Single-cell identification with quantum-enhanced nuclear magnetic resonance

TL;DR

This work presents a single-cell identification approach using quantum-enhanced NMR with diamond nitrogen-vacancy centers for label-free detection of intracellular proton signals, and lays the groundwork for label-free sorting applications in rare cell analysis, personalized medicine, and single-cell diagnostics.

Abstract

Identification of individual cells within heterogeneous populations is essential for biomedical research and clinical diagnostics. Conventional labeling-based sorting methods, such as fluorescence-activated cell sorting and magnetic-activated cell sorting, enable precise sorting when reliable markers are available. However, their applicability is limited in cells lacking defined markers or sensitive to labeling, as labeling can compromise cellular viability and function. We present a single-cell identification approach using quantum-enhanced NMR with diamond nitrogen-vacancy centers for label-free detection of intracellular proton ($^1$H) signals. Using this method, we distinguish two human tumor cell lines by their proton spin-lattice ($T_1$) relaxation times, which serve as a cell-intrinsic physicochemical signature. It lays the groundwork for label-free sorting applications in rare cell analysis, personalized medicine, and single-cell diagnostics.

Figures (4)

• Figure 1: NV-enhanced single-cell NMR for label-free cell identification. (a) Single-cell identification workflow. Individual cells are measured with NV-enhanced NMR, where NMR contrasts such as proton spin-lattice relaxation enable label-free discrimination of cell types. A diamond-based microfluidic chip with a liquid channel is aligned below a diamond hosting near-surface NV centers and a microstructured surface region. NV spins are optically initialized and read out with laser pulses. Microwaves, delivered via a coplanar waveguide, drive the NV electron spin, while a radio-frequency (RF) coil addresses intracellular $^{1}\mathrm{H}$ nuclei . A static magnetic field $B_0$ sets the quantization axis. Time-domain signals are processed to extract quantitative NMR parameters, which serve as features for label-free classification. (b) Scanning electron micrograph of the diamond pillar array. Inset, higher-magnification view of a single diamond pillar. (c) Cross-sectional schematic of the single-cell NMR chip, showing the magnet, RF coil, glass coverslip, polydimethylsiloxane (PDMS) layer, diamond with pillar array, polymethyl methacrylate (PMMA) support and coplanar waveguide (CPW). Further structural details are provided in the Supplementary Information. (d) Principle of NV-based single-cell NMR, where an NV sensor in a diamond pillar detects the nuclear spins ($^1$H) of a single cell under an external magnetic field. Enlarged view of the dipolar interaction between an NV spin and nuclear spins within a cell. The interaction depends on the NV-nucleus separation $r$ and the relative angle $\theta$ to magnetic field (see Supplementary Information).
• Figure 2: Localization of the detected NMR signal to a single cell. (a) Optical micrograph of the diamond pillar array with cells sealed on top. Cells are highlighted with a blue pseudo-color overlay and the central band corresponds to the CPW. (b) NV fluorescence scan of the same region as a function of the lateral scan position, with dashed contours indicating the positions of individual cells. (c) Three-dimensional stack of NV fluorescence images at different focal planes, showing that the fluorescence contrast is confined to the NV pillars located beneath a single cell. (d) Zoom-in of the NV fluorescence map around one selected pillar used for single-cell measurements. (e) Integrated NMR signal strength as a function of the lateral integration radius around the NV sensor. The red marker indicates the radius at which half of the total signal is accumulated and the signal saturates for radii comparable to the cell size, demonstrating that the detected NMR signal is dominated by nuclear spins within a single cell.
• Figure 3: Single-cell NMR detection and identification of proton signals. (a) Top: The detection sequence consists of three parts: initialization with real-time feedback charge-state preparation, the ENDOR sequence with dynamical decoupling, and readout via SWAP operations that transfer the electron state to the nuclear spin for repetitive readout. Bottom: The full experiment begins with DNP to build nuclear polarization, followed by repeated sensing cycles. Polarization is sampled in each cycle and relaxes between cycles toward Boltzmann equilibrium with $T_1$. (b) Proton resonance spectra obtained at different magnetic fields, with the horizontal axis showing the applied RF frequency. Solid lines are fits to the data. (c) Linear fit of resonance frequency versus magnetic field yields a slope of 4.24(2) kHz/G, in agreement with the proton gyromagnetic ratio (4.26 kHz/G).
• Figure 4: Measurement and statistical analysis of single-cell relaxation times for label-free cell identification. (a) Representative $T_1$ relaxation curves for an individual MCF7 cell. Markers show a three-point moving average of the raw signal. Because the detected NMR signal exceeds the linear regime, solid curves are fits to the function $C_0\sin(Ae^{-t/T_1})$. The inset shows the reconstructed polarization decay $P(t)\propto e^{-t/T_1}$. (b) Representative $T_1$ relaxation curves for an individual HeLa cell. (c) Statistical comparison of nuclear relaxation times. Population statistics of $T_1$ show a significant difference between MCF7 and HeLa cells ($p=0.0049$, Mann-Whitney U test), with a large effect size (Cliff's $\delta=0.78$) and a median $T_1$ difference of about 36.5 ms (Hodges-Lehmann estimator). Mean $T_1$ values are $76(6)$ ms for MCF7 and $109(7)$ ms for HeLa. (d) Envisioned multi-parametric single-cell analysis with NMR-based microfluidics. Cells flow through microfluidic channels into an NV-based detection region, where intrinsic NMR signatures ($T_1$, $T_2$, chemical shift, NMRD) are measured for label-free identification and sorting. This platform can be extended to high-throughput analysis and integration with other single-cell omics.