M-DNA/Transition Metal Dichalcogenide Hybrid Structure-based Bio-FET sensor with Ultra-high Sensitivity

Here, we report a high performance biosensor based on (i) a Cu2+-DNA/MoS2 hybrid structure and (ii) a field effect transistor, which we refer to as a bio-FET, presenting a high sensitivity of 1.7 × 103 A/A. This high sensitivity was achieved by using a DNA nanostructure with copper ions (Cu2+) that induced a positive polarity in the DNA (receptor). This strategy improved the detecting ability for doxorubicin-like molecules (target) that have a negative polarity. Very short distance between the biomolecules and the sensor surface was obtained without using a dielectric layer, contributing to the high sensitivity. We first investigated the effect of doxorubicin on DNA/MoS2 and Cu2+-DNA/MoS2 nanostructures using Raman spectroscopy and Kelvin force probe microscopy. Then, we analyzed the sensing mechanism and performance in DNA/MoS2- and Cu2+-DNA/MoS2-based bio-FETs by electrical measurements (ID-VG at various VD) for various concentrations of doxorubicin. Finally, successful operation of the Cu2+-DNA/MoS2 bio-FET was demonstrated for six cycles (each cycle consisted of four steps: 2 preparation steps, a sensing step, and an erasing step) with different doxorubicin concentrations. The bio-FET showed excellent reusability, which has not been achieved previously in 2D biosensors.

the concentration dependent physical characteristics of various ions and doxorubicin coordinated DNA molecules in order to estimate the optimum concentration for structural stability and determine significant changes in functionalities [13][14][15][16][17][18][19] . Considering the parameters required to maintain deformation-free DNA nanostructures, we added the appropriate amounts of Cu 2+ ions and doxorubicin molecules into the DNA nanostructures. As shown in Fig. 1a, the Cu 2+ ions became intercalated between base-pairs and bound onto the phosphate backbone sites because they were size compatible. The DNA nanostructures with Cu 2+ ions were mostly stable up to 4 mM, but above this concentration the structures were not well formed. This can be explained according to specific and non-specific binding. While specific binding dominates at a low Cu 2+ ion concentration in the DNA, non-specific binding also plays a role at a higher concentration. In the case of specific binding, the lattices are minimally deformed because Cu 2+ ions are systematically coordinated with particular sites. In non-specific binding, the lattices may be more severely damaged and deformed because the Cu 2+ ions are randomly coordinated within the DNA duplex. When we coordinated the doxorubicin into DNA, the binding of doxorubicin had periodic orientation via an intercalation mode between the nucleosides. Because doxorubicin has a rough plane shape and is larger than the single nucleotide, doxorubicin molecules are expected to reside between layers of base-pairs of DNA duplexes through chemical bonds with nucleosides, as shown in Fig. 1. Figure 1b,c show the absorption spectra and the analysis (intensity and maximum peak positions) of the absorption peaks, which were obtained on various DNA nanostructures with doxorubicin, Cu 2+ ions, and a combination of Cu 2+ ions and doxorubicin. For reference, the absorption peak of pristine DNA is located at about 260 nm. The DNA absorption peak intensity increased with coordinating doxorubicin, Cu 2+ ions, and a combination of Cu 2+ ions and doxorubicin into DNA nanostructures. Additionally, new absorption peaks appeared at approximately 505 nm for doxorubicin, 660 nm for Cu 2+ ions, and 596 nm for the combination of Cu 2+ ions and doxorubicin into DX DNA nanostructures. Tremendous changes in intensity of the DNA absorption peaks and a significant shift in the new absorption peak position by the coordinating molecules indicate that the Cu 2+ ions and doxorubicin were properly bound to the DNA nanostructures. The absorption peak intensity values of Cu 2+ ions or doxorubicin coordinated DNA were much lower than that of DNA nanostructures with the combination of Cu 2+ ions and doxorubicin. This interesting behavior indicates that the Cu 2+ ions in DNA worked as mediators for binding more doxorubicin molecules into DNA nanostructures, because the opposite polarity induced strong interactions between positively charged Cu 2+ ions and negatively charged doxorubicin molecules. Furthermore, we performed atomic force microscopy (AFM) to confirm the DX nanostructure formation with Cu 2+ ions, doxorubicin, and a combination of Cu 2+ ions and doxorubicin. In Fig. 1d, the noise-filtered 2D fast Fourier transform spectrum images show the periodicity of the unit building block (DX tile). The AFM images revealed that the surface morphology of pristine DNA and coordinated DNA nanostructures were similar, indicating that the coordinating molecules (Cu 2+ ions and doxorubicin) did not damage the DNA nanostructures.

The influence of receptor (DNA and Cu 2+ -DNA) and target (doxorubicin) molecules on MoS 2 .
To investigate the influence of (i) DNA/Cu 2+ -DNA nanostructures (receptors) and (ii) doxorubicin molecules (targets), Raman analysis was performed on the MoS 2 flakes coated by DNA or Cu 2+ -DNA nanostructures before and after being exposed to doxorubicin molecules with various concentrations (10 −4 μ M, 10 −3 μ M, 10 −2 μ M, 10 μ M, 30 μ M, and 50 μ M). Figure 2a shows the Raman spectra measured on pristine MoS 2 , DNA/MoS 2 , and DNA/MoS 2 after being exposed to a 50 μ M solution of doxorubicin molecules. The conventional peaks (E 1 2g and A 1g ) were observed at 382 cm −1 and 408 cm −1 on pristine MoS 2 , which respectively indicate the in-plane and out-of-plane vibrations for bulk MoS 2 . After coating the pristine MoS 2 with DNA nanostructures, the E 1 2g and A 1g peaks were shifted towards the negative direction (Δ E 1 2g = − 1.9 cm −1 and Δ A 1g = − 2.0 cm −1 ) 20 . This is because the DNA nanostructures have negative charges originating from the phosphate backbones (PO 4 − ), and those subsequently attract holes in MoS 2 to the interface region between DNA nanostructures and MoS 2 21,22 . As shown in Fig. 2b, both peak shift values (Δ E 1 2g and Δ A 1g ), which indicate differences between MoS 2 Raman peak positions before and after detecting doxorubicin, were further shifted to the negative direction as the concentration of doxorubicin increased from 10 −4 μ M to 50 μ M. We attributed this behavior to the negative charges from doxorubicin (originating from -OH functional groups), which become stronger on the MoS 2 surface. The Raman spectra of pristine MoS 2 , Cu 2+ -DNA/MoS 2 , and Cu 2+ -DNA/MoS 2 after being exposed to 50 μ M doxorubicin are shown in Fig. 2c. The Cu 2+ -DNA nanostructures attract electrons in MoS 2 and then hold them at the interface between Cu 2+ -DNA and MoS 2 , resulting in positive shifts of Raman peaks (Δ E 1 2g = 5.1 cm −1 and Δ A 1g = 5.2 cm −1 ). When doxorubicin molecules were detected by the Cu 2+ -DNA, the shifted Raman peaks moved back in the negative direction. The degree of the positive shift was determined by the concentration of doxorubicin because of the -OH functional groups that had negative charges in the doxorubicin molecules. In the case of a low doxorubicin concentration between 10 −4 μ M and 10 −2 μ M, the peak shift values were relatively small (− 1.1 to − 2.1 cm −1 ) due to the existing positive charges from the Cu 2+ ions, as seen in Fig. 2d. However, above 10 μ M of doxorubicin, the peaks were more negatively shifted as the Cu 2+ ions were overwhelmed by doxorubicin, resulting in high negative peak shift values (Δ E 1 2g = − 7.5 cm −1 and Δ A 1g = − 7.5 cm −1 ). In the Cu 2+ -DNA/MoS 2 hybrid structure, we observed relatively larger shifts in both E 1 2g and A 1g peaks about − 1.1 and − 1. In order to confirm the change of work function in MoS 2 flakes coated by DNA or Cu 2+ -DNA nanostructure before and after being exposed to doxorubicin molecules, we also conducted Kelvin probe force microscopy analysis. Before the measurement, KPFM tip was calibrated on a highly oriented pyrolytic graphite (HOPG) surface. Then, we obtained the work function (W) values from the contact potential difference (Δ V CPD ) between the KPFM tip and MoS 2 surface. Figure 2e shows the work function mapping images taken on the surfaces of the MoS 2 flakes coated by DNA nanostructure before and after being exposed to doxorubicin molecules. After coating the pristine MoS 2 with DNA nanostructures, the brighter image was observed, indicating that the work function of MoS 2 was Scientific RepoRts | 6:35733 | DOI: 10.1038/srep35733 decreased. After being exposed to a 50 μ M doxorubicin molecules, the image was brighter than that of DNA/ MoS 2 surface because the work function of MoS 2 was further decreased by the negative charges from doxorubicin molecules. However, in the other case using Cu 2+ -DNA nanostructure (Fig. 2f), the darker image was obtained when compared to the pristine MoS 2 surface after being coated by Cu 2+ -DNA nanostructures, because the work function was increased by the positive charges in Cu 2+ -DNA. After doxorubicin molecules were detected by the Cu 2+ -DNA, the image became brighter, even compared to the case that doxorubicin molecules were detected by DNA nanostructures. Figure 2g shows the work function values extracted on the pristine MoS 2 , DNA/MoS 2 , and DNA/MoS 2 after being exposed to a 50 μ M solution with doxorubicin molecules. The work function values were extracted at the specific line in MoS 2 KPFM images, where the decreasing trend was observed (W pristine MoS2 = 4.79 eV and W DNA/MoS2 = 4.77 eV) due to the negative charges originating from the phosphate backbones (PO 4 − ). After being exposed to 50 μ M doxorubicin solution, the work function values were further decreased (W DNA/MoS2+DOX = 4.65 eV) because of the negative charges from doxorubicin molecules. In the case of MoS 2 coated by Cu 2+ -DNA nanostructures, work function value was increased (W pristine MoS2 = 4.80 eV and W Cu2+-DNA/MoS2 = 4.85 eV) because the positive polarity of Cu 2+ ions attracted the electrons. When doxorubicin molecules were detected by the Cu 2+ -DNA, the lower work function value was obtained (W Cu2+-DNA/MoS2+DOX = 4.58 eV) due to the -OH functional groups with negative polarity in the doxorubicin molecules, which was even lower than the case of the DNA/MoS 2 hybrid structure. Based on these Raman and KPFM analyses, the receptor template consisting of Cu 2+ -DNA is expected to further improve the ability to detect doxorubicin, compared to the DNA receptor template. Figure 2i presents schematic diagrams of DNA/MoS 2 and Cu 2+ -DNA/MoS 2 hybrid structures before and after doxorubicin molecules were detected, which also explain how the detected doxorubicin   the Cu 2+ -DNA nanostructure (p-doping of MoS 2 ). Figure 3d shows the energy band diagrams at the Ti/MoS 2 junctions for the three cases of pristine MoS 2 , DNA/MoS 2 (1 st process), and DNA/MoS 2 after detecting doxorubicin (2 nd process). After the DNA nanostructures were coated on pristine MoS 2 , the Ti/MoS 2 junction had a lower effective electron barrier height for electron injection. This effective barrier height decreased further when doxorubicin was detected by the DNA/MoS 2 hybrid structure because of the negative charges originating from DNA nanostructures and doxorubicin. Figure 3e shows the I D -V G characteristics of DNA/MoS 2 -based bio-FET for various concentrations of doxorubicin. Since the effective electron barrier height was reduced as the concentration of doxorubicin increased from 10 −4 μ M to 50 mM, an increase in off-current (from 1.4 × 10 −12 A/μ m to 2.7 × 10 −10 A/μ m) was observed. Then, we calculated sensitivity, (I off_after sensing − I off_before sensing )/I off_before sensing , as a function of gate voltage at different drain voltages when detecting 50 mM of doxorubicin, as shown in Fig. 3f. Here, a very high sensitivity of 82.5-325.9 A/A was observed at a V GS of about − 30 V because a much lower base current (8.4 × 10 −13 A/μ m) was obtained in the off-state as compared to other V GS regions higher than − 22 V. A higher V DS resulted in a little bit higher sensitivity because a higher electric field occurred in the MoS 2 region under a biased condition. This consequently increased the probability of collecting injected electrons when doxorubicin was introduced. Figure 3g presents the sensitivity of the DNA/MoS 2 -based bio-FET as a function of the concentration of doxorubicin at different drain voltages. The sensitivity at V DS = 5 V increased from 0.6 to 325.9 A/A as the concentration of doxorubicin increased from 10 −4 μ M to 50 mM; these sensitivity values were higher than the sensitivity values at V DS = 1V for all concentrations. Overall, in the DNA/MoS 2 bio-FET, we obtained a high sensitivity of 325.9 A/A through a proper voltage bias condition (a more negative V GS for a lower base current and a higher V DS for better electron collection).
Similarly, Fig. 3h presents an energy band diagram of Ti/MoS 2 junctions for another three samples: pristine MoS 2 , Cu 2+ -DNA/MoS 2 (1 st process), and Cu 2+ -DNA/MoS 2 after detecting doxorubicin (2 nd process). Unlike the case of the DNA nanostructure on MoS 2 , the effective electron barrier height was higher than that of pristine MoS 2 when the Cu 2+ -DNA nanostructure was introduced onto the MoS 2 . Since the Cu 2+ ions bonded with phosphate backbones and base pairings provide a positive polarity, electron carriers were held at the interface between Cu 2+ -DNA and MoS 2 . Because of this p-doping effect, the energy band was up-shifted compared to the pristine MoS 2 , eventually increasing the effective electron barrier height. When doxorubicin was detected by the Cu 2+ -DNA/MoS 2 hybrid structure, negative charges from doxorubicin molecules seemed to compensate for the strength of the Cu 2+ positive ions, subsequently shifting down the MoS 2 energy band and increasing the electric field at the Ti/MoS 2 junction. This phenomenon consequently reduced the effective barrier height of the Ti/MoS 2 junction, eventually increasing the current (i.e., a Schottky barrier lowering effect). The I D -V G characteristics of Cu 2+ -DNA/MoS 2 -based bio-FETs for various concentrations of doxorubicin are plotted in Fig. 3i, where the off-current at V GS = − 30 V increased significantly from 7.2 × 10 −12 to 4.6 × 10 −9 A/μ m in the same doxorubicin concentration range (between 10 −4 μ M and 50 mM) as compared to the case of the DNA/MoS 2 -based bio-FET. This is because the Cu 2+ ions incorporated into the DNA improved the ability of the DNA-based receptors to detect negatively polarized doxorubicin. In this case, phosphate backbone (PO 4 − ) sites allowed capture of doxorubicin via the Cu 2+ ions in addition to the base parings. As shown in Fig. 3j, the corresponding sensitivity was also calculated as a function of gate voltage at different drain voltages, where we obtained very high sensitivity in a much wider V GS region (maximum 219.3 A/A at V DS = 1 V and maximum 1757.1 A/A at V DS = 5 V under V GS = − 14 V). The sensitivity was roughly five times higher than that of the DNA/MoS 2 -based bio-FET. This increased operating region (V GS < − 14V) of the bio-FET was attributed to an increase in electron barrier height by the Cu 2+ -DNA-based p-doping effect and the subsequent positive shift in V TH . In addition, like the case of the DNA/ MoS 2 hybrid structure, an 8-fold higher sensitivity was observed at higher V DS because electron carriers injected after detecting doxorubicin were also expected to be more easily collected under stronger electric fields at higher V DS . Figure 3k shows the sensitivity of the Cu 2+ -DNA/MoS 2 based bio-FET as a function of the doxorubicin concentration at different drain voltages. Here, we confirmed that the sensitivity increased from 1.7 to 1757.1 A/A at V DS = 5 V as a function of doxorubicin concentration, which was also higher than the values when V DS = 1 V was applied. Additionally, we investigated the long-term stability of the Cu 2+ -DNA/MoS 2 based bio-FET by exposing the device to air for two weeks (Supporting information Figure S3).

Reusability of Cu 2+ -DNA/MoS 2 based-biosensor.
Finally, we investigated the reusability of the Cu 2+ -DNA/MoS 2 hybrid structure-based bio-FETs by repeating the preparation, sensing, and erasing steps. Figure 4a shows a schematic diagram that graphically explains the 1 st /2 nd preparation, sensing, and erasing steps. In the first step, DNA nanostructures were coated and dried several times on a MoS 2 layer to cover the surface of MoS 2 (1 st preparation step), followed by the addition of Cu 2+ ions (2 nd preparation step) for improving the doxorubicin detecting ability. Then, various concentrations of doxorubicin were dropped onto the bio-FET device and I D -V G measurement was performed (3 rd sensing step). After sensing, the used doxorubicin molecules and Cu 2+ ions were removed by deionized (DI) water to reuse the bio-FET with DNA receptor template (4 th erasing step). Because DI water rinsing does not destroy the DNA/MoS 2 hybrid structure (Cu 2+ ions and doxorubicin are only soluble in DI water), we started the second detection cycle by again transferring Cu 2+ ions onto the DNA/TMD bio-FET (2 nd preparation step). Then, another sensing process was performed, and the used doxorubicin molecules and Cu 2+ ions were removed again using the DI water rinsing process. In this way, the proposed Cu 2+ -DNA/ MoS 2 bio-FET can be used repeatedly, and it is also possible to detect various target molecules with one device if the molecules have negative polarity. Figure 4b presents Raman spectra of the MoS 2 measured over four detecting steps. After the Cu 2+ ions were dropped on the DNA/MoS 2 , both E 1 2g and A 1g peaks in MoS 2 shifted in the positive direction as already confirmed in the previous Raman analysis of Cu 2+ -DNA/MoS 2 . Then, the E 1 2g and A 1g peaks were moved in the negative direction by the negative polarity of the doxorubicin, which compensated for the effect of the Cu 2+ ions. After the fourth erasing step, the position of E 1 2g and A 1g peaks returned to their original position (i.e., the position observed for DNA on MoS 2 ). The I D -V G characteristics of Cu 2+ -DNA based bio-FET were also investigated in the four detecting steps, as shown in Fig. 4c. We confirmed that the overall change in current level was similar to previous results and the current level returned to its initial shape after the 4 th erasing step. Then, we extracted the off-current values at V GS = − 30 V in each detecting cycle. These data are presented in Fig. 4d, where the change in off-current level in each step was repeated in all detecting cycles. For reference, this reusability analysis was repeated for ten detecting cycles and these results can be found in the Supporting information Figure S4. Finally, we changed the concentration of doxorubicin in six detecting cycles and then plotted the extracted sensitivity values in Fig. 4e. We first flowed a low concentration of doxorubicin (10 −4 μ M) in the first two cycles, and then increased the concentration up to 50 μ M. As a result, a sensitivity of approximately 1.7, which was observed in the first two cycles, dramatically increased to 1.7 × 10 3 in the 3 rd and 4 th cycles. In the final two cycles, the sensitivity was reduced down to about 54 because we lowered the concentration to 10 μ M. Based on the observation that the sensitivity was almost same when the same concentration of doxorubicin was used, the proposed Cu 2+ -DNA/MoS 2 bio-FET seems to provide reproducible results. We also compared our work with the previously reported biosensors in terms of materials, type of biosensor, detection molecules/range, sensitivity, and reusability (Supporting information Table S1).

Conclusions
In conclusion, we demonstrated a Cu 2+ -DNA/MoS 2 based bio-FET with an extremely high sensitivity of 1.7 × 10 3 A/A. The high sensitivity was accomplished because the Cu 2+ -DNA nanostructures had a positive polarity, and these improved the detecting ability for doxorubicin-like molecules with negative polarity as compared to the DNA nanostructures alone. In addition, the short distance between the biomolecules and the sensor surface, which could be achieved without using high-κ dielectric layers in bio-FETs, also contributed to the high sensitivity. We first found the optimum conditions for the formation of Cu 2+ -DNA by considering the deformation phenomenon of Cu 2+ -DNA according to the concentration of doxorubicin. We then studied the influence of doxorubicin on Cu 2+ -DNA/MoS 2 structure using Raman spectroscopy and Kelvin probe force microscopy.  concentrations (each cycle consists of four steps; 1 st and 2 nd preparation, 3 rd sensing, and 4 th erasing steps). This excellent reusability has not been reported previously for 2D biosensors.

Experimental Methods
Fabrication of DX DNA Lattice. High-performance liquid chromatography purified synthetic oligonucleotides of DNA strands were purchased from Bioneer (Daejeon, Korea). Two DX tiles were used for the fabrication of a DX DNA nanostructure using a conventional free solution annealing process. Complexes were formed by mixing a stoichiometric quantity of each strand in physiological 1 × TAE/Mg 2+ buffer (40 mM Tris base, 20 mM acetic acid, 1 mM EDTA (pH 8.0), and 12.5 mM magnesium acetate) to form the DX structure. The complexes were cooled slowly from 95 to 25 °C to facilitate hybridization after placing the microtubes in 2 L of boiling water in a Styrofoam box for at least 24 h. The final concentration of DX DNA lattices was 200 nM.
Absorbance Measurements. A Varian Cary 5G spectrophotometer was used to conduct the optical absorbance measurements of the DNA, DNA with doxorubicin, Cu 2+ -DNA, and Cu 2+ -DNA with doxorubicin in solution form (wavelengths between 800 and 200 nm). The spectrophotometer was equipped with two light sources: a deuterium arc lamp (near-infrared and visible) and a quartz W− halogen lamp (ultraviolet). It has also two detectors: a cooled PbS detector for the near-infrared region and a photomultiplier tube for the visible and ultraviolet regions. The spectrophotometer measure the frequency-dependent light intensity passing either through a vacuum or through the sample. AFM Measurements. AFM images were taken by pipetting 5 μ L of the solutions on freshly cleaved mica, after which 30 μ L of 1× TAE/Mg 2+ buffer was pipetted onto the mica surface and another 10 μ L of 1× TAE/Mg 2+ buffer was dropped onto the AFM tip (Veeco Inc.). All AFM images were obtained on a Digital Instruments Nanoscope III (Veeco, USA) with a multimode fluid cell head in tapping mode using NP-S oxide-sharpened silicon nitride tips (Veeco, USA).
Characterization of DNA/MoS 2 or Cu 2+ -DNA/MoS 2 without and with doxorubicin. First, DNA or Cu 2+ -DNA/MoS 2 /SiO 2 /Si samples were investigated and compared with a control sample (MoS 2 /SiO 2 /Si). Then, the DNA or Cu 2+ -DNA/MoS 2 /SiO 2 /Si samples with doxorubicin were measured using PL/Raman spectroscopy (Alpha300M + , WITec). Here, TMD bulk flakes with similar thicknesses (~32 nm) were selected to avoid the thickness effect. Raman spectroscopy with an excitation wavelength of 532 nm was used, where the laser beam size was approximately 0.7-0.9 μ m, and the instrumental spectral resolution was less than 0.9 cm −1 . An integration time of 5 seconds and a spectrometer with 1800 grooves/mm were used. For the KPFM measurement, a platinum/iridium (Pt/Ir) coated Si tip was used and the tip was calibrated on a HOPG surface. First, we calculated the work function of the KPFM tip (W tip − W HOPG = Δ V CPD_HOPG ) using the well-known work function of the HOPG (W HOPG = 4.6 eV) and the contact potential difference between tip and HOPG (Δ V CPD_HOPG = 324 meV). We then found the work function of the TMD layers with the calculated W tip (4.92 eV) value and the measured Δ V CPD_TMD between KPFM tip and TMD surface (W tip − W TMD = Δ V CPD_TMD ). (channel length and width were 5 μ m) on MoS 2 /SiO 2 /Si samples by optical lithography, followed by deposition of 10-nm-thick Ti and 50-nm-thick Au in an e-beam evaporator. Back-gated MoS 2 transistors were coated by DNA or Cu 2+ -DNA and were electrically analyzed using an HP 4155A semiconductor parameter analyzer (I D -V G ). The threshold voltage (V TH ) and sensitivity, which is calculated as (I off_after sensing − I off_before sensing )/I off_before sensing , were calculated from the I D -V G data. All drain currents (I DS ) were normalized by the channel width (W).