All the DNA were purchased from Integrated DNA Technologies (IDT, Coralville, IA, USA), and their sequences are shown in Table S1. Commercial citrate-capped AgNPs (20 nm diameter) were purchased from Nanocomposix (San Diego, CA, USA). Trisodium citrate and 3-(N-morpholino)propanesulfonic acid (MOPS) were obtained from Sigma-Aldrich (St. Louis, MO). Sodium nitrate (NaNO₃) was purchased from Mandel Scientific (Guelph, ON, Canada).

UV-vis absorption spectra were recorded on a spectrometer (Agilent 8453A). The morphology of AgNPs was examined by a transmission electron microscope (TEM, Philips CM10). The etching kinetics of AgNPs were monitored using a microplate reader (SpectraMax M3). Dynamic light scattering (DLS) measurements were carried out using a Zetasizer Nano 90 (Malvern) at 25°C. Circular dichroism (CD) spectra were collected on a Jasco J-715 spectrophotometer (Jasco, Japan).

In a typical experiment, a 15-mer DNA (20 μM, 35 μL) was incubated with an equal volume of AgNPs (10 μg/mL) at 37°C for 1.5 h. The final concentration of DNA was 10 μM. Then, the sample was analyzed by a spectrometer.

Typically, PO-C₁₅, PS₁₄-C₁₅, or PS₁₄-T₁₅ (20 μM, 50 μL) was mixed with the AgNPs (10 μg/mL, 50 μL) in a 96-well plate. Then, 10 μL of 10 mM buffer with different pH values (citrate buffer for 4.0, 5.0, and 6.0; MOPS for 7.0 and 7.9) was added and incubated at 37°C for 1.5 h. The absorbance intensity was monitored at 395 nm every 0.5 min under the kinetic mode using the plate reader.

First, PS-R DNA (40 μM, 17.5 μL) was mixed with 17.5 μL of its cDNA of different concentrations (5.0, 10, 20, 40, 60, 80, and 100 μM), followed by the addition of NaNO₃ (0.25 M, 2.0 μL). Then, the mixture was heated to 95°C for 5 min and cooled slowly to room temperature. Finally, AgNPs (10 μg/mL, 35 μL) were added and incubated at 37°C for 1.5 h. The final concentrations of the cDNA were 1.25, 2.5, 5.0, 10, 15, 20, and 25 μM, respectively.

To test the effect of PS modification (see Figure 1A for its structure), we used the four types of 15-mer DNA homopolymers both with the normal phosphodiester (PO) backbone and with full PS modifications (each bridging phosphate contained a PS modification). Our 20 nm AgNPs had a strong surface plasmon peak at 395 nm (Figure 1C). Adding the normal PO-A₁₅ DNA had no effect and the UV-vis spectrum retained its original shape. In contrast, the PS₁₄-A₁₅ (note that a 15-mer DNA has only 14 bridging phosphates) dropped the extinction peak intensity by over 80%. From this experiment, we concluded that the sulfur atoms in the PS-DNA were the reason for the decreased extinction of the AgNPs. From TEM (Figure 2A and Figures S1, S2A), our starting AgNPs were monodispersed ~20 nm spheres. After adding the PS₁₄-A₁₅, overall the AgNPs became smaller (Figures 2B,F), indicating its etching.

Similar experiments were also performed with the other DNA sequences, and the same observations were also made with the two T₁₅ DNAs from both their UV-vis spectra (Figure 1D) and TEM (Figure 2C). The lack of etching by PO-A₁₅ and PO-T₁₅ is in agreement with the relatively low affinity between these two DNA bases and silver surface (Basu et al., 2008; Wu et al., 2014).

When the PO-G₁₅ was added, the extinction intensity of the AgNPs dropped by about 20% (Figure 1E), while a nearly 80% drop was observed when PS₁₄-G₁₅ was added. At the same time, the peak red shifted by 18 nm. This suggested formation of larger AgNPs, which was confirmed by TEM (Figure 2D). Therefore, with this DNA concentration and reaction time, the PS₁₄-G₁₅ helped Ostwald ripening of the AgNPs. Etching was the first step of the interaction, where the AgNPs were dissolved by the added DNA. Extensive etching and deposition of dissolved silver species on larger AgNPs (thus with lower solubility) resulted in the subsequent ripening.

The PO-C₁₅ DNA was very effective in etching the AgNPs and it decreased extinction by 65% (Figure 1F). The extinction of the PS₁₄-C₁₅-treated sample further decreased the extinction peak to nearly 90%. Interestingly, under this condition, the ripening process caused only a 3 nm redshift, which was much smaller than that induced by PS₁₄-G₁₅ (18 nm). Meanwhile, the average size of the PS₁₄-C₁₅-treated AgNPs (Figure 2E) was smaller than that of the PS₁₄-G₁₅ treated sample (Figure 2D), indicating the cytosine and guanine bases also played a role on etching.

With these four pairs of DNA, we plotted the peak intensity drop in Figure 1B. All the PS sequences dropped the intensity by a similar value (the red bars), while a much larger difference was observed with the normal PO-DNA (blue bars). Since all the PS-DNA sequences had a significant etching effect, PS-DNA can etch the AgNPs in a less sequence-independent manner. This might be useful since etching can be general to different DNA sequences.

The above experiments indicated that PO-T₁₅ is one of the least effective sequences for etching AgNPs. Therefore, the etching effect of PS₁₄-T₁₅ should mainly come from the PS sites. PO-C₁₅ is the best PO DNA and only the bases are available for etching, while PS₁₄-C₁₅ is likely to be a most effective sequence overall. For PS₁₄-C₁₅, both the PS sites and the cytosine bases are likely to contribute to silver binding. Since these three sequences are representative, they were chosen for further studies. Their silver coordination sites are marked in black and red circles in Figure 1A.

We first studied the effect of DNA concentration. In each case, the peak intensity decreased with increase of DNA concentration (Figures 3A–C). At the same time, some redshifted peaks were observed at high DNA concentrations. All these experiments were performed with an incubation time of 1.5 h, when the systems were approaching equilibrium (Figure 4).

By plotting the decrease of peak height against DNA concentration, we obtained their apparent binding curves (Figure 3D). Among these DNAs, PO-C₁₅ had the lowest response with an apparent K_d of 3.87 μM. PS₁₄-C₁₅ had a tighter binding with a K_d of 1.78 μM. Interestingly, PS₁₄-T₁₅ had a K_d similar to that of PS₁₄-C₁₅, and thus from this standpoint, the base's contribution was minimal. Even at a relatively low DNA concentration, the base did not contribute much to etching. Adding PS modifications decreased the concentration requirement for C₁₅ by 2.2-fold, while the improvement for T₁₅ was close to infinity (compared to PO-T₁₅).

We then plotted the shift of peak wavelength (Figure 3E), where the upper half of the figure is for red shifted samples and the lower half for blue shifts. With low concentrations of PO-C₁₅, a gradual blue-shift of the AgNPs peak was observed and the maximal shift was achieved with 1.5 μM DNA, indicating etching of the AgNPs to form small particles. Then, the peak started to red shift attributable to Ostwald ripening. When the DNA concentration was more than 7.5 μM (crossing the dashed line in Figure 3E), the peak red-shifted relative to the original AgNPs.

For PS₁₄-C₁₅, most of their spectra were red shifted (e.g., ripening) except for the DNA concentration below 1 μM. From TEM, the overall size was indeed decreased for the 0.5 μM PS₁₄-C₁₅ treated sample (Figure S2B). A wide size distribution with both larger and smaller AgNPs was observed for more PS₁₄-C₁₅ (7.5 μM), confirming the ripening (Figure S2C). However, we cannot rule out slight aggregation of AgNPs occurring at the same time, which also caused the red shifted spectra. By comparing PO-C₁₅ and PS₁₄-C₁₅, both showed etching at low DNA concentrations and then ripening with more DNA added. PS₁₄-C₁₅ has a tighter affinity with silver allowing it to achieve the etching-to-ripening transition at a lower DNA concentration.

Interestingly, for PS₁₄-C₁₅, the red shift initially increased but later decreased when the concentration of PS₁₄-C₁₅ was more than 3.75 μM (red trace in Figure 3E). A similar trend was also observed for PS₁₄-T₁₅ despite smaller shifts compared to that of PS₁₄-C₁₅. This difference may be ascribed to their different DNA bases, suggesting that the cytosine bases of PS₁₄-C₁₅ also participated in the etching process. We reason that PS₁₄-C₁₅ had a complex multi-stage etching process. Low concentration of DNA contributed to AgNPs etching and further ripening (Figures S2A–C). Further increased PS₁₄-C₁₅ DNA could further etch the larger AgNPs from the previous ripening step, which yielded the decreased red shift. The etching and thus size decrease was also confirmed by TEM (Figure S2D). As a result, we proposed a three-stage mechanism for PS-DNA to interact with AgNPs: etching, ripening and further etching (Figure 3F). For the PO-DNA, we only observed two stages (etching-ripening of AgNPs) indicating that cytosine bases alone were incapable of etching larger AgNPs, which were thermodynamically more stable than the originally used 20 nm ones. Since PS₁₄-T₁₅ was also not very obvious than PS₁₄-C₁₅ for this three-stage process, both cytosine bases and PS of PS₁₄-C₁₅ contribute to the etching-to-ripening transition (with PS being the major contributor).

To further study etching, we followed the reaction kinetics. Since the conformation of poly-C DNA is strongly affected by pH (Dong et al., 2014; Huang Z. et al., 2016), we also measured the etching kinetics at different pH values. For PO-C₁₅, etching was strongly inhibited at low pH (Figure 4A). In particular, when pH was at 6 or lower, etching was essentially fully inhibited. In contrast, PS₁₄-C₁₅ had the same rate of etching regardless of pH from 4 to 7.9 (Figure 4B). We fitted the kinetic data of the PS-DNA to a first-order equation and obtained a rate constant of 41.3 h⁻¹, which was much faster than the PO kinetics of 183.4 h⁻¹ at pH 7.9 (the rate of the PO samples was even slower at lower pH). Since the only difference here was the base, we reason that the inhibited PO-C₁₅ etching must be related to its base protonation and formation of secondary structures such as the i-motif (Figure 4D).

Using circular dichroic (CD) spectroscopy, a strong positive peak at around 285 nm and a small negative peak near 260 nm were observed suggesting an intramolecular i-motif structure of PO-C₁₅ at pH 4.0 (the black spectrum in Figure S3) (Liu and Balasubramanian, 2003). Such a folded conformation could shield the bases and inhibit their interaction with AgNPs or with Ag⁺. For the PS₁₄-C₁₅-mediated etching, pH had no effect on etching. Since the PS modifications could cause a reduced melting temperature compared to the PO counterpart (Gonzalez et al., 1991), the PS₁₄-C₁₅ was incapable of forming i-motif at 37°C even under acidic conditions (the red spectrum in Figure S3). Therefore, the exposed PS and the bases in the random-coil structured PS₁₄-C₁₅ could serve as the ligand for etching the AgNPs. The pH-independent etching also appeared for PS₁₄-T₁₅ (Figure 4C), demonstrating the generality of PS-DNA-mediated etching of AgNPs.

The above experiments used 15-mer DNA with full PS modification. We then varied the number of DNA length and PS modifications (see Figure 5A for the DNA sequences). First, the DNA length was explored. To minimize the effect of the DNA base, PS-modified poly-T DNAs were tested. The length of DNA varied from 5-mer to 15-mer, and the total PS modification was maintained to be the same (e.g., the molar concentration of PS₄-T₅ was 3.5 times of that of PS₁₄-T₁₅). The peak of the PS₁₄-T₁₅ sample dropped more than that of PS₄-T₅, suggesting that longer DNA was more effective and thus the importance of polyvalent binding (Figure 5B).

We then varied to the number of PS modifications, while the DNA length was maintained at 15-mer. The number of PS modifications was reduced from 14 to 7, 4, 2, and 1 (Figure 5C). The peak intensity gradually dropped with increased PS modifications. For the poly-T DNAs with 1–7 PS modifications, the drop in the peak intensity was linearly proportional to the number of PS (Figure S4), further highlighting that the PS responsible for the AgNP etching. This also provided a method to quantitatively tune the extent of etching. Further increase of the PS modifications to 14 did not bring in much more changes, suggesting that seven PS modifications could be sufficient with the 1.5 h incubation time.

Finally, we explored the effect of the location of PS modifications. Compared to the uniform distribution of PS in the whole DNA backbone of PS₇-T₁₅, the 7 PS modifications in PS_7r-T₁₅ were concentrated on the 3′-terminus of the DNA. Interestingly, the evenly distributed PS₇-T₁₅ had a stronger decrease (Figure 5D), implying that PS coordination is more effective when they are separated.

Effective adsorption of PS-DNA on AgNPs could be important for the etching process. All the above experiments used flexible single-stranded DNA oligonucleotides, while a rigid DNA structure (e.g., duplex) may hinder the attachment of DNA to AgNPs due to restricted binding sites (Figure 6A). To test this hypothesis, we explored the effect of DNA conformation on etching by forming duplex DNA. However, PS modifications can weaken the stability of duplex DNA as reflected from the reduced melting temperature (T_m) (Gonzalez et al., 1991). Furthermore, the A-T base pair with a PS modification showed more decreased T_m than that of the C-G base pair (Stein et al., 1988). Therefore, we designed a PS-modified random DNA (named PS-R DNA) with a high GC content (Figure 6B). This DNA could etch the AgNPs (the black spectrum in Figure 6C), and the etching efficiency was gradually inhibited with increasing dose of the complementary DNA (cDNA). The inhibiting efficacy was sharply decreased when the misDNA with a single mismatched base was added, while a full non-complementary DNA (T₃₀) had little inhibition effect (Figure 6D). Therefore, we can attribute the cDNA-dependent etching to the formation of the duplex DNA. In other words, single-stranded DNA is much more effective for etching the AgNPs, although the PS backbone is still fully exposed in duplex DNA. This implies that DNA needs to fold into optimal binding structures, and etching cannot take place effectively on isolated PS sites.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.