Ultrasensitive detection of miRNA with an antimonene-based surface plasmon resonance sensor

MicroRNA exhibits differential expression levels in cancer and can affect cellular transformation, carcinogenesis and metastasis. Although fluorescence techniques using dye molecule labels have been studied, label-free molecular-level quantification of miRNA is extremely challenging. We developed a surface plasmon resonance sensor based on two-dimensional nanomaterial of antimonene for the specific label-free detection of clinically relevant biomarkers such as miRNA-21 and miRNA-155. First-principles energetic calculations reveal that antimonene has substantially stronger interaction with ssDNA than the graphene that has been previously used in DNA molecule sensing, due to thanking for more delocalized 5s/5p orbitals in antimonene. The detection limit can reach 10 aM, which is 2.3–10,000 times higher than those of existing miRNA sensors. The combination of not-attempted-before exotic sensing material and SPR architecture represents an approach to unlocking the ultrasensitive detection of miRNA and DNA and provides a promising avenue for the early diagnosis, staging, and monitoring of cancer.

B iomarkers have potential in the prediction, diagnosis and monitoring of diseases 1 . In particular, there is a need to discover biomarkers that can be used to detect diseases for which an early diagnosis is crucial or diagnosis is currently difficult 2 . MicroRNA (miRNA), which constitutes a class of short RNA, is emerging as ideal candidates as noninvasive biomarkers for applications in toxicology, diagnosis, and monitoring treatment responses or adverse events 3,4 . The aberrant expression of miRNA has been found in all types of tumours, including pancreatic cancer, lung cancer, prostate cancer, colorectal cancer, triple-negative breast cancer and osteosarcoma 4 . The detection of tumour-specific circulating miRNA at an ultrahigh sensitivity is of utmost significance for the early diagnosis and monitoring of cancer 5 . Unfortunately, miRNA detection remains challenging because miRNA are present at low levels and comprise ∼0.01% of the total RNA mass in a given sample. Therefore, the development of new approaches or sensing media for miRNA detection at the molecular level is urgently needed for clinical disease diagnosis.
Traditionally, the use of miRNA detection techniques, such as quantitative real-time PCR (qRT-PCR) 6 , northern blotting 7 and microarray-based hybridization 8 , is limited in early diagnosis in clinical practice due to the difficulty in amplification, the high cost, complex operations and low sensitivity. Due to its many advantages, such as non-destructive label-free detection, high reproducibility and low cost, the surface plasmon resonance (SPR) technique has proven to be versatile in investigations of molecular interactions by assessing the refractive index change on a chip surface [9][10][11] . Nevertheless, using the traditional SPR technique to detect biomolecules at very low concentrations remains challenging due to the limited quantity of immobilized probe DNA and miRNA on the chip surface (normally a thin gold film) [12][13][14][15][16][17] . Therefore, there is an urgent need to identify advanced material with large adsorption energy and work function increment to improve the performance of the SPR biosensor. Recently, numerous emerging two-dimensional (2D) nanomaterials have been tested for DNA molecule sensing, including graphene [18][19][20] , transition-metal dichalcogenides (TMDs) 21,22 , topological insulators 23,24 , black phosphorus 25,26 and MXenes 27 . However, most nanomaterials are subject to certain limitations due to weak interactions with biomolecules or poor chemical stability. Identifying a new 2D material with a stronger molecular-level interaction with biomarkers is critical.
Antimonene has been described as a 2D material that can be exfoliated from bulk antimony (Sb) and has quickly attracted the attention of the scientific community because its physicochemical properties are superior to those of typical 2D materials (e.g., graphene, MoS 2 , and black phosphorus) 28,29 . Similar to graphene materials, antimonene has an sp 2 -bonded honeycomb lattice, but antimonene exhibits strong spin-orbit coupling, tremendous stability and hydrophilicity that is significantly better than that of graphene 29,30 . Antimonene nanosheets and quantum dots have already been used in nonlinear optics 31 , photothermal therapy (PTT) 32 , thermophotovoltaic (TPV) cells 33 , and field effect transistors (FET) 34 . Although the photoelectrical properties of antimonene nanomaterials have been studied, the interaction between DNA and antimonene and its application in optical sensing remain elusive.
Here, we firstly explored via first-principles density functional theory (DFT) calculations the chemical interactions of singlestranded DNA (ssDNA) and double-stranded DNA (dsDNA) with antimonene, and find that antimonene has much better sensitivity than graphene previously used in DNA molecule sensing. Motivated by this theoretical finding, we developed a SPR sensor by using antimonene materials and performed trace attomolar-level quantification of miRNA molecules. This method reached an extremely low limit of detection (LOD), surpassing that of existing sensing methods. In addition, the sensor can distinguish miRNA that differ by one nucleobase mutation. Because of the extremely large adsorption energy between ssDNA and antimonene, we can envision an ultrasensitive RNA and DNA sensor device for early cancer diagnosis. This proposed methodology based on antimonene materials for nucleic acid detection holds intriguing potential for the development of multiplexed lab-on-chip platforms, which can be further applied for clinical purposes.

Results
First-principles calculations. DFT-based energetic calculations including dispersive Van der Waals forces were performed to investigate the differential interactions of ssDNA and dsDNA with antimonene, as summarized in Fig. 1. The changes in work function after ssDNA/dsDNA absorption are shown in Table 1. The interactions with graphene were studied for direct comparison. The adsorption energies (E ad ) of the bases on antimonene are higher than those of the base-pairs on antimonene, indicating the stronger interaction between the nucleobases and antimonene. The work function (ΔW) shows substantial increase after DNA absorption. These behaviors are consistent with those of current and previous calculations of graphene-based systems 35 . Distinctly and importantly, we found that by comparison with graphene, antimonene exhibits the much stronger interaction with ssDNA, as indicated by the higher adsorption energies (for the ssDNA absorption case, Fig. 1g) as well as the about 1.5 times larger work function increment (Table 1). This is further supported by the charge density difference map between antimonene/graphene with DNA absorption and noninteracting counterparts (Fig. 1h), where the stronger charge transfer and electronic orbital hybridization occurs in the antimonene case (upper panel). Further calculations of adsorption energies and work functions for the nucleobases on top of antimonene/graphene with varied adsorption orientations indicate that the bases adsorption orientations have negligible effect on the above results obtained (see Supplementary Fig. 1). The underlying mechanism might be related to the more delocalized 5s/5p orbitals or the buckling honeycomb lattice of antimonene. These results indicate that antimonene is more sensitive to ssDNA than graphene in terms of sensing ability.
Methodology of an antimonene-based miRNA sensor. The strategy adopted for the highly sensitive detection of miRNA hybridization events on antimonene-modified SPR chips is depicted in Fig. 2. The SPR signal is sensitive to changes in the refractive index of the analyte. First, AuNRs are employed to connect with ssDNA to amplify the SPR signal. Then, AuNR-ssDNA complex is adsorbed onto the antimonene nanosheet due to the strong interaction between ssDNA and antimonene. Following the addition of complementary miRNA, the hybridized targets are easily desorbed from the antimonene interface since double-stranded DNA has a low affinity to antimonene. The amount of miRNA can be typically determined based on the negative shift of the SPR signal. The investigation of AuNR-ssDNA-modified antimonene interfaces is beneficial for detecting miRNA hybridization events.
Preparation of the few-layer antimonene. Liquid-phase sonication is an effective method for preparing few-layer antimonene by breaking weak van der Waals forces. We obtained few-layer antimonene nanosheets using sonication liquid exfoliation as shown in Fig. 3a. The antimonene nanosheets consist of β-phase antimony based on the hexagonal coordinate system. The chemical composition and morphology of the prepared samples were systematically investigated. As shown in Fig. 3b, the Faraday-Tyndall effect, which is known as light scattering by particles in a fine suspension, was clearly observed, indicating the existence of antimonene nanosheets in the solution. Figure 3c shows a transmission electron microscope (TEM) image of the few-layer antimonene flake, in which very thin 2D nanosheets are resolved. The high-resolution TEM (HRTEM) image shows that the lattice is pure without any defects. The Fast Fourier Transform (FFT) image shows that the single-crystalline antimonene is a cubic system (Fig. 3d). The atomic force microscope (AFM) topography of typical antimonene nanosheets is shown in Fig. 3e. The overall lateral dimensions of the nanosheets are greater than 300 nm, and the thinnest piece is~3 nm thick.
The crystal structure of antimonene was confirmed by XRD spectrum 36 as depicted in Fig. 3f. The diffraction peaks of antimonene were identical to the spectrum of β-Sb precursor (JCPDS No. 35-0732). To further investigate the crystal structure and quality of the antimonene nanosheets, the Raman spectra of typical few-layer antimonene and bulk antimony were measured (Fig. 3g). Two characteristic Raman peaks, i.e., E g at 117 cm −1 and A 1g at 153 cm −1 , were observed in the Raman spectra of the few-layer antimonene. The degenerate modes of E g symmetry, which correspond to the in-plane transversal and longitudinal vibrations of the sublayers in opposite directions, cause the experimentally observed Raman peak at 117 cm −1 . The peak at 153 cm −1 is caused by the third mode opposite-in-phase out-ofplane vibrations of the sublayers of the A 1g symmetry 37 . Compared to the bulk material, in the few-layer system, a strong contraction of the in-plane lattice constant occurred as the film thickness decreased. Thus, the bulk A 1g mode blue shifted from 150 cm −1 to 153 cm −1 , and the film thickness decreased. The Sensitivity simulation and LBL assembly of antimonene. To investigate the key point of antimonene for improving SPR sensor performance, a numerical simulation was performed to evaluate the effect of the antimonene thickness on the sensitivity of the SPR sensor. The sensitivity can be defined as S = △θ/△n, which is the ratio of the change in the resonance angle to the change in the refractive index of analyte 38 . The electric field distribution is shown in Supplementary Fig. 2, in which further enhancment rather than immediate drop in the electric field is observed while four layers antimonene is used. Figure 4a shows the variation in sensitivity with respect to the refractive index of the sensing medium and the number of antimonene layers in the proposed SPR biosensor. The simulation results suggest that antimonene materials can greatly improve the sensitivity of the SPR sensor. Figure 4b shows the variation in the sensitivity of the antimonene-based miRNA SPR sensor concerning the antimonene layer when the refractive index of the sensing medium is 1.37 + △n. The sensitivity first increases to the maximum (171°R IU −1 ) when the number of antimonene layer is 4 and then begins to decrease when L > 4. The highest sensitivity was obtained with four layers of antimonene (details shown in Supplementary Information).
The above simulations provide clear guidance for the assembly of antimonene nanosheets on SPR chip surfaces. Experimentally, we assembled the antimonene nanosheets on a gold chip surface using a layer-by-layer technique. Figure 4c shows the representative AFM topography of antimonene nanosheets assembled on an Au chip surface. Although the surface of Au film exhibits rough, the distribution of the nanosheets is relatively uniform, and the average thickness of the nanosheets is~5 nm (see Supplementary  Fig. 2e). The consecutive build-up of the layer-by-layer (LBL) antimonene film was monitored by AFM ( Supplementary Fig. 2). The increase in the surface coverage and thickness as a function of the number of assembled layers indicates that a very uniform increase in the average layer thickness occurred during each dipping cycle. To further confirm the results of the antimonene assembly, contact angle experiments were carried out (Fig. 4d). After the antimonene nanosheets were assembled on the sensor chip, the contact angle was approximately 38°due to the high hydrophilicity of the antimonene material. The long-time chemical durability of antimonene nanosheets is outstanding (shown in Supplementary Fig. 3). Hence, we can control the assembly of antimonene films on SPR chip surfaces using layerby-layer assembly techniques.
miRNA sensing performance of antimonene. We measured the angle-resolved SPR spectra of the target miRNA-21 at very low concentration. The SPR responses upon the addition of complementary miRNA-21 are displayed in Fig. 5a. A prominent shift can be observed in the resonance angle, revealing the desorption of   The calculated work functions of isolated antimonene and graphene are 4.389 and 4.208 eV, respectively ARTICLE AuNR-ssDNA. The hybridization of the target miRNA-21 results in an obvious left-shift in the SPR angles even at very low concentrations of 10 −17 M. Using the same process but with unmodified-ssDNA, the shift in the SPR angle is barely detectable with the same concentrations (Fig. 5b). The calibration curve of the SPR angle shift versus the miRNA-21 concentration is shown in Fig. 5c. In the case of AuNR-ssDNA adsorption by antimonene, the detection limit of miRNA-21 was determined to be 10 aM according to the IUPAC guideline of a 3:1 signal to noise ratio. This detection limit is 10 5 times lower than that using the non-modified ssDNA. Figure 5d shows the real-time desorption process of AuNRss DNA, which reaches a plateau after 5 min (Fig. 5d). When one base mismatched miRNA was used, the SPR angle slightly shifted to the right relative to that of AuNR-ssDNA (Fig. 5e), indicating good selectivity. Thus, opposite signals are obtained for mismatched miRNA, indicating that mismatched miRNA also bound the antimonene surface instead of binding to the AuNR-ssDNA conjugates 39,40 . Importantly, we observed similar results in the detection of miRNA-155 and ssDNA ( Supplementary Fig. 4, 5), suggesting that the antimonene nanomaterials are universal for the detection of miRNA and ssDNA.
To further demonstrate the sensitivity of our device, we compared the LOD of the antimonene 2D materials with that of previously reported miRNA biosensors 41-47 as presented in Fig. 5f. Our antimonene-based miRNA SPR biosensor outperforms other miRNA biosensors based on conventional 2D nanomaterials [41][42][43][44][45][46][47] . In particular, the LOD of miRNA sensing can approach 10 aM (~30 molecules for a 5 μL sample), indicating the great potential of antimonene nanosheets in applications, such as single-molecule biological imaging, clinical therapy, and environmental monitoring 48,49 .

Discussion
The unprecedented high sensitivity of the SPR sensor not only relies on the strong interaction between antimonene and singlestranded DNA but also benefits from the enhanced coupling between the localized-SPR (LSPR) of the gold nanorods and  propagating-SPR of the gold film. The preparation of AuNRs is characterized by TEM images (Fig. 6a). As shown in Fig. 6b, AuNRs modified with ssDNA show red-shifted plasmon bands at 513 and 747 nm. The signal enhancement due to the electromagnetic field coupling between the plasmonic properties of the AuNRs and propagating plasmons was expected. As shown in Fig. 6c-e, the electromagnetic field intensity distributions of single AuNRs on gold films with a 5 nm thick antimonene spacer were calculated by the finite-difference time-domain (FDTD) method 50 . The FDTD simulation parameters were consistent with the experimental conditions 51 . Compared with conventional SPR using gold film, a considerable electromagnetic enhancement was observed based on the evanescent field excitation and LSPR of the AuNRs with certain thickness of antimonene nanosheets. The simulation results of the local electric field distribution around antimonene nanosheets with different thicknesses were shown in Supplementary Fig. 6, which further verify the trend of sensitivity shown in Fig. 4b. The electric field at the optimal gap (antimonene layer) between the gold film and gold nanorod is increased by~300 times at incident light wavelengths of 632.8 nm, where |E| = |E local /E in |, E local and E in are the local and incident electric fields, respectively. Thus, the sensor sensitivity is significantly improved by AuNRs.
In this paper, we demonstrated the efficiency and ultrasensitivity of the antimonene-based SPR sensor in the quantitative detection of cancer-associated miRNA. Specifically, we applied the sensor to detect miRNA-21 and miRNA-155, which are promising biomarkers for cancer diagnosis. This antimonene-based biosensor has a LOD of 10 aM, representing the highest sensitivity described thus far in miRNA detection based on direct detection and quantification of miRNA levels. More importantly, the signal amplification of the AuNRs and the interaction between antimonene and ssDNA/dsDNA were computationally and experimentally investigated. Consequently, the proposed biosensor represents the first methodology reported using antimonene materials for clinically relevant nucleic acid detection and constitutes an extraordinary opportunity for the development of lab-on-chip platforms. Nevertheless, for clinical and practical applications of the antimonene-based SPR sensor to be successfully used in early cancer diagnosis and the realization of pointof-care systems, future investigations of the specificity and high throughput are critically needed.

Methods
First-principles calculations. Calculations were performed using DFT-based planewave pseudopotential methods as implemented in the Vienna Ab initio Simulation Package 52,53 . We described the electron-ion interactions using the projected augmented wave pseudopotentials with 5s 2 5p 3 for Sb, 2s 2 2p 2 for C, 2s 2 2p 3 for N, 1s for H, and 2s 2 2p 4 for O as valence electrons 54 . The generalized gradient approximation formulated by Perdew, Burke, and Ernzerhof was used as the exchange correlation functional 55 . To simulate DNAs absorbing antimonene, a separation of 20 Å in the z direction was adopted to avoid interactions between adjacent antimonenes (in 6 × 6 or larger supercells). The nucleobase molecules were terminated with a methyl group to replace the sugar ring and generate an electronic environment as similar to a DNA chain as possible. The kinetic energy cut off for wave function expansion was set to 520 eV, and the single Г point of the supercell was used for sampling the electronic Brillouin zone. Equilibrium structures of DNAs absorbing antimonene were obtained through total energy minimization with the energy convergence threshold of 1 meV. We used the vdW-optB86b functional to properly take into account the long-range van der Waals interactions 56 .
FDTD simulation. The near-field distribution of AuNRs on gold film under a 632.8 nm incident laser was simulated by the FDTD method. To model the AuNRs on a gold film, the FDTD method was used with antisymmetric and symmetric boundary conditions at the x-y axis and z-y axis, respectively. The propagation of the plane waves was directed along 45°from the x axis. For all simulations, the parameters of the AuNRs nanostructures were set according to the average sizes (70 nm) measured from the experimental results. The mesh size was 1 nm. The electromagnetic field distribution of the Au nanorod was calculated at incident light wavelengths of 632.8 nm. The refractive index of Au used for the simulation was taken from the source program.
Synthesis of antimonene nanosheets. The antimonene nanosheets were prepared by probe-sonication liquid-phase exfoliation in ethanol. Pulverized antimony (Sb) powder at an initial concentration of 30 mg mL −1 was dispersed in a glass vial containing ethanol. Subsequently, the Sb powder solution was sonicated for 1 h in an ice-bath at 450 W and 22 kHz with ultrasound probe 0.5 s pulses. Then, the resulting solution was centrifuged at 1509.3×g for 10 min. Finally, the supernatant containing the antimonene nanosheets were carefully collected in a clean glass vial for future use.
Bioconjugation of AuNRs with ssDNA. The AuNRs were chemically modified with 5′-thiol-called oligonucleotides according to the procedure described by Mirkin et al. 57   Characterization. SPR measurements were performed with a commercially available Time-Resolved Surface Plasmon Resonance Spectrometer (DyneChem, China). TEM and HRTEM images were obtained under a JEM-3200FS microscope (JEOL, Japan). The Raman spectra were collected using an iHR 320 spectrometer (Horibai, Japan). The AFM images were taken under an L01F4C8 microscope (Bruker, Germany). The UV-visible spectroscopy was performed using Cary60 (Agilent, Malaysia). XRD was determined by a D8 Advance instrument (Bruker, Germany). The XPS data were collected using an ESCALAB 250Xi XPS spectrometer (Thermo Fisher, America). The contact angle experiments were performed with a Theta instrument (Biolin Scientific, Sweden).