + -mediated duplex formation. We avoid this issue by using C 4 G 4 C 4 and G 4 C 4 G 4 , which are complementary for Ag + pairing regardless of strand orientation (because Ag + preferentially pairs C to C and G to G, rather than C to G [23]). The abundance weighted Ag + stoichiometry was 17.4 and 20.1 Ag + per duplex, respectively (figures 3(a) and (b)), corresponding to 1.45 and 1.68 Ag + per paired base, similar to the additional Ag + incorporated by single base mutations to G 11 . Full mass spectra are shown in figure SI.4(a) and (b).
The CD spectra for these palindromic strands, C 4 G 4 C 4 and G 4 C 4 G 4 , suggest that the duplex structures are similar despite their quite different C, G content ( figure 3(d)). To represent the case of no structural perturbation from heterobase inclusion, we used CD spectra for Ag + -assembled cytosine homobase duplexes and Ag +assembled guanine homobase duplexes. These exhibit almost entirely negative CD over the spectral range studied (figures 5(a) and (b), red traces), in striking contrast to CD spectra of WC duplexes or quadruplexes [23]. We formed linear combinations of these homobase duplex spectra with weights chosen to match the G, C content in the palindromic mixed-base sequences ( figure 3(e)). Using 33% weighting of the cytosine duplex component and 66% weighting of the guanine duplex component (green curve, figure 3(e)) gives good agreement with the measured CD signal for Ag + -paired G 4 C 4 G 4 (green curve, figure 3(d)) . Similarly, using 66% weighting of the cytosine component and 33% weighting of the guanine component (blue curve, figure 3(e)) gives good agreement with the measured CD signal for Ag + -paired C 4 G 4 G 4 (blue curve, figure 3(d)). The agreement of the spectral shapes from these linear combinations of homobase duplex spectra (figure 3(e)) with the measured CD spectra for the palindromic strands C 4 G 4 G 4 and G 4 C 4 G 4 ( figure 3(d), blue and green curves) indicates that the underlying secondary structure is preserved, despite the presence of additional bound silver associated with the heterobase crossovers. This agreement also suggests that the strand orientations favored by homobase runs of C-Ag + -C and G-Ag + -G pairings may be the same. If the Ag + -paired C 4 G 4 C 4 and G 4 C 4 G 4 duplexes instead had different secondary structures, we would not expect both CD spectra (figure 3(d)) to agree with the weighted linear combinations of homobase duplex components in figure 3(e).
It appears that the major effects of the heterobase crossovers in these palindromic sequences is incorporation of additional Ag + , without major disruption to the underlying duplex structures of the constituent G homobase and C homobase components. To test the effects of increasing the number of heterobase crossovers we use the strand CG 3 C 3 G 2 C 2 G (5 heterobase crossovers). Consistent with the expectation that base crossovers incorporate additional Ag + , the abundance weighted Ag + stoichiometry rose further, to 22.6 Ag + per duplex (figure 3(c)), or 1.88 Ag + per paired base. (Full mass spectra are shown in figure SI.4.) The CD spectrum for CG 3 C 3 G 2 C 2 G shows the same overall peak structure as for C 4 G 4 G 4 and G 4 C 4 G 4 , albeit with reduced amplitude (figure 3(d), orange dashed curve). This is consistent with preservation of the same underlying duplex structure, given that heterogeneous dipolar couplings introduced by base crossovers can also affect CD amplitudes.
Overall, results from CD spectra and silver binding stoichiometries suggest a fairly simple structural picture in which the underlying duplex motifs from Ag + pairings are stable in the presence of heterobase crossovers between C homobase motifs with lengths of up to 5 or more bases, and G homobase motifs with lengths of up to 4 bases. In all cases, heterobase inclusion within C, G rich strands appears to be stabilizing, as indicated by the increase in numbers of incorporated Ag + .
In order to create high yields of larger nanostructures from smaller duplex subunits, the subunits should form with minimal polydispersity. In particular, it is important to minimize Ag + -bearing strand monomer products, such as those identified previously for cytosine homobase strands [23]. These prior studies were made in low ionic conditions. Investigating whether increased ionic strength could improve Ag + -paired product homogeneity, as is the case for WC mediated assemblies, requires an alternative assay to ESI-MS, which is incompatible with solutions of high ionic strength. Therefore we turn to high performance liquid chromatography (HPLC) with inline absorbance detection to monitor the retention times of the various Ag + -DNA products formed in the presence of additional salts. To test this approach we focus on the strand T 2 C 20 T 2 . We select the 24 base length because suitable solvent gradients in HPLC produce well-separated peaks for strand monomer versus strand dimer products. This is because for the ion-pair reverse-phase HPLC with a C18 column used in this study, the retention time of the DNA products depends most strongly on the amount of column-accessible phosphate backbone, and to a lesser extent the hydrophobicity of the nucleobases [35,36]. Therefore more rigid, compact structures with less exposed backbone will elute earlier.
The HPLC chromatogram for T 2 C 20 T 2 annealed in low ionic conditions (10 mM ammonium acetate) displays a high degree of heterogeneity in peak location and amplitude ( figure 4(a)). The two dominant peaks ( figure 4(a)) have large separation in retention time. The aliquots corresponding to these dominant peaks were Figure 3. Distribution of Ag + binding stoichiometry in Ag + -duplexes of 12 base C, G strands, and corresponding CD spectra. (a) C 4 G 4 C 4 , (b) G 4 C 4 G 4 , and (c) CG 3 C 3 G 2 C 2 G at 80 μm DNA, 1 Ag + /base and in 10 mM ammonium acetate. Abundance-weighted Ag + per duplex is 17.4, 20.1 and 22.6 for C 4 G 4 C 4 , G 4 C 4 G 4 and CG 3 C 3 G 2 C 2 G respectively. Overall binding is much higher than expected for 1 Ag + per C-Ag + -C or G-Ag + -G base pair (dashed gray line). (d) Circular dichroism spectra of the strands at 7.5 μM and 1 Ag + / base exhibit signals similar to (e) linear combinations of normalized C 20 and G 20 spectra at 1 Ag + per base. collected and identified through ESI-MS (full mass spectra are shown in figure SI.5). The early-eluting peak is comprised of strand monomers with 8 and 9 attached Ag + (left inset, figure 4(a)). The late eluting peak is comprised of Ag + -bound duplexes with 19 and 20 attached Ag + (right inset, figure 4(a)). That the Ag + -mediated duplexes elute last is expected because the length of column-exposed backbone is greater than for strand monomer products.
Annealing at higher ionic conditions, 500 mM ammonium acetate ( figure 4(b)) or 60 mM MgSO 4 ( figure 4(c)) produced a much cleaner chromatogram with one dominant absorbance peak. (The running HPLC buffer remained constant at 10 mM ammonium acetate and 35 mM triethylamine acetate in all cases.) This single dominant peak has a nearly identical elution time to that of the Ag + -duplexes in low ionic conditions. Therefore we conclude that they are the Ag + -bound duplex structure. Full mass spectra for solutions synthesized across a wider range of ionic conditions, then solvent exchanged into 10 mM ammonium acetate, are shown in figure SI.6. In particular, because 10 mM Mg 2+ is commonly used to create WC paired DNA nanostructures by assembly techniques such as origami, we tested relative yields of duplex and monomer products for annealing at 10 mM magnesium sulfate with 10 mM ammonium acetate (figure SI.6(c)). We found that this lower Mg 2+ concentration is sufficient to produce high yields of Ag + -assembled duplexes with almost no contamination by monomer byproducts. Thus formation of double-stranded regions paired by Ag + appears consistent with DNA nanotechnology conditions for stable WC pairings.
CD spectra for T 2 C 20 T 2 in 60 mM MgSO 4 has the same overall shape as the CD with no MgSO 4 , but with slightly higher magnitude peaks ( figure SI.7), indicating that the structure of the Ag + -assembled duplex does not vary significantly with ionic strength. Given the apparent removal of the strand monomer products by the use of higher ionic conditions (figures 4(b) and (c)), it appears that charge screening from the additional ions promotes The insets in (a) display mass spectra of the marked peaks in the HPLC chromatogram, which were caught and injected in to an ESI-MS in negative ion mode. The lower retention time peak in (a) consists of Ag + -strand monomer products containing 8 and 9 Ag + while the higher retention time peak consists of Ag + -mediated duplexes containing 19 and 20 Ag + . For labelled, full range mass spectra of these chromatogram peaks please see figure SI.5. Higher ionic conditions in the chromatograms for (b) and (c) produce one narrow peak with a retention time corresponding to the Ag + -bound duplex in (a). monodisperse duplex formation, an important finding for developing protocols for Ag + -assembly of more elaborate structures. With the bound Ag + alone, insufficient screening of the phosphate backbone at low ionic strength may be responsible for the presence of monomer products.
Ag + -mediated pairings of C base mismatches has been shown to provide surprising thermal robustness to otherwise canonical, WC paired DNA duplexes [21,22], and recent studies of Ag + -assembled C 6 and G 6 homobase duplexes found very little change in CD signal even at 90°C [23]. Apparently thermal stability is excellent, at least for these tested base motifs. To determine how resistant these Ag + -mediated base pairings are to chemical perturbation, we use urea because it is a common additive for denaturing DNA [37]. The stability of Ag + -mediated C 20 and G 20 homobase duplexes in varying concentrations of urea, at low ionic conditions, was tested by CD spectroscopy ( figure 5). The Ag + -C 20 solution showed high resistance to urea, with CD magnitude dropping by only 20% at 270 nm in 8 M urea ( figure 5(a)), while retaining overall spectral shape. Conversely, for the Ag + -G 20 solution the CD magnitude dropped by 76% at 282 nm in 8 M urea ( figure 5(b)), though the spectral shape remained similar. It is not clear whether the reduction in CD signal for G 20 relates to a loss of Ag +mediated base pairing or an alteration in the structure of the duplex. The surprising resistance of Ag + -mediated C 20 duplexes to denaturation by urea suggests that structure designs which incorporate abundant C-Ag + -C base pairs could be highly resistant to denaturing.
The high thermal stabilities of G-Ag + -G and C-Ag + -C duplexes and the high chemical stabilities of C-rich Ag + -mediated pairings suggest that two-stage assemblies for incorporating both Ag + and WC pairings may be achievable. For example, an Ag + -mediated assembly stage could be made under thermal or chemically denaturing conditions to suppress WC pairings that might impede Ag + assembly. A subsequent, lower temperature (or denaturant-free) assembly stage could then be made to form WC pairings. Any such a two-stage assembly will require the Ag + -assembled strands to remain stable in the presence of Na + or Mg 2+ cations, which are used to stabilize WC-based DNA nanotechnology. To examine the stability of Ag + -mediated pairings in the presence of additional salts, we measured CD spectra of C 20 and G 20 with 100 mM NaCl (figures 6(a) and (c)) at 1 Ag + /base and with 60 mM MgSO 4 (figures 6(b) and (d)), also at 1 Ag + /base. A control spectrum for each strand at 1 Ag + /base, but no additional salt is included for reference in every graph (figure 6, red curves). For C 20 with 1 Ag + per base (figures 6(a) and (b)), 100 mM NaCl and 60 mM MgSO 4 had similar effects, reducing the magnitude of the CD spectrum while retaining similar spectral shapes. This surprising resistance of C-Ag + -C base pairing to AgCl precipitation is very promising for DNA nanotechnology. For G 20 with 1 Ag + per base and 100 mM NaCl (figure 6(c), green curve) the Ag + -duplex structure appears to be completely disrupted, as evidenced by similarity in dichroic signal to the bare G 20 strand in 100 mM NaCl (figure 6(c), blue curve). Apparently G-Ag + -G base pairs do not persist in high amounts of NaCl. However the CD signal for G 20 in 60 mM MgSO 4 with Ag + (figure 6(d), gold curve) remained spectrally similar in magnitude and shape to the control with no additional salt (figure 6(d), red curve). Thus it seems both C-Ag + -C and G-Ag + -G base parings would be compatible with DNA nanostructures stabilized by Mg 2+ cations.

Towards Ag + -stabilized DNA nanotechnology
Success in any two-stage assembly, with an Ag + -pairing stage followed by a WC pairing stage, will require that the 1st step not introduce a significant amount of disruptive Ag + pairings in the regions for planned WC pairing in the 2nd step. The notion that this two-stage assembly may be achievable for diverse WC pairing motifs is suggested by the recent success in using pre-formed Ag N -DNA, synthesized on strands with mixed base hybridization tails, to decorate DNA nanotubes [38]. The hybridization tails were designed using recent advances in machine learning of 'bright' multi-base motifs that promote formation of fluorescent silver clusters, and 'dark' motifs that do not [39]. To achieve dense decoration of the DNA nanotube, the tails appended to the silver cluster template had to include C and G bases in order to be short enough to hybridize the Ag N -DNA onto closely spaced sites along the nanotube [38]. The successful decoration of the DNA nanotubes with Ag N -DNA designed with select 10-base tails with significant (40%) G, C content indicates that if any silver was attached to these mixed-base tails, it did not significantly interfere with WC hybridization.
We conclude that it is reasonable to expect properly designed strands with up to ∼40% G and C bases to form duplexes primarily through WC pairing after a prior, Ag + mediated assembly stage. This is consistent with the preservation of an overall WC geometry after addition of Ag + to canonical mixed-base WC duplexes with individual C-C mismatches [40], and with recent quantum chemical calculations which found diverse, geometrically distinct but energetically similar Ag + -mediated base pairings [23]. The small differences found in calculated binding energies for differently configured Ag + pairings imply that geometries of Ag + -mediated pairs that are embedded within a background of WC pairs may be strongly influenced by steric constraints from neighboring regions.
The critical issue of strand orientation for Ag + -paired duplexes is not yet established, though early studies of Ag + -paired d(C 8 ) strands, and our results above, provide some evidence for a parallel orientation [25]. Parallel orientations are also suggested by the most stable configurations found in theoretical calculations for C-Ag + -C and G-Ag + -G pairings (see supporting information in [23]).
Detailed structural information on duplexes formed by Ag + -mediated base pairs has yet to be elucidated. However trends in CD spectra for C and G homopolymer strands of increasing length (figure 7) suggest simple Figure 7. CD spectra of varying lengths of (a) G and (b) C homobase oligonucleotides in 7.5 mM MOPS at 1 Ag + per base. CD spectra correspond to sequences of length 2 (red), 3 (green), 6 (purple) and 20 (teal). The trace for C 2 in (a) was reproduced with permission from [24]. helical duplex structures. For Ag + -paired C N strands, this is indicated by the similarity of CD spectra for strand lengths, N, ranging from 2 to 20 ( figure 7(a)). The increase in the per-base CD signal with increasing N (figure 7(a)) is qualitatively as expected for increasingly ordered helices in which the relative contributions from less conformationally constrained end bases decrease with strand length. Also suggestive of simple helical structures is the agreement between strand length, N, and the numbers of Ag + in the dominant C homobase duplex product for lengths N of 2-20 bases [23,24]. For the longer Ag + -paired G N strands, the number of Ag + in the dominant duplex product also agrees with strand length (6 Ag + for N=6 and 20 Ag + for N=20), again suggestive of a simple helical structure [23]. However for N=2 and 3, the CD spectrum changes dramatically ( figure 7(b), right axis), with a large increase in normalized CD magnitude that suggests Ag + -induced strand aggregation [41,42]. We conclude that for both C and G homobase strands, the CD spectra show high amplitudes indicative of Ag + -assembled pairings at lengths of just 2 bases. Thus it appears that Ag + -mediated pairings may enable much shorter base motifs to be used for 'sticky ends' than is the case for WC pairings, which typically require runs of 6 or more bases for adequate stability at room temperature.
Based on the above results, we anticipate that it may be possible to exploit the high stability of Ag + pairings in new types of dynamic DNA nanostructures. Figure 8(a) proposes a simple prototype structure, assuming parallel strand orientations of Ag + -paired motifs. The brown segments of the upper and lower strands contain base motifs that do not Ag + -pair. These are flanked by segments containing different G N -and C N -rich motifs that Ag + -pair with the two outer strands. This would form sturdy side 'frames' supporting a WC-paired 'picture' formed by subsequent WC hybridization (gray strands). We expect that the WC 'picture' could be both thermally and chemically manipulated without disturbing the 'frame.' For example in an extended, ribbon-type structure ( figure 8(b)), disruption of WC paired stem-loop structures in the 'picture' would alter the ribbon width and rigidity, but the strands would be retained by the frame. Such a change in structure, without loss of strands, could be very useful (for example, the change in width could be used to alter flow through nanoconstrictions without any loss of strands).

Conclusions
We have shown that Ag + -base pairing holds promise for DNA nanotechnology through surprising thermal and chemical robustness of Ag + -base pairing, the formation of monodisperse Ag + -paired duplexes, and the consistent structures of Ag + -duplexes with both mixed base and homobase motifs, as indicated by CD spectroscopy. The examples in figure 8 are intended only to convey that new DNA nanostructure function with improved robustness could arise from Ag + pairing of select DNA strands. There are important issues that must be resolved before such structures can be made a reality. These include the issue of strand orientation for Ag + - paired motifs, how best to design strands that engage in both Ag + and WC pairing, and the structural regularity of mixed base Ag + -pairing motifs. We are continuing to investigate these questions and are optimistic about the inclusion of Ag + -base pairing in DNA nanotechnology.