*3.1. Ag+ Induces a Conformational Change from Duplex to MDB*

One-dimensional (1D) 1H NMR experiments were first performed to investigate if Ag+ could bind to C2·C6 mispair of the CCTG MDB. It showed that upon adding Ag<sup>+</sup> to the CCTG MDB, the H6 proton signals of C2 and C6 became broadened while those of T3 and T7 remained sharp and almost unchanged, suggesting that Ag<sup>+</sup> bound to the C2·C6 mispair (Figure 2). Besides, C1 H6, G4 H8, C5 H6, and G8 H8 peaks were also found to be broadened, as it has been reported that Ag+ could also bind to C-G base pairs [44].

We then tested if Ag+ could promote MDB formation to induce a DNA conformational change, which is the prerequisite of most DNA sensors. For this aim, we prepared a DNA duplex formed by the CCTG MDB strand (5 -CCTGCCTG-3 ), namely *CCTG2*, and its complementary strand (5 -CAGGCAGG-3 ), namely *CAGG2*, at pH 8/7/6 and collected CD spectra to monitor DNA conformational change upon Ag+ titration at 25 ◦C. These two strands formed a duplex in the absence of Ag+, as indicated by a positive CD band at 265 nm (Figure 3A–C, black lines) [45]. Upon adding Ag+ to the duplex, a new major band at 290 nm was observed at pH 6, but not obvious at pH 7 and 8, when the DNA:Ag<sup>+</sup> ratio was 1:2 (Figure 3A–D, red lines). The CD band at 290 nm was characteristic of the CCTG MDB [46], suggesting that Ag+ efficiently induced a conformational change from the duplex to the MDB at pH 6. Notably, the DNA:Ag<sup>+</sup> ratio of 1:2 showed the maximum population of Ag+-induced MDB (Figure 3C). This may because Ag+ is also non-selectively bound to C-G base pairs in the MDB (Figure 2), and thus more Ag+ is required to promote MDB formation.

**Figure 2.** NMR spectra of 0.1 mM CCTG MDB in 1 mM sodium cacodylate (pH 6), 90% H2O/10% D2O, with various Ag<sup>+</sup> concentrations at 25 ◦C. Peak broadenings of C2 H6 and C6 H6 in the presence of Ag+ suggest that Ag+ is bound to C2·C6 mispair.

**Figure 3.** CD spectra of 15 μM *CCTG2* and *CAGG2* with 0, 5, 15, and 30 μM Ag<sup>+</sup> in 10 mM NaPi at pH 8 (**A**), pH 7 (**B**), and pH 6 (**C**). (**D**) CD spectra of 15 μM *CCTG2* and *CAGG2* in 30 μM Ag+ at pH 8, 7, and 6 at 25 ◦C. (**E**) CD spectra of 15 μM *CCTG2* and *CAGG2* without Ag<sup>+</sup> and with 30 μM Ag<sup>+</sup> (pH 6) at 25 ◦C and 35 ◦C. Absorbance at 290 nm is characteristic of the free CCTG MDB.

We did not further lower the pH as previous work has demonstrated that the CCTG MDB completely dissociated from the duplex owing to its much higher thermodynamic stability than the duplex at pH 5 [43], therefore there would not be further conformational change upon adding Ag+. We also performed the Ag+ titration at 35 ◦C to examine if this system could function at an elevated temperature. However, the CD signal of MDB was observed without adding Ag+ (Figure 3E), which could be attributed to the relatively higher thermodynamic stability of MDB than duplex at 35 ◦C and pH 6. Zhang et al. have also reported that a higher temperature leads to partial melting of the initial DNA duplex and thus a lower sensitivity [47].

#### *3.2. Design and Optimization of the CCTG MDB-Based DNA (M-DNA) Sensor*

Based on the Ag+-induced formation of CCTG MDB at pH 6 (Figure 3C,D), we designed the *M-DNA* sensor, which was simply composed of the 8-bp duplex formed by *CCTG2* and *CAGG2*. SYBR Green I (SGI) was used as a fluorescence reporter and it was expected to emit strong fluorescence when bound to the duplex in the absence of Ag<sup>+</sup> while giving weak fluorescence when the duplex was converted to MDB in the presence of Ag+ (Figure 4A). To ensure SGI will not affect the DNA conformational change, CD spectra were collected without and with adding SGI, and the results showed that Ag+-induced conformational change still effectively occurred (Figure S1).

**Figure 4.** (**A**) Schematic of the *M-DNA* sensor for Ag<sup>+</sup> detection. (**B**) Normalized fluorescence intensity at 520 nm as a function of time for the *M-DNA* sensor in the absence of Ag<sup>+</sup> (black) and after adding 50 nM Ag+ (green). (**C**) Fluorescence spectra of the *M-DNA* upon titrating Ag<sup>+</sup> ranging from 0 to 200 nM (**left**) and the fitting curve constructed using fluorescence intensity at 520 nm and log[Ag+]/log[*M-DNA*] (R2 = 0.99) (**right**). Error bars were standard deviations obtained from three replicative experiments.

At pH 6, the *M-DNA* concentration and SGI:*M-DNA* ratio were further optimized. Two *M-DNA* concentrations (50 and 200 nM) and four SGI:*M-DNA* ratios (0.1:1, 0.5:1, 1:1, and 5:1) were tested to find the condition that would give the largest fluorescence change in response to Ag+. The DNA concentration and SGI:*M-DNA* ratio were finally optimized to be 50 nM and 1:1, respectively (Figure S2). Therefore, the *M-DNA* used for Ag<sup>+</sup> sensing in the following experiments contained 50 nM *CCTG2*, 50 nM *CAGG2*, and 50 nM SGI in 10 mM NaPi at pH 6, unless otherwise specified.

To further verify whether the CCTG MDB played an important role in the *M-DNA* sensor for Ag+ detection, we also performed Ag<sup>+</sup> titration on a controlled DNA (named *C-DNA*), which was an 8-bp self-complementary duplex. When the mixture of 50 nM *C-DNA* and 50 nM SGI in 10 mM NaPi at pH 6 was titrated with Ag+, there was only a little change in fluorescence intensity (Figure S3), suggesting that the CCTG MDB played an irreplaceable role in Ag+ sensing.

#### *3.3. Kinetics, Sensitivity, and Selectivity of the M-DNA Sensor*

One of the most interesting features of this *M-DNA* sensor is using an ultrashort 8-nt oligonucleotide, which is expected to undergo a much faster conformational change than longer i-motif and hairpin sequences [26,29,30,32]. Therefore, we also evaluated the kinetics of this *M-DNA* for Ag<sup>+</sup> sensing. The fluorescence intensity (520 nm) of the *M-DNA* sensor without Ag<sup>+</sup> was recorded from 0 to 180 s with a step time of 2 s. Ag<sup>+</sup> was then added to the same sample, and the fluorescence intensity was immediately recorded from 0 to 180 s with a step time of 2 s. Figure 4B shows that immediately after adding Ag+, the fluorescence intensity drastically decreased and remained almost unchanged through the entire monitoring process for 180 s. Therefore, it is safe to conclude that the reaction was completed within the acquisition time for the first data point, i.e., 2 s. It was reported that the Ag+-triggered conformational change from a single-stranded DNA to a 21-nt i-motif was complected in ~15 s [26,29,30,32], therefore it is reasonable that the conformational change to an 8-nt MDB was much faster.

The *M-DNA* was then used to sense Ag+ at various concentrations ranging from 0 to 200 nM (Figure 4C). There was a good linear correlation between the fluorescence intensity and log[Ag+]/log[M-DNA]. Following the rule of three times the standard deviation over the blank response [48], the Ag+ detection limit was determined to be ~2.1 nM. As the tolerable level of Ag<sup>+</sup> in drinking water is ~927 nM [7], the detection limit of the *M-DNA* sensor should be sufficient for detecting Ag+ in real samples containing Ag+.

The anti-interference capability of the *M-DNA* sensor for Ag+ detection in a complex environment was also evaluated. As the drinking water source may also contain other metal ions, we evaluated the fluorescence response of *M-DNA* to K+, Li+, Ca2+, Mg2+, Mn2+, Co2+, Cu2+, Ba2+, and Ni2+, and the result showed only tiny fluorescence changes upon adding these ions (Figure 5A). Furthermore, an additional experiment was also performed to examine if the *M-DNA* could detect Ag+ in the presence of these interfering metal ions. Upon adding 50 nM Ag+ to the solutions containing the respective interfering metal ions, the fluorescence change became significant and achieved a similar level to that of only 50 nM Ag<sup>+</sup> (Figure 5B). Na<sup>+</sup> was not included as an interference ion in this study because the buffering system contained 10 mM NaPi. Approximately 10 to 200 mM Na+ are also commonly used in buffering systems for many DNA-based sensors to neutralize the negatively charged phosphodiester backbones [26,29–31,34]. The concentrations of non-Ag+ ions vary in different water samples, e.g., few mM Na+ in most China river and lake basins [49] and hundreds mM Na+ in sea water [50]. The *M-DNA* sensor should be applicable for detecting Ag<sup>+</sup> in common river and lake basins, and its performance may need to be further improved for sensing Ag<sup>+</sup> in water samples containing high concentrations of interfering ions (e.g., sea waters).

**Figure 5.** Fluorescence changes at 520 nm of the *M-DNA* in the presence of (**A**) 50 nM non-Ag+ metal ions (blue) and (**B**) 50 nM non-Ag<sup>+</sup> metal ions plus adding 50 nM Ag<sup>+</sup> (blue). The fluorescence change in the presence of only 50 nM Ag+ was shown as a reference (red). Error bars were standard deviations obtained from three replicative experiments. *F0*: initial fluorescence intensity in the absence of Ag+; *F*: fluorescence intensity after adding 50 nM AgNO3 or other metal ions.
