Multiplexed sensing of biomolecules with optically detected magnetic resonance of nitrogen-vacancy centers in diamond

Significance

Bioassay applications that use magnetic nanotags to convert specific molecular interactions into a detectable magnetic dipole represent a rapidly advancing field of nanotechnology to develop next-generation diagnostics while providing the capability for simultaneous noninvasive manipulation in an external magnetic field. Although absorption, autofluorescence, scattering, photobleaching, blinking, and chemical quenching are major noise components for fluorescent markers in thick and opaque heterogeneous biological samples, they are not the significant limiting factors for the magnetic detection systems using magnetic nanotags. Here, we report progress toward a multiplexed magneto-DNA assay platform with position-indexed hydrogel microstructures decorated for sequence-specific nucleic acid detection and a complementary magnetic-imaging system using nitrogen-vacancy quantum centers in diamond.

Abstract

In the past decade, a great effort has been devoted to develop new biosensor platforms for the detection of a wide range of analytes. Among the various approaches, magneto-DNA assay platforms have received extended interest for high sensitive and specific detection of targets with a simultaneous manipulation capacity. Here, using nitrogen-vacancy quantum centers in diamond as transducers for magnetic nanotags (MNTs), a hydrogel-based, multiplexed magneto-DNA assay is presented. Near–background-free sensing with diamond-based imaging combined with noninvasive control of chemically robust nanotags renders it a promising platform for applications in medical diagnostics, life science, and pharmaceutical drug research. To demonstrate its potential for practical applications, we employed the sensor platform in the sandwich DNA hybridization process and achieved a limit of detection in the attomolar range with single-base mismatch differentiation.

Thanks to their nanoscale size, MNTs exhibit a reversible, hysteresis-free magnetization and high magnetophoretic mobility, which make them attractive for biomedical applications (1). The extremely large surface-to-volume ratio allows efficient interaction with the surrounding medium. Their surface can be chemically modified to conjugate biomolecules of different sizes, including drugs, nucleic acids, and peptides (2, 3, 4). Compared to fluorescence-based labeling approaches, MNTs provide distinct advantages, such as lower background noise that is related to the negligible magnetic property of biological samples, stable detection without label-bleaching, and feasibility for integration to magnetophoresis-based external manipulation systems (5). Exploiting these properties, MNTs have already been used for a wide range of applications in biomedicine and the life sciences, including bioseparation (6), controlled drug delivery to living cells (7), hyperthermic cancer treatment (8), magnetogenetics (9), and NMR imaging (10).

Although the size and morphology of magnetic nanotags (MNTs) can be characterized with existing analytical techniques such as dynamic light-scattering (11) and high-resolution transmission electron microscopy (12), the methods to probe their magnetic property in a physiological environment are limited. Superconducting quantum interference devices (SQUIDs) can achieve very high sensitivities but have limited spatial resolution, exhibit a size in the millimeter scale, and require cryogenic temperatures to operate (13), making them not appropriate for biologically relevant systems. While force microscopy with a magnetic tip allows to achieve high spatial resolution, it is impractical for the investigation of biological samples (14). First, physical contact with the sample can contaminate and damage the scanning probe, and incorporation in closed microfluidic platforms is difficult. Second, the measured signal is not only due to the magnetic field from the target but also includes other interactions such as Van der Waals and electrostatic forces, which undermine the sensitivity. Besides, as with SQUID magnetometers, magnetic force microscopy represents a time-consuming and low-throughput mechanical-scanning process. Similarly, various other methodologies such as tunneling magnetoresistance (15), Hall effect sensors (16), and atomic vapor cells (17) have been widely used for the detection of local magnetic fields, but complex experimental conditions, difficulties in miniaturization, and measurements with very slow readout limit their application in the biomedical field. An effective imaging technique for multiplex magneto-DNA assay sensors should not rely on a single point sensor or require precise positioning of the probe hardware that scans a local environment to extract the magnetic field distribution. Here, we use a thin layer of nitrogen-vacancy (NV) quantum centers in a diamond chip to map the magnetic profile of hydrogel microstructures that form a magneto-DNA assay platform. The platform exploits a nucleic acid sandwich hybridization technique in which the target nucleic acid binds to a capture-DNA immobilized within the hydrogel microstructure network and a complementary reporter DNA conjugated with an MNT.

NV quantum centers, atomic-scale defects in diamond crystals, are another promising candidate for various applications in sensing, quantum optics, and metrology. Optically addressable spin states of NV quantum centers can be manipulated with an externally applied microwave signal (MW) and detected from the spin state–dependent photoluminescence (PL) signal. The NV centers are well suited for precise detection of physical quantities such as magnetic fields (18–20), electric fields (21), temperature (22), pressure (23), ion concentrations (24), and spin densities (25). The sensing capabilities of NV quantum centers under ambient conditions are extremely valuable for applications in biological systems.

Bulk, diamond-based sensors employing an ensemble of NV quantum centers have been used to detect action potentials at the single-neuron level in whole organisms (26), image magnetically labeled cell lines (27) and localize magnetosomes generated in magnetotactic bacteria (28). Recently a chemically modified diamond pillar array has been introduced for the detection of individual, nitroxide spin-labeled DNA duplexes immobilized within ∼10-nm detection range of shallowly embedded NV quantum centers (29). Although this study did not focus on the detection of unmodified DNA duplex, it demonstrates the detection capacity of NV centers for individual spin labels in aqueous buffer solutions when immobilized in close proximity.

In addition, in the form of single fluorescent nanodiamonds (FNDs), unique spin and optical properties of NV quantum centers have promoted the development of numerous fluorescent microscopy techniques such as superresolution microscopy (30), fluorescence resonance energy transfer experiments (31), long-term in vivo tracking (32), fluorescence lifetime imaging microscopy, and stimulated emission depletion microscopy (33). FNDs are also emerging as promising candidates for sensor applications operating at the nanoscale. They have been validated for nanoscale thermometry in living human cells with high resolution and precision (22). A recent study on a hybrid approach with magnetic nanoparticles and NV quantum centers in FNDs has shown that the sensitivity can be further enhanced with the critical magnetization near the ferromagnetic–paramagnetic transition temperature (34). It has been demonstrated that nanodiamonds with NV centers are nontoxic for cells and do not induce any detectable stress in organisms (35). They have been employed to investigate a wide range of biomedical processes, including early embryogenesis (36), organ development (37), and drug delivery (38). More recently, they have been used as ultra-sensitive fluorescent labels that can be manipulated by microwave irradiation for a better signal-to-background ratio and allow a detection limit of 8.2 × 10⁻¹⁹ molar for a biotin–avidin model (39).

Results

Diamond Chip with a Near-Surface, Thin NV Layer.

In this work, the optically detected magnetic resonance (ODMR) measurements were performed using a single-crystal diamond plate with high purity. An electronic grade chip of size 3 mm × 3 mm × 0.1 mm was fabricated by chemical vapor deposition with boron and nitrogen concentrations of less than 1 ppb and 5 ppb, respectively (Element Six, Ltd). Then the diamond chip was mounted on a carrier silicon wafer for nitrogen ion, ¹⁵N⁺, implantation process. The depth profile estimation of ¹⁵N⁺ in the diamond chip was carried out using Monte Carlo simulations in the Stopping and Range of Ions in Matter (SRIM) software platform. As shown in Fig. 1A, the mean depth of NV centers created by an implantation process with a dose of 10¹² cm⁻² and an acceleration energy of 15 keV was estimated to be 37 nm. The ion beam was applied with a tilt angle of 7° to minimize the ion-channeling effect. The implanted sample was vacuum annealed at 850 °C for 3 h to diffuse the generated vacancies and form a high density of NV quantum centers in the diamond chip. Then, another annealing process was applied, at 465 °C for 24 h in air, to enhance the ratio of negatively charged states.

Fig. 1.

(A) Stopping and Range of Ions in Matter (SRIM) simulation results for three-dimensional ion density distribution profile in diamond and depth of NV quantum centers created by ¹⁵N⁺ ion-implantation process. The estimated average depth of the NV quantum centers formed with the ions at 20-keV energy is about 37 nm. (B) Schematic of energy-level diagram of the negatively charged NV quantum centers. A laser beam at 532-nm wavelength polarizes the NV center into the triplet-excited state, ³E. After the phonon-relaxation process within the phonon-sideband (PSB), spin-conserving dipole transitions to the triplet-ground state, ³A, generates a PL signal at 637-nm wavelength. The notation |i> represents sublevels with spin quantum number ms = i. The nonradiative intersystem crossing (ISC) process involving the metastable singlet states, ¹A and ¹E, is represented with dashed black arrows. Compared to |0> spin state, |±1> states have a higher probability of nonradiative transition via ISC. The infrared transition between the singlet states at 1,042 nm is not in the detection band of the experimental setup used in this work. The electron spin resonance (ESR) of the NV quantum center is driven by an MW at the resonant frequency D and D±γB for zero-field splitting (ZFS) and Zeeman interaction, respectively. The Inset shows the atomic structure of an NV quantum center in the diamond lattice. (C) Full ODMR spectrum of ensemble NV quantum centers in the diamond chip, presenting eight resonance frequencies. The splitting value for each pair of resonance peaks is proportional to the bias magnetic field projection along with the corresponding NV quantum center orientation in the diamond lattice. (D) The excitation and readout scheme for the ODMR experiments with NV quantum centers. Continuous optical excitation is used to both initialize and readout the NV spin state. The frequency of the MW in one channel, $f_{-}$, is swept linearly over time, while in the other channel, it is kept at a stationary far-off resonant frequency, $f_{+}$. The synchronized amplitude modulation of these signals allows lock-in detection. The direct-current bias field, of strength $\approx$ 50 G from a permanent magnet, is aligned and applied to the ensemble NV quantum centers to resolve the resonant frequencies.

The NV quantum center is formed by a nitrogen atom and adjacent vacancy in the diamond lattice, and in the negatively charged state, it can be treated as a system with two unpaired electrons presenting trigonal C3v symmetry (40). The unpaired electrons form a ground-state triplet, S = 1, that strongly interacts with optical and microwave fields (14). The total ground-state Hamiltonian for the spin system of NV quantum center is given by (41):

$$H=D(s_z^2-s\left(s+1\right)/3)+E\left(s_x^2-s_y^2\right)+\mathit{\gamma }\overrightarrow{B}.\overrightarrow{S}+H_{SI}+H_I\text{,}$$

where $D$ is the zero-field splitting (ZFS) parameter due to the spin–spin interaction of two unpaired electrons, $E$ is local, strain-dependent ZFS parameter, $\mathit{\gamma }/2\mathit{\pi }\approx 28\hspace{0.17em}\mathit{GHz}/T$ is the spin gyromagnetic factor, $\overrightarrow{B}$ is the external vector magnetic field applied to the system, $\overrightarrow{S}$ is the vector of spin-1 Pauli matrices, $H_{SI}$ is the hyperfine coupling to the nitrogen, ¹⁵N, nucleus, and $H_I$ represents the nuclear Zeeman and quadrupole interactions. Neglecting the nuclear terms, the energy-level diagram of an NV quantum center is given in Fig. 1B. It presents a spin-conserving optical transition between the ground-state triplet, ³A, and excited-state triplet, ³E, which is responsible for the zero-phonon line at a wavelength of 637 nm. Due to the phonon sideband (PSB), an NV quantum center can undergo a transition to the excited state by a photon at 532-nm wavelength and then either relax directly to the ground state or follow a nonradiative path involving intermediate singlet states, ¹A-¹E. Given that m_s= ±1 spin states have a higher probability of undergoing intersystem crossing, ISC, the PL signal provides optical identification of the spin state. In absence of external fields or strain, when an MW at a frequency of $D=2.87 \mathit{GHz}$ is applied to the populated ground state, transitions to the degenerate spin sublevels, m_s= ±1, are induced. This leads to a Lorentzian decay of the PL signal and gives rise to ODMR. Under an external magnetic field along NV axis, the Zeeman interaction results in a linear splitting of m_s= ±1 spin states with a value equal to Δ$\approx 56\hspace{0.17em}\mathit{KHz}/\text{µ}T$. Since NV quantum centers can have four different crystallographic orientation in the diamond lattice, $\left[111\right], \left[1\overline{1}\overline{1}\right]$, $\left[\overline{1}1\overline{1}\right]$, and $\left[\overline{1}\overline{1}1\right]$, the full ODMR spectrum can exhibit up to eight resonance lines (Fig. 1C). While the probability of crystallographic orientations for NV centers in the diamond chip is similar, the optical dipole associated with the orientation along the [111] axis is parallel to the chip surface, thus maximizing the photon-collection efficiency for higher sensitivity (42, 43). The splitting value for each resonance line pair is proportional to the magnetic field component along with the crystallographic orientation. Therefore, a quantitative approach to resolve the magnetic field strength exerted along with a specific crystallographic orientation, employed in the present work, is the precise detection of the ODMR resonance peaks in the spectrum.

Magnetic-imaging with the fabricated thin NV quantum center layer was carried out using a camera-based, lock-in detection scheme (Fig. 1D). This approach is a promising technique to improve the signal-to-noise ratio for background-free imaging applications (44, 45). As the local magnetic field to be detected was low in amplitude, a bias magnetic field was first applied to split transitions, m_s = 0 to m_s= +1 and m_s= 0 to m_s= −1, and then to probe the local fluctuations for imaging. After the alignment of the bias magnetic field along the most-sensitive NV axis, the [111] orientation, microwave excitation was applied in a dual-channel mode in which the signal was swept across the corresponding resonant frequency for one channel and a far-off resonant frequency for the other. With the far-off frequency component, at which the MW has no ODMR response, the environmental noise associated with microwave radiation, including thermal effects, was canceled out. However, it is important to note that the technique does not account for temperature-dependent resonance frequency shifts. In parallel, sequential image frames that are synchronized to the microwave channels were processed to extract microwave frequency–dependent PL contrast images. This approach provides a built-in ability to self-correct for low-noise frequency fluctuations. In this way, the background noise (e.g., autofluorescence, absorption, and scattering of ambient light), excitation-related fluctuations (e.g., irregular spatiotemporal dynamics of the laser beam), and inhomogeneity in the Complementary Metal Oxide Semiconductor (CMOS) sensor response were effectively eliminated.

We postprocessed the PL signal detected through camera-based imaging to extract the magnetic field pattern on the NV layer in the diamond chip. Compared to scanning probe magnetometry approaches, camera-based imaging provides a multiplexed data acquisition process to reconstruct stray magnetic fields from a large field of view in a very short period of time. On a pixel-by-pixel basis, the ODMR data were curve fitted to map the distribution of the shifts in the resonance frequency. Given that the shape of the measured contrast in the PL signal follows a double-Lorentzian model, $I\left(f\right)$, a custom set of Matlab scripts were developed to apply a least-squares algorithm and extract the fitting parameters for each pixel data in the image stack:

$$I\left(f\right)=\frac{p_1/2}{1+{\left(\frac{f+\mathit{\delta }/2-p_2}{p_3}\right)}^2}+\frac{p_1/2}{1+{\left(\frac{f-\mathit{\delta }/2-p_2}{p_3}\right)}^2}\text{,}$$

where $f$ is the frequency of the microwave excitation, $\mathit{\delta }=3.05$ MHz is the splitting term reflecting the hyperfine interaction with the ¹⁵N nucleus, $p_1$ is the amplitude of the PL contrast, $p_2=\mathit{\gamma }B$ is the shift in the resonance frequency due to the local magnetic field strength, $B, p_3$ is the full-width at half maximum of Lorentzian profile, and $\mathit{\gamma }$ is the electron spin gyromagnetic ratio.

Hydrogel Microstructures Indexed by Position.

To fabricate DNA-incorporated hydrogel microarrays, we built a glass, capillary-based, noncontact-print setup (Methods) to deliver picoliter volumes of reagents onto a target surface. With that, we were able to spot micrometer-scale poly(ethylene glycol) diacrylate (PEG-DA, Mw: 700 g/mol)–based hydrogel prepolymer solution droplets indexed by position, whereby each synthesized sensor element in the array can be chemically modified for a specific function. Compared to a planar surface, a hydrogel structure enables the integration of sensing moieties within a three-dimensional matrix, which improves the performance of the sensor system by increasing target interactions and immobilization capacity. In particular, PEG-based hydrogel scaffolds provide a unique platform to accommodate a wide range of detection configurations, while filtering nonspecific binding of biomolecules, allowing for a greater signal-to-noise ratio (46). The elasticity of the hydrogel networks at high water content and the lack of complex surface-attachment chemistry, which is the case for planar surface arrays, render them attractive for DNA-based sensing systems, as the ability to form Watson–Crick base pairs is preserved (47, 48). To this end, PEG-DA hydrogels were covalently connected with DNA molecules by copolymerization with acrylamide-functionalized oligonucleotides as recognition elements to probe specific molecular interactions. As shown in Fig. 2, we employed these hydrogel microstructures to transduce DNA hybridization events into a measurable, effective magnetic dipole that can be detected through the ODMR signal from NV quantum centers.

Fig. 2.

Schematic illustration of the sandwich hybridization assay in a PEG-DA–based hydrogel microstructure. The microgel is immobilized on a glass substrate and serves as a carrier platform for 5′-acrydite–modified, capture-DNA oligonucleotides. Streptavidin-coated MNTs conjugated to biotin-modified reporter DNA molecules are trapped on the hydrogel when complementary target DNA sequences are added to the sample solution. The sequences (F1, F2, and F3) are given in SI Appendix, Table T1.

The Magnetic Property of a Hydrogel with Magnetic MNTs in an External Field.

We developed a dipole model to estimate the magnetic field strength generated by the immobilized MNTs in the hydrogel network. In this model, we assumed that the particles are uniformly distributed over the hemispherical shell of hydrogels (SI Appendix, Figs. S6 and S7) and each particle functions as an infinitesimal dipole contributing to the overall field. The magnetic scalar potential, ${\mathit{\varphi }}_{\mathit{dipole}}(\overrightarrow{\mathit{\rho }})$ generated at a point, $\overrightarrow{x}$, by a magnetic dipole is given by (49):

$${\mathit{\varphi }}_{\mathit{dipole}}(\overrightarrow{\overrightarrow{\mathit{\rho }}})=\frac{\overrightarrow{m}.\widehat{n}}{4\mathit{\pi }{\mid \overrightarrow{\mathit{\rho }}\mid }^2}\text{,}$$

where $\overrightarrow{m}$ is the magnetic dipole moment and $\widehat{n}$ the unit vector parallel to $\overrightarrow{\mathit{\rho }}$. Using this property, the magnetic scalar potential, in SI units and spherical polar coordinates of the uniformly magnetized hemispherical shell given in Fig. 3A, can be determined by integrating the contributions of the infinitesimal dipoles, $\overrightarrow{m}$, forming the shell:

$$\mathit{\varphi }\left(k\right)=\frac{M}{4\mathit{\pi }}\underset{0}{\overset{\mathit{\pi }}{{\displaystyle \int }}}\underset{0}{\overset{\mathit{\pi }}{{\displaystyle \int }}}\underset{R_2}{\overset{R_1}{{\displaystyle \int }}}\frac{k-r\cos \mathit{\theta }}{{(r^2+k^2-2rk\cos \mathit{\theta })}^{3/2}}{r}^2\sin \mathit{\theta }\mathit{drd}\mathit{\theta }d\mathit{\phi }\text{,}$$

where $M$ is the magnetization value in $\widehat{z}$ and $k$ is the spacing between the detection point and the glass substrate to which the hydrogels are tethered. Using Legendre polynomial, we find the general solution as $\mathit{\varphi }\left(r,\theta \right)=\frac{M(R_1{}^3-R_2{}^3)\cos \mathit{\theta }}{6r^2}$. This provides the corresponding magnetic field as

$$\overrightarrow{B}\left(r,\theta \right)=\nabla \text{x} \mathit{\varphi }\left(r,\mathit{\theta }\right)=\frac{\text{Δ}VM{\mu }_o}{8\mathit{\pi }{r}^3}\left(2\cos \mathit{\theta }\widehat{r}+\sin \mathit{\theta }\widehat{\mathit{\theta }}\right),$$

where µ₀ is the magnetic permeability of free space and ΔV is the volume of the shell, which is equivalent to an effective dipole located on the glass substrate. As shown in Fig. 3 B–D, simulation results for 100 MNTs of size 50 nm with a mass magnetization of 80 emu/g and a density of 1 g/mL yield magnetic field strengths at various distances from the glass substrate in the range of microtesla (µT), which is readily detectable with the ODMR signal from NV quantum centers in diamond.

Fig. 3.

(A) Schematic representation of model hydrogel microstructure with the immobilized MNTs acting as infinitesimal point-like magnetic dipoles. Depending on their spatial position in the populated shell, each contributes to the magnetic field strength on the NV quantum centers in diamond. The shell is defined through homocentric hemispheres with radius, ${\text{R}}_1$ and ${\text{R}}_2$, while $\text{k}$ is the spacing value between the glass substrate hosting the hydrogel and the NV quantum centers layer in the diamond. $r$ and $\theta$ denote radial and polar parameter values in the spherical coordinate system, respectively. The particles are exposed to an external magnetic field, resulting in a longitudinal magnetization value of $M$. (B) Simulation results for the magnetic field pattern of the effective magnetic dipole produced by 100 MNTs of size 50 nm and mass magnetization of 80 emu/g given in the model shown in A. The color bar shows the strength of the magnetic field in microtesla (μT) units. The magnetic field strength applied to NV quantum centers plane k = 1 μm above the glass substrate is in the range of 0 to 150 μT. (C) For a distance of k = 5 μm, the magnetic field strength is in the range 0 to 1 μT. (D) When the distance is increased to k = 10 μm, the magnetic field profile broadens in the transverse plane as the magnetic field strength falls to the 0 to 0.15 μT range.

Magnetic Profile Measurements with NV Quantum Centers.

To probe the target interactions by magnetic labeling, the prepared hydrogel microstructures incorporating capture-DNA molecules were incubated in 20 µL of phosphate-buffered saline (PBS) containing various concentrations of target DNA molecules. The hydrogels were then rinsed with PBS buffer to remove unbound or loosely bound components from the surface. As shown in Fig. 4A, the hydrogels were brought into the close vicinity of the NV quantum centers by integrating the supporting glass substrate with the diamond chip platform. The measurements were performed using the custom ODMR experimental setup described in Methods. Conjugated MNPs immobilized in the hydrogel network were magnetized by applying an external magnetic field that is simultaneously used to separate the resonance lines in the ODMR spectrum. After resolving the most-sensitive resonance line in the spectrum, arising from [111]-oriented NV centers, using a single photon-counting module (SPCM), a 10-dBm MW was frequency modulated across this line using a parametric sweeping method with 10-kHz step size, 50-ms dwell time, and 10-MHz frequency range. The time-resolved PL signal from NV quantum centers was detected with a CMOS camera operating at 400 fps, and images were postprocessed on a pixel-by-pixel basis using the aforementioned curve-fitting approach, which extracts shifts in resonance frequency due to local magnetic field fluctuations. As shown in Fig. 4, we successfully mapped the magnetic field profile of hydrogels with the conjugated MNTs after incubation with target DNA solutions at very low concentrations. The maximum ODMR shift in the mapped profile, reflecting the maximum magnetic field strength applied to the NV centers within the associated pixel, is positively correlated with target concentration (Fig. 4B). The saturation effect at high concentrations is attributed to complete hybridization of MNPs to capture-DNA strands available at the surface of the hydrogels. The limit of detection is around 100 attomolar, which is about six orders of magnitude improvement over the conventional fluorescent labeling (SI Appendix, Fig. S1), and the magnetic field pattern is matching well with the simulation results, showing an effective magnetic dipole property. It is important to note that the high number of biotin-binding sites provided by MNPs increases the stability of the biotin–streptavidin system, which is essential at very low target concentrations. To test the selectivity of the sensor platform, we also performed measurements using the PBS solution of the target DNA molecules with a single-base mismatch in the sequence. As shown in Fig. 4, the magnetic image does not present a magnetic field pattern with the designed hydrogels when the buffer contains DNA molecules with a single-base mismatch, similar to measurements made with PBS buffer without target DNA molecules.

Fig. 4.

(A) Schematic illustration of the diamond-based sensor chip assembled for ODMR measurements. The diamond chip with near-surface NV quantum centers is mounted on a glass substrate using crystal wax at the corners. The fabricated hydrogel microstructure array on a thin, glass coverslip is brought into contact with the NV layer and secured in place using Kapton polyimide tape at the edges. The microwave antenna is aligned and placed on the glass coverslip through a three-dimensional positioning stage without blocking optical access for excitation and detection. (B) Maximum ODMR frequency shift in response to target DNA concentration. The experimental data are fitted to a Hill function (red dash line) with 10% and 90% of the maximal response in the concentration range of 100 attomolar to 10 picomolar. (C) Bright-field optical image of the fabricated hydrogel microstructure array on glass slide via glass micro capillary–based spotting and photo-polymerization process. All hydrogels were prepared in the same batch on the same diamond chip. The hydrogel network incorporates acrydite-modified capture-DNA molecules that copolymerize with PEG-DA monomers. Silanization of the glass surface with methacrylate moieties provides a stable, micropatterned hydrogel after exposure to UV light. (D) Measured magnetic field profile of the hydrogel microstructure, shown by the zoomed-in bright-field view in C, which is incubated in a PBS solution with target DNA molecules at a concentration of 1 femtomolar and magnetically labeled reporter DNA molecules at a concentration of 10 micromolars. The magnetic field projection on NV quantum centers along the sensitive crystallographic axis in the diamond is reconstructed on a pixel-by-pixel basis through the shifts in ODMR frequencies. (E) As a control experiment, the magnetic field profile of the hydrogel microstructure (incorporating capture-DNA molecules) incubated in PBS solution containing only the probe DNA conjugated MNPs was recorded. (F) Magnetic field profile of the hydrogel microstructure when incubated in PBS solution that contains target DNA molecules with a single-base mismatch (SI Appendix, Fig. S2). As the result of the control experiment in E, the image does not present the magnetic field profile of an effective magnetic dipole, which reveals the selectivity of the sensor system toward the target molecules in the presence of interfering nonspecific compounds. The color bar represents the strength of the magnetic field in the microtesla (µT) unit. (Scale bar, 5 μm.)

Discussion

In this work, using position-indexed hydrogel microstructures and diamond-based, magnetic-imaging platform, we demonstrated a multiplexed sensor system for the DNA hybridization process with high sensitivity and selectivity. Compared to the optical response of a fluorophore attached to the probe DNA molecule, the magnetic response of a conjugated MNT allows for detection of the target without requiring any enzymatic amplification process. We achieved a detection limit in the attomolar range that is much lower than reported for optical and electrochemical DNA detection techniques in a planar surface configuration (50, 51). As efficient transducers for MNTs, NV quantum centers facilitate a precise detection mechanism, while the sandwich DNA-based hybridization process ensures high specificity in discriminating a single-base mismatch. Here, magnetic detection through the ODMR of NV quantum centers is a key element to overcome the challenges associated with optical detection. While biological samples (e.g., whole blood and urine) and complex physiological environments are subject to absorption, autofluorescence, and scattering, they do not present strong intrinsic magnetic susceptibility. Although microwave-modulated FNDs as biomarkers can significantly improve detection sensitivity and imaging capabilities, methods used in signal extraction require optical access to the FNDs and thus to the target sample. Given that the ODMR measurements with the lock-in detection mechanism presented in this work only probe the magnetic field–dependent fluorescence signal generated by the NV quantum centers that do not have direct contact with the sample, the optical noise inherent to fluorescent labeling such as photobleaching, blinking, and chemical quenching is not the main limiting factor. Therefore, we anticipate that this approach could introduce clinical diagnostic tests for screening genetic mutations, rare diseases, and infections such as the SARS-CoV-2.

A main limitation of the detection approach presented in this work is the transduction efficiency of the hydrogel architecture that converts the hybridization events into a detectable magnetic field. Therefore, covalent modification of the diamond chip for miniaturized, surface-tethered hydrogel structures might improve the coupling of the magnetic field to the NV layer for a lower detection limit. Fabrication of the diamond chip with higher sensitivity is another challenge. In situ nitrogen-doping and He⁺ ion-implantation process enabling narrower spin-resonance line widths (52), engineering preferentially aligned NV centers (43), and optimization of the photon-collection efficiency with waveguide architectures (53) are some potential approaches for further improvements.

Methods

ODMR Experimental Setup.

A schematic of the experimental setup for the ODMR measurements with NV quantum centers is given in Fig. 5A. A single-mode, fiber-coupled collimated laser beam (Laser Quantum, Ventus) with a wavelength of 532 nm and a maximum power of 1.5 W was used for the optical excitation of NV quantum centers in the diamond chip. For a tunable field of view in the sample plane, the laser beam was defocused through an achromatic lens (f = 100 mm) adjusting the illumination area at the focal plane. Using a microscope objective lens (20X Nikon CFI60 TU Plan), the laser beam was applied to the NV quantum centers layer after reflecting off a dichroic mirror (LaserMUX Dichroic Beam splitter 514 to 543). The PL signal from the NV layer is collected by the same objective lens, transmitted through a long-pass filter (Semrock, LP02-633RU-25), and directed into the detection optics. A flip mirror in the detection optics allows the signal to follow an imaging path or photon-counting detection path. While in the imaging path, the signal is focused on a CMOS sensor, in the other path, it is coupled to a single-mode optical fiber via intermediate optics. For the controlled excitation and the subsequent readout, a complementary electronic interface was developed to address, manipulate, and detect the spin states of NV quantum centers. A digital filter was implemented on a field-programmable gate array (FPGA) device to perform real-time analysis and control the experimental parameters. The FPGA device converts transistor–transistor logic (TTL) pulses from a single-photon avalanche diode (Excelitas, SPCM-AQRH-15-FC) into the spin state–dependent PL signal, generates TTL pulse trains to synchronize excitation and detection units, and processes data for the acquisition of the ODMR spectrum. Frequency-modulated MW, which is driving electron spin resonance, is generated through a dual-channel radio-frequency (RF) signal source (Windfreak SynthHD). The output of the generator amplitude modulated through an RF switch and fed into an RF amplifier (Mini-Circuits ZHL-16W-43-S+), followed by an RF waveguide connected to a microwave antenna. A circulator (Pasternack PE83CR1004) has been placed between the antenna and the amplifier to prevent high-power back reflections.

Fig. 5.

(A) Schematic drawing of the experimental setup used for the ODMR measurements with NV quantum centers. After passing through a defocusing lens, which adjusts the size of the beam in the sample plane, the laser beam at 532-nm wavelength is focused on the diamond chip with the microscope objective lens. An antenna is delivering the MW to coherently control the electron spins of NV quantum centers in the diamond chip. The microwave source has two independent channels for MWs that are amplitude and frequency modulated for lock-in detection. An RF switch is used to select the frequency channel through a field-programmable gate array (FPGA) device. A high-voltage amplifier is used to modulate the microwave power, while a circulator with a load is dumping the reflected power. The PL signal passes through a dichroic mirror and a long-pass filter for the detection units. The flipping mirror enables optical detection paths for the camera and a single photon-counting module (SPCM). The FPGA device is used to synchronize the RF switch and the CMOS camera through transistor–transistor logic pulses (TTL_b, TTL_c), to process the pulse train (TTL_a) generated by the SPCM and to communicate with the computer for control and post data processing. (B) Schematic drawing of the capillary-based microfluidic platform printing hydrogel microstructures. Under microscope alignment, the glass microcapillary and the position control of the substrate allow the fabrication of a specific prepolymer droplet pattern. A long-pass optical filter is used to block the transmission of photons inducing undesired UV photo-polymerization during the alignment and spotting process. The hydrogel precursor solution is delivered onto the substrate surface by a syringe pump driving the hydrodynamic flow inside the glass microcapillary. Subsequent formation of the hydrogel network, and surface-tethering is achieved using the UV photo-polymerization process.

Noncontact Microspotting Setup.

Borosilicate glass capillaries with an inner diameter of 0.5 mm and an outer diameter of 1 mm (Science Products) were used for the microarray spotting process. Before pulling, the capillaries were cleaned by sonicating in acetone, isopropanol, and deionized water, successively, and then dried in air at 60 to 70 °C. A programmable micropipette puller (P-1000, Sutter Instrument) with the following parameters was used to produce the tapered capillary needles with a tip size of 1 to 2 µm: Heat = Ramp+10, Pull = 15, Velocity = 30, Time = 255, and Pressure = 550. The ramp value is the reference temperature parameter required to melt the glass and depends on the ambient temperature and humidity.

Using a polymeric tube, the printing microcapillary needle was connected to a syringe pump to control the flow rate of the ejected prepolymer solution in the range of 0.01 to 1 mL/h. As shown in Fig. 5B, the position of the microcapillary needle on the prepared glass substrate was controlled through a three-dimensional translational stage. After the deposition of prepolymer solution in the form of microdroplets, the pattern was irradiated by ultraviolet (UV) light at 365 nm (∼20 μW/cm⁻²) to crosslink DNA-incorporated hydrogels and immobilize them on the substrate.

Data Availability

All study data are included in the article and/or SI Appendix.