Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Spin-enhanced nanodiamond biosensing for ultrasensitive diagnostics


The quantum spin properties of nitrogen-vacancy defects in diamond enable diverse applications in quantum computing and communications1. However, fluorescent nanodiamonds also have attractive properties for in vitro biosensing, including brightness2, low cost3 and selective manipulation of their emission4. Nanoparticle-based biosensors are essential for the early detection of disease, but they often lack the required sensitivity. Here we investigate fluorescent nanodiamonds as an ultrasensitive label for in vitro diagnostics, using a microwave field to modulate emission intensity5 and frequency-domain analysis6 to separate the signal from background autofluorescence7, which typically limits sensitivity. Focusing on the widely used, low-cost lateral flow format as an exemplar, we achieve a detection limit of 8.2 × 10−19 molar for a biotin–avidin model, 105 times more sensitive than that obtained using gold nanoparticles. Single-copy detection of HIV-1 RNA can be achieved with the addition of a 10-minute isothermal amplification step, and is further demonstrated using a clinical plasma sample with an extraction step. This ultrasensitive quantum diagnostics platform is applicable to numerous diagnostic test formats and diseases, and has the potential to transform early diagnosis of disease for the benefit of patients and populations.

Access options

Rent or Buy article

Get time limited or full article access on ReadCube.


All prices are NET prices.

Fig. 1: Schematic illustration of the use of FNDs in LFAs.
Fig. 2: Microwave modulation of FNDs on paper.
Fig. 3: Characterizing the fundamental limit of detection using biotin–avidin binding of FNDs on LFAs.
Fig. 4: Single-copy detection of HIV-1 RNA on LFAs using RT–RPA and FNDs.

Data availability

The datasets generated during and/or analysed during the current study, and the computer code used are available from the corresponding author on reasonable request, in line with the requirements of UCL and the funder (EPSRC policy framework on research data).


  1. 1.

    Childress, L. & Hanson, R. Diamond NV centers for quantum computing and quantum networks. MRS Bull. 38, 134–138 (2013).

    CAS  Google Scholar 

  2. 2.

    Mochalin, V. N., Shenderova, O., Ho, D. & Gogotsi, Y. The properties and applications of nanodiamonds. Nat. Nanotechnol. 7, 11–23 (2012).

    ADS  CAS  Google Scholar 

  3. 3.

    Boudou, J.-P. et al. High yield fabrication of fluorescent nanodiamonds. Nanotechnology 20, 235602 (2009).

    ADS  PubMed  PubMed Central  Google Scholar 

  4. 4.

    Schirhagl, R., Chang, K., Loretz, M. & Degen, C. L. Nitrogen-vacancy centers in diamond: nanoscale sensors for physics and biology. Annu. Rev. Phys. Chem. 65, 83–105 (2014).

    ADS  CAS  PubMed  Google Scholar 

  5. 5.

    Igarashi, R. et al. Real-time background-free selective imaging of fluorescent nanodiamonds in vivo. Nano Lett. 12, 5726–5732 (2012).

    ADS  CAS  PubMed  Google Scholar 

  6. 6.

    Leis, J., Martin, P. & Buttsworth, D. Simplified digital lock-in amplifier algorithm. Electron. Lett. 48, 259 (2012).

    ADS  Google Scholar 

  7. 7.

    Shah, K. G. & Yager, P. Wavelengths and lifetimes of paper autofluorescence: a simple substrate screening process to enhance the sensitivity of fluorescence-based assays in paper. Anal. Chem. 89, 12023–12029 (2017).

    CAS  PubMed  Google Scholar 

  8. 8.

    Childress, L. et al. Coherent dynamics of coupled electron and nuclear spin qubits in diamond. Science 314, 281–285 (2006).

    ADS  CAS  PubMed  Google Scholar 

  9. 9.

    Chang, H.-C., Hsiao, W. W.-W. & Su, M.-C. Fluorescent Nanodiamonds Ch. 11 (Wiley, 2018).

  10. 10.

    Yu, S. J., Kang, M. W., Chang, H. C., Chen, K. M. & Yu, Y. C. Bright fluorescent nanodiamonds: no photobleaching and low cytotoxicity. J. Am. Chem. Soc. 127, 17604–17605 (2005).

    CAS  PubMed  Google Scholar 

  11. 11.

    Shenderova, O. A. & McGuire, G. E. Science and engineering of nanodiamond particle surfaces for biological applications (review). Biointerphases 10, 030802 (2015).

    PubMed  Google Scholar 

  12. 12.

    Chang, Y. R. et al. Mass production and dynamic imaging of fluorescent nanodiamonds. Nat. Nanotechnol. 3, 284–288 (2008).

    CAS  PubMed  Google Scholar 

  13. 13.

    Maze, J. R. et al. Nanoscale magnetic sensing with an individual electronic spin in diamond. Nature 455, 644–647 (2008).

    ADS  CAS  PubMed  Google Scholar 

  14. 14.

    Balasubramanian, G. et al. Nanoscale imaging magnetometry with diamond spins under ambient conditions. Nature 455, 648–651 (2008).

    ADS  CAS  PubMed  Google Scholar 

  15. 15.

    Tetienne, J. P. et al. Magnetic-field-dependent photodynamics of single NV defects in diamond: An application to qualitative all-optical magnetic imaging. New J. Phys. 14, 103033 (2012).

    Google Scholar 

  16. 16.

    Acosta, V. M. et al. Temperature dependence of the nitrogen-vacancy magnetic resonance in diamond. Phys. Rev. Lett. 104, 070801 (2010).

    ADS  CAS  PubMed  Google Scholar 

  17. 17.

    Hsiao, W. W. W., Hui, Y. Y., Tsai, P. C. & Chang, H. C. Fluorescent nanodiamond: a versatile tool for long-term cell tracking, super-resolution imaging, and nanoscale temperature sensing. Acc. Chem. Res. 49, 400–407 (2016).

    CAS  PubMed  Google Scholar 

  18. 18.

    Vaijayanthimala, V. & Chang, H.-C. Functionalized fluorescent nanodiamonds for biomedical applications. Nanomedicine 4, 47–55 (2009).

    CAS  PubMed  Google Scholar 

  19. 19.

    Fu, C.-C. et al. Characterization and application of single fluorescent nanodiamonds as cellular biomarkers. Proc. Natl Acad. Sci. USA 104, 727–732 (2007).

    ADS  CAS  PubMed  Google Scholar 

  20. 20.

    Chang, B. M. et al. Highly fluorescent nanodiamonds protein-functionalized for cell labeling and targeting. Adv. Funct. Mater. 23, 5737–5745 (2013).

    CAS  Google Scholar 

  21. 21.

    Waddington, D. E. et al. Nanodiamond-enhanced MRI via in situ hyperpolarization. Nat. Commun. 8, 15118 (2017).

    ADS  PubMed  PubMed Central  Google Scholar 

  22. 22.

    Hegyi, A. & Yablonovitch, E. Molecular imaging by optically detected electron spin resonance of nitrogen-vacancies in nanodiamonds. Nano Lett. 13, 1173–1178 (2013).

    ADS  CAS  PubMed  Google Scholar 

  23. 23.

    Sarkar, S. K. et al. Wide-field in vivo background free imaging by selective magnetic modulation of nanodiamond fluorescence. Biomed. Opt. Express 5, 1190 (2014).

    PubMed  PubMed Central  Google Scholar 

  24. 24.

    Chapman, R. & Plakhoitnik, T. Background-free imaging of luminescent nanodiamonds using external magnetic field for contrast enhancement. Opt. Lett. 38, 1847 (2013).

    ADS  CAS  PubMed  Google Scholar 

  25. 25.

    Doronina-Amitonova, L., Fedotov, I. & Zheltikov, A. Ultrahigh-contrast imaging by temporally modulated stimulated emission depletion. Opt. Lett. 40, 725 (2015).

    ADS  CAS  PubMed  Google Scholar 

  26. 26.

    Bhutta, Z. A., Sommerfeld, J., Lassi, Z. S., Salam, R. A. & Das, J. K. Global burden, distribution, and interventions for infectious diseases of poverty. Infect. Dis. Poverty 3, 21 (2014).

    PubMed  PubMed Central  Google Scholar 

  27. 27.

    Global HIV & AIDS statistics — 2018 fact sheet. (UNAIDS, 2018).

  28. 28.

    May, M. et al. Impact of late diagnosis and treatment on life expectancy in people with HIV-1: UK Collaborative HIV Cohort (UK CHIC) Study. Br. Med. J. 343, d6016 (2011).

    Google Scholar 

  29. 29.

    Gray, E. R. et al. p24 revisited: a landscape review of antigen detection for early HIV diagnosis. AIDS 32, 2089–2102 (2018).

    CAS  PubMed  PubMed Central  Google Scholar 

  30. 30.

    Price, C. P. Point of care testing. Br. Med. J. 322, 1285–1288 (2001).

    CAS  Google Scholar 

  31. 31.

    World Malaria Report. (WHO, 2018).

  32. 32.

    Land, K. J., Boeras, D. I., Chen, X. S., Ramsay, A. R. & Peeling, R. W. REASSURED diagnostics to inform disease control strategies, strengthen health systems and improve patient outcomes. Nat. Microbiol. 4, 46–54 (2019).

    CAS  PubMed  Google Scholar 

  33. 33.

    Walter, J. G. et al. Protein microarrays: reduced autofluorescence and improved LOD. Eng. Life Sci. 10, 103–108 (2010).

    CAS  Google Scholar 

  34. 34.

    Kim, J. et al. Rapid and background-free detection of avian influenza virus in opaque sample using NIR-to-NIR upconversion nanoparticle-based lateral flow immunoassay platform. Biosens. Bioelectron. 112, 209–215 (2018).

    CAS  PubMed  Google Scholar 

  35. 35.

    Paterson, A. S. et al. A low-cost smartphone-based platform for highly sensitive point-of-care testing with persistent luminescent phosphors. Lab Chip 17, 1051–1059 (2017).

    CAS  PubMed  PubMed Central  Google Scholar 

  36. 36.

    Boudou, J. P., David, M. O., Joshi, V., Eidi, H. & Curmi, P. A. Hyperbranched polyglycerol modified fluorescent nanodiamond for biomedical research. Diamond Relat. Mater. 38, 131–138 (2013).

    ADS  CAS  Google Scholar 

  37. 37.

    Hermanson, G. T. Zero-length crosslinkers. In Bioconjugate Techniques 3rd edn (ed. Hermanson, G. T.) Ch. 4 (Academic, 2013).

  38. 38.

    González Flecha, F. L. & Levi, V. Determination of the molecular size of BSA by fluorescence anisotropy. Biochem. Mol. Biol. Educ. 31, 319–322 (2003).

    Google Scholar 

  39. 39.

    Reth, M. Matching cellular dimensions with molecular sizes. Nat. Immunol. 14, 765–767 (2013).

    CAS  PubMed  Google Scholar 

  40. 40.

    Ngom, B., Guo, Y., Wang, X. & Bi, D. Development and application of lateral flow test strip technology for detection of infectious agents and chemical contaminants: a review. Anal. Bioanal. Chem. 397, 1113–1135 (2010).

    CAS  PubMed  Google Scholar 

  41. 41.

    Armbruster, D. A. & Pry, T. Limit of blank, limit of detection and limit of quantitation. Clin. Biochem. Rev. 29, S49–S52 (2008).

    PubMed  PubMed Central  Google Scholar 

  42. 42.

    Daher, R. K., Stewart, G., Boissinot, M. & Bergeron, M. G. Recombinase polymerase amplification for diagnostic applications. Clin. Chem. 62, 947–958 (2016).

    CAS  PubMed  PubMed Central  Google Scholar 

  43. 43.

    Lillis, L. et al. Cross-subtype detection of HIV-1 using reverse transcription and recombinase polymerase amplification. J. Virol. Methods 230, 28–35 (2016).

    CAS  PubMed  PubMed Central  Google Scholar 

  44. 44.

    Crannell, Z. A., Rohrman, B. & Richards-Kortum, R. Equipment-free incubation of recombinase polymerase amplification reactions using body heat. PLoS ONE 9, e112146 (2014).

    ADS  PubMed  PubMed Central  Google Scholar 

  45. 45.

    Rohrman, B. A. & Richards-Kortum, R. R. A paper and plastic device for performing recombinase polymerase amplification of HIV DNA. Lab Chip 12, 3082 (2012).

    CAS  PubMed  PubMed Central  Google Scholar 

  46. 46.

    Boyle, D. S., Lehman, D. A. & Lillis, L. Rapid detection of HIV-1 proviral DNA for early infant diagnosis using rapid detection of HIV-1 proviral DNA for early infant diagnosis. MBio 4, e00135-13 (2013).

    PubMed  PubMed Central  Google Scholar 

  47. 47.

    Dineva, M. A., Mahilum-Tapay, L. & Lee, H. Sample preparation: a challenge in the development of point-of-care nucleic acid-based assays for resource-limited settings. Analyst 132, 1193 (2007).

    ADS  CAS  PubMed  Google Scholar 

  48. 48.

    Jauset-Rubio, M. et al. Ultrasensitive, rapid and inexpensive detection of DNA using paper based lateral flow assay. Sci. Rep. 6, 37732 (2016).

    ADS  CAS  PubMed  PubMed Central  Google Scholar 

  49. 49.

    Phillips, A. et al. Sustainable HIV treatment in Africa through viral-load-informed differentiated care. Nature 528, S68–S76 (2015).

    PubMed  PubMed Central  Google Scholar 

  50. 50.

    Kuo, Y., Hsu, T.-Y., Wu, Y.-C., Hsu, J.-H. & Chang, H.-C. Fluorescence lifetime imaging microscopy of nanodiamonds in vivo. In Proc. Advances in Photonics of Quantum Computing, Memory, and Communication VI Vol. 8635, 863503 (SPIE, 2013).

  51. 51.

    Kim, E. Y. et al. A real-time PCR-based method for determining the surface coverage of thiol-capped oligonucleotides bound onto gold nanoparticles. Nucleic Acids Res. 34, e54 (2006).

    PubMed  PubMed Central  Google Scholar 

  52. 52.

    Besnier, C., Takeuchi, Y. & Towers, G. Restriction of lentivirus in monkeys. Proc. Natl Acad. Sci. USA 99, 11920–11925 (2002).

    ADS  CAS  PubMed  Google Scholar 

  53. 53.

    Bainbridge, J. W. et al. In vivo gene transfer to the mouse eye using an HIV-based lentiviral vector; efficient long-term transduction of cornealendothelium and retinal pigment epithelium. Gene Ther. 8, 1665–1668 (2001).

    CAS  PubMed  Google Scholar 

  54. 54.

    Foley, B. et al. HIV Sequence Compendium 2017. LA-UR-18-25673 (Los Alamos National Laboratory, 2018).

  55. 55.

    Kong, J. & Yu, S. Fourier transform infrared spectroscopic analysis of protein secondary structures. Acta Biochim. Biophys. Sin. (Shanghai) 39, 549–559 (2007).

    CAS  Google Scholar 

  56. 56.

    Zadeh, J. N. et al. NUPACK: Analysis and design of nucleic acid systems. J. Comput. Chem. 32, 170–173 (2011).

    CAS  PubMed  Google Scholar 

  57. 57.

    SantaLucia, J. A unified view of polymer, dumbbell, and oligonucleotide DNA nearest-neighbor thermodynamics. Proc. Natl Acad. Sci. USA 95, 1460–1465 (1998).

    ADS  CAS  PubMed  Google Scholar 

  58. 58.

    Laitinen, M. P. & Vuento, M. Affinity immunosensor for milk progesterone: Identification of critical parameters. Biosens. Bioelectron. 11, 1207–1214 (1996).

    CAS  PubMed  Google Scholar 

Download references


We thank M. Schormans for help with circuit design, M. Thomas for assistance with dynamic light scattering measurements and M. Towner for assistance with FTIR measurements. This work was funded by the i-sense EPSRC IRC in Early Warning Sensing Systems for Infectious Diseases (EP/K031953/1); i-sense EPSRC IRC in Agile Early Warning Sensing Systems in Infectious Diseases and Antimicrobial Resistance (EP/R00529X/1); the National Institute for Health Research University College London Hospitals Biomedical Research Centre; Royal Society Wolfson Research Merit Award to R.A.M. (WM130111); LCN Departmental Studentship to B.S.M.; EPSRC Centre for Doctoral Training in Delivering Quantum Technologies to G.D. (EP/L015242/1); H2020 European Research Council Local quantum operations achieved through the motion of spins to J.J.L.M. (771493); and the UCLH NHS Foundation Trust to J.H. and E.N.

Author information




B.S.M. and R.A.M. conceived the research and led the study; P.J.D. advised on nanodiamonds and J.J.L.M. on microwave modulation. B.S.M. demonstrated the initial proof-of-concept; B.S.M. and L.B. designed and optimized the lock-in analysis, functionalization and LFA design; B.S.M., L.B. and D.H. performed all the FND LFA experiments; H.D.G. designed, optimized and performed RT-RPA assays including primer design and template generation; D.H. adapted and performed RT-RPA assays and purification; J.J.L.M. and G.D. designed the microwave delivery including resonators; E.R.G. performed clinical RNA extraction, and advised on virology including primer design; J.H. performed qPCR on the seroconversion panel; E.N. provided clinical expertise; B.S.M. and E.R.G. designed and performed binding-site quantification experiments; B.S.M., L.B. and R.A.M. drafted the manuscript; and all authors reviewed and revised the manuscript.

Corresponding authors

Correspondence to Benjamin S. Miller or Rachel A. McKendry.

Ethics declarations

Competing interests

B.S.M., L.B., G.D., P.J.D., J.J.L.M. and R.A.M. are inventors on the UK patent application number 1814532.6 filed by University College London Business.

Additional information

Peer review information Nature thanks Takuya Segawa and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data figures and tables

Extended Data Fig. 1 Optimization of microwave modulation.

af, A linear resonator was designed to have a wideband response over the range 1–4 GHz, and an omega narrowband resonator was designed to have a stronger, narrower resonance at 2.87 GHz with quality factor Q = 100. The schematic printed circuit board layouts for the two resonators are shown in a and d, respectively. The resulting simulated fields are shown in b and e, respectively. The reflected power (S11) is plotted against frequency in c and f. The narrowband resonator shows 5–6 orders of magnitude greater absorption than the wideband resonator at 2.87 GHz, indicating resonant coupling giving strong absorption. Panel f also shows the corresponding FND intensity dip. g, Emission spectra of FNDs acted on by a 2.87 GHz microwave field. The powers listed in decibel-milliwatts are the output power of the microwave generator (before the 17 dB amplifier). h, Each spectrum is integrated over the whole wavelength range to give a total intensity, which is plotted against preamplifier power. This shows a linear relationship between fluorescence intensity and microwave power (in dBm) above a threshold power, and up to 7 dBm, where the amplifier reaches its 1 dB compression power. At this point, the fluorescence starts to increase again owing to a loss in the quality of the sinusoid leading to power lost in harmonics. Dots show means and error bars show the s.d., with n = 3 measurement repeats.

Extended Data Fig. 2 Optimization of lock-in analysis.

a, Schematic of the computational lock-in algorithm used to extract the microwave modulated FND signal from the background. The input signal is high-pass filtered using a moving average filter to remove low-frequency drift. It is subsequently multiplied by cosine and sine functions with frequency Fm, and the resulting signals are low-pass filtered to generate the in phase and quadrature components, respectively, of the vector representation of the signal. The magnitude of this vector is calculated to remove the effect of phase, giving the output magnitude. b, The variation of lock-in amplitude with modulation rate (Fm) at various sampling rates (Fs). A single strip with very high intensity was modulated at Fm values between 1–450 Hz, and sampled at various Fs values between 3.89–996 Hz. The resulting plot shows that lock-in amplitude is independent of Fs when Fs > 2Fm. c, d, The relationships between lock-in amplitude, exposure time (Te) and modulation frequency (Fm). An identical LFA strip was measured with exposure times between 10–50 ms, using the maximum possible Fs for each Te, and Fm values between 1–15 Hz. d shows Fm against lock-in amplitude at various exposure times. It is shown that the lock-in amplitude has its maximum at around 5 Hz for all frequencies, and reduces when Fm is close to Fs/2, its maximum possible value. This is evident in the raw signal plots in c for each Fm at a fixed exposure time of 30 ms. As Fm approaches Fs/2, the sampling effects obscure the square wave, decreasing lock-in amplitude. For maximum lock-in amplitude, the highest possible Te should be used. Here, we are limited to 50 ms by the background autofluorescence of the nitrocellulose, which saturates the camera above this value. A corresponding Fm of 4 Hz was chosen as it is in the optimal range and is a power of 2, so can be achieved by simply dividing the temperature compensation crystal oscillator (TCXO) frequency. e, The variation of lock-in amplitude with total measurement time at Fm = 4 Hz and Fs = 20 Hz for five different concentrations of FNDs and a negative control, immobilized with a biotin–avidin interaction. The positive amplitudes stabilize quickly, reaching 5% of their 15 s value in 3.9 s for positive results. The negative results take longer to stabilize, reaching 5% of their 15 s value in 13 s. A measurement time of 15 s (300 frames) was used for subsequent measurements. f, Schematic circuit design of temperature compensation crystal oscillator (TCXO)-based modulated microwave source. It is powered by a 5 V source which powers a TCXO, which outputs a 32.768 kHz square wave. This is converted to a 4 Hz signal by a 4060 counter chip. This square wave controls two transistors which deliver 12 V stepped up power (d.c. converter) to the microwave VCO. The bias voltage is regulated from 12 V to 8.15 V by a voltage regulator. The VCO microwave output is amplified by the MW amplifier and transmitted to the omega resonator. g, Printed circuit board layout of the prototype (65 mm × 38 mm). Outputs for the microwave amplifier and microwave VCO are at the top right and bottom right, respectively. A photo of the printed circuit board with a £1 coin (GBP) for scale is shown below.

Extended Data Fig. 3 FND characterization and functionalization.

a, Comparison of the non-specific binding of various commercial FNDs with various surface functionalizations on LFAs. The lock-in amplitude at the test line was measured to quantify non-specific binding. The LFAs were also pre-blocked with a polyvinylpyrrolidone-sucrose solution (proprietary method, Mologic). The lowest non-specific binding was from the PG-functionalized particles (FND–PG), as the PG adds a hydrophilic layer. b, Dynamic light scattering of three different FND particle core diameters: 120, 200 and 600 nm. c, A schematic of antibody functionalization of FND–PG. DSC activates hydroxyl surface groups to form succinimidyl carbonates, which can then react with antibodies to form stable carbamate or urethane bonds. df, Scanning electron microscope images of FNDs with particle core diameters of 120, 200 and 600 nm, respectively. g, Dynamic light scattering was also used to measure the size and aggregated fraction after functionalization of 120 nm FND–PG before and after functionalization with BSA-biotin or antibodies. Dots show the means of n = 3 measurement replicates. Fitting the number plots to skew exponentials (equation (3), plotted as lines) gave peak particle hydrodynamic diameters of 106, 121 and 128 nm. h, The fitted peak diameters are plotted with error bars denoting their 95% confidence intervals (n = 3 measurement replicates), showing no significant difference between the bio-functionalized diameters (FND–biotin, FND–antibody), but both are significantly different from the pre-functionalization diameter (FND–PG): *P ≤ 0.05; **P ≤ 0.01, ANOVA with Tukey HSD post hoc test. i, FTIR spectroscopy of FND–PG and antibody-functionalized FND–PG. Lines show means of n = 6 measurement replicates for FND samples and n = 2 measurement replicates for the blank. C–O and C–H peaks, indicative of the PG layer, can be seen in both FND–PG and FND–PG–antibody at around 1,100 cm−1 and at around 2,900 cm−1, respectively. The FND–PG-antibody spectrum displays additional peaks at around 1,640 cm−1 and at around 1,540 cm−1, suggesting protein amide I and amide II bonds, respectively55, showing that protein functionalization was successful.

Extended Data Fig. 4 Quantification of the number of available binding sites per FND.

a, Initially, binding constants of the anti-DIG antibody binding to DIG were measured using interferometry. Full experimental details are shown in Supplementary Information section 1. Binding at different concentrations was measured and the resulting curves were fitted to exponentials. To find the equilibrium dissociation constant (KD), equilibrium binding values, B, were plotted here against concentration, C. A Langmuir adsorption isotherm was fitted \(({B}^{{\rm{\infty }}}=\frac{a\times C}{{K}_{{\rm{D}}}+C})\) giving a KD value of 5.1 × 10−10 M. b, In order to find the on- and off-rates, kon and koff, the observed reaction rates, kobs, at each concentration were plotted and fitted to the linear relationship: kobs = koff + C × kon. The resulting fitted values are kon = 1.6 × 105 M−1 s−1 and koff = 9.1 × 10−5 s−1. c, A schematic of the assay to quantify the number of available binding sites per FND. After functionalization of FNDs with anti-DIG antibodies, an approximately 50-fold excess of DIG-modified DNA was added and left to bind for 2 h. The negative DNA control used the same sequence, but with no DIG modification to compensate for non-specific binding and adequate washing. After multiple washes by centrifugation to remove the excess DNA, the remaining DNA (bound to FNDs) was quantified by qPCR. See Extended Data Fig. 8d for template, primer and probe sequences, and Methods for full experimental details. d, A kinetic binding simulation was performed to verify that all available sites would be occupied after 2 h with the above excess. The graph shows the fraction of sites on the FNDs which are occupied, with this approximately 50-fold excess, over a range of KD, kon and koff values. The red cross in circle marks the location of the anti-DIG antibody used in this paper (using the values measured in a and b), indicating that more than 99.9% of available sites will be occupied after 2 h. This means that quantifying the DNA gives a true measure of available binding sites. e, Amplification plot showing the normalized fluorescence intensity against the number of cycles. A standard curve of each decade from 40 copies to 4 × 108 copies is plotted, along with the sample and negative control FND samples described above. The negative diluent controls are also plotted along with the Cq threshold. The lines show means and shaded areas show the s.d. of repeats (n = 3 technical replicates for standard curve, and n = 6 for samples). f, The resulting Cq values are plotted against copy number per reaction. Dots show means and error bars show s.d. (n = 3 technical replicates for standard curve and n = 6 for samples). The standard curve was fitted to a logarithmic curve (Cq = −3.2log10 copies + 39), enabling calculation of the number of copies in the DIG–DNA sample and negative DNA control. Dividing by the particle concentration (measured as shown in Extended Data Fig. 5c) and subtracting the negative DNA control value gives the number of available binding sites per particle as 4,300 sites. This is within what is geometrically plausible, giving an area per antibody of at least 200 nm2 (assuming at least 1 paratope available of at least 75% of the bound antibodies). The corresponding calculated values for 120 and 200 nm particles are 172 and 477 available binding sites per FND respectively, assuming the same loading density.

Extended Data Fig. 5 Lateral flow and FND benchmarking.

a, Measurement of flow rate of lateral flow strips. During wetting, the flow follows the Washburn equation, where \(V\approx {t}^{\frac{1}{2}}\) (inset), and during fully-wetted flow, Darcy’s law for capillary flow is followed (Vt), with a constant flow rate of 6.9 μl min−1. b, Using a one-to-one receptor–ligand binding approximation, the binding of biotinylated FNDs to streptavidin was modelled kinetically, indicating that all the FNDs bind with a residency time of more than about 10−3 s. Here, the residency time is measured as 4 s, using the flow rate from a, so all the FNDs should bind. c, An example of the measurement of FND concentration. FND fluorescence is unaffected by surface chemistry, so is used to quantify concentration. A serial dilution of FND suspensions from a known stock concentration was performed (dots showing means with error bars showing s.d., n = 6 measurement replicates). This was then fitted with a linear regression (lines) to find a relationship between fluorescence intensity and concentration. After each FND functionalization, the fluorescence intensities of the final suspensions were measured, and the linear fit was used to estimate concentration (crosses). d, Fundamental LODs for different sized FNDs on LFAs, using a model biotin–avidin interaction. Suspensions (55 μl) of BSA–biotin-functionalized FNDs were run at different concentrations on poly-streptavidin strips. Concentrations were chosen to span the dynamic range of the camera, limited by over-exposure, as seen with the top concentration of 200 and 600 nm FNDs. Dots show means and error bars show s.d. (n = 3 technical replicates, n = 3 measurement replicates). Each series is fitted to a simple linear regression, shown as the solid line, with 95% confidence intervals shown shaded. LODs for 120, 200 and 600 nm diameter FNDs are 200 aM, 46 aM, and 820 zM respectively, defined by the intersection of the lower 95% confidence intervals of the linear fit with the upper 95% confidence intervals of the blanks for each particle size.

Extended Data Fig. 6 Assay optimization by buffer selection.

Sensitivity is limited by the non-specific binding of FNDs at the LFA test line. LFA strip blocking, running buffer and washing step are, therefore, key factors in improving LOD. In this section 120 nm FNDs were used for optimization. a, Signal-to-background comparison for the FNDs in different running buffers. There is no wash step. Error bars show s.d. (n = 3 measurement replicates). Milk was selected as the basis for the running buffer. b, Subsequently, a sweep of different surfactants was performed (n = 1 measurement replicate). The best signal-to-background ratio came from adding 0.05 vol% Empigen, showing a significant increase in the signal-to-background. There is no wash step. c–e, The best running buffer was then used for a washing buffer pH sweep (n = 1 measurement replicate) (c). All washing buffers were run at a volume of 75 μl, chosen because preliminary experiments showed it to be a good compromise between assay time and washing success. Although results were similar, pH 5 gave the best signal-to-background ratio, so acetate buffer at 10 mM pH 5 was used as the basis for a second washing buffer sweep, shown in (d), testing a number of detergents and adding casein at 0.2 wt% as a blocking protein (n = 1 measurement replicate). As a final test, the three best running buffers were tested, each with the three best washing conditions, displayed as a grid in (e). Each square is the average of three measurements (n = 3 measurement replicates). The results were consistent with previous sweeps, the combination of the best running buffer and best washing buffer giving the best signal-to-background. Milk and protein percentages are by weight and detergent percentages are by volume.

Extended Data Fig. 7 Optimization of FND concentration.

The background was reduced by optimising the particle concentration, shown here for 120-nm FNDs. a, A positive LFA strip (500 pM of ssDNA) and a negative control (deionized water) were run at varying FND concentrations between 3.88 fM and 496 fM, plotted against FND concentration, and fitted to simple linear regressions. The dots show means and error bars show the s.d. of repeat measurements (n = 3 measurement replicates). Linear regressions are shown by solid lines, and shaded areas show the 95% confidence intervals of the fits. b, Signal-to-background ratio, found by dividing the fitted linear regressions in a, is plotted against FND concentration. At higher concentrations, where the gradient term of the linear regression dominates, the positive and negative lock-in values tend to a constant separation on the log–log plot, so the signal-to-background ratio tends to a constant value of around 27. At low concentrations, the positive and negative curves converge as the negative lock-in amplitude levels off at the noise threshold, and the signal-to-background ratio tends to 1. c, The fitted linear regressions in a were used, along with the antibody equilibrium dissociation constant measured in Extended Data Fig. 4, to estimate the variation of lock-in amplitude with analyte concentration at different FND concentrations. The principles and equations are described in full in Supplementary Information section 2. The LOD for each FND concentration is defined as the intersection of this plot with the value of the blank plus two times the 95% confidence interval at that value, assuming a low concentration positive would have a similar confidence interval. d, The estimated LODs and dynamic ranges from c, plotted against FND concentration, to determine the optimum.

Extended Data Fig. 8 Primer optimization.

a, List of forward primers (F1–F5) and reverse primers (R1–R5) tested for the initial primer screen. b, An initial primer screen was performed to achieve the highest amplification efficiency (n = 3 technical replicates) using the TwistAmp Exo Reverse Transcription Kit (TwistDx). The yield of each primer combination was measured by the fluorescence of the exo probe with a fluorescence microplate reader (SpectraMax i3, Molecular Devices LLC). Primers F5 and R3 gave the highest yield, although all the yields were above 63% of this value. c, Interactions between forward primers and reverse primers to predict the minimum free energy structures for the ten primer combinations that gave the largest yield of RPA product in the primer screen. The table shows the results of simulations in NUPACK56, using an input of 10 μM for each oligonucleotide. The minimum free energy secondary structures are the most energetically favourable secondary structures that can be assumed for oligonucleotides of a given primary sequence, calculated using the nearest-neighbour method57. Primers F1 and R4 were selected for future work since the energetics of their hybridization are much less favourable than that of F3 and R5, yet they still gave a high RPA yield in the primer screen (93% of the highest yield pair). d, A list of oligonucleotides used for PCR, RPA and qPCR assays. The PCR reverse primer included a T7 promoter for RNA transcription (underlined) and a spacer (bold). e, Gel electrophoresis of 1,503 bp template sequence produced by PCR using a 1% agarose gel. f, Gel electrophoresis of 181 bp double-stranded RT–RPA products using a 1% agarose gel.

Extended Data Fig. 9 Comparison of LODs of model ssDNA with real RPA amplicons and gold nanoparticles.

a, The dilution series of the real RPA amplicons and the model ssDNA ‘amplicons’ were plotted against concentration for 600 nm FNDs (dots showing means with error bars showing s.d., n = 3–9 technical replicates, n = 3 measurement replicates) with their respective linear fits (solid lines with 95% confidence intervals of the fit shown shaded). The curves are similar, with fitted KD values of 29 and 22 fM for model and real amplicons, respectively, and similar dynamic ranges. The real amplicons showed increased variation in the blanks, leading to a higher blank cutoff giving a higher LOD, and slightly reduced signal-to-blank ratio. b, The dilution series of model ssDNA ‘amplicons’ were plotted against concentration for 120, 200 and 600 nm FNDs (dots showing means with error bars showing s.d., n = 3 technical replicates, n = 3 measurement replicates) with their respective linear fits (solid lines with 95% confidence intervals of the fit shown shaded). The LODs are 3.7, 3.6 and 0.8 fM respectively. c, Comparison of 600 nm FNDs with 40 nm gold nanoparticles on LFAs, often used in LFAs owing to a good compromise between stability (and therefore ease of functionalization), and sensitivity58. Serial dilutions are plotted (dots showing means with error bars showing s.d., n = 3 technical replicates, n = 3 measurement replicates for the FNDs; and dots with error bars showing the s.d. across the test line, n = 1 technical replicate, n = 1 measurement replicate for the gold nanoparticles). LODs are calculated as previously, giving 800 aM and 6.0 pM, respectively. d, e, A Monte Carlo simulation of the signal variation that can be explained by the FND size distribution (from DLS measurements in Extended Data Fig. 3b) was performed (n = 200,000). The violin plots (d) show the normalized simulated random variation in lock-in amplitudes due to the 600-nm FND size distribution in the clinical sample assays in Fig. 4d (negative plasma control and clinical standard). The experimental data are overlaid, showing that FND size distribution explains approximately 8–9% of the total experimental signal variance. Full details of the simulation are given in Supplementary Information section 3. A further approximately 0.1–2% of the variance is explained by periodic drift in modulation amplitude, shown over 45 min in e, normalized to the mean. f, A plot of the variation in lock-in amplitude due to small changes in the modulation frequency, Fm. The variance of the frequency is 3 × 10−8% over the same period, giving negligible differences in lock-in amplitude.

Extended Data Fig. 10 Further analysis of RT-RPA samples.

a, ANOVA analysis was performed on the measured lock-in amplitudes of the FND LFAs, giving a P value of 7.4 × 10−29 and F value of 95.6, with 71 total degrees of freedom. Box plots of the data groups are shown (grouped by RNA concentration). The horizontal red lines represent the medians, the horizontal blue lines represent the 25th and 75th percentiles and the notches represent the 95% confidence intervals of the medians. The black dashed lines represent the range for each group. b, A graphical comparison of the means of the groups (grouped by RNA concentration). The circles represent the means, and the horizontal lines represent the comparison intervals of the means from Tukey HSD post hoc test (overlap of these intervals denotes statistical similarity). The negative control, highlighted in blue, is shown to be not significantly different from the 10−2 and 10−1 RNA copy number samples (P values >0.999, shown in grey), but it is significantly different from the 1, 101 and 102 RNA copy number samples (P values ≈10−8, shown in red). c, A table of ANOVA P values. The P value for the null hypothesis that the difference between the means of the two groups is zero. d, Comparison of amplification time for a low copy number RT–RPA sample (average of 1.26 RNA copies). Multiple RPA reactions were run and stopped after different times, before adding to FND LFAs, as described in Methods. A negative control is shown for comparison, and the dashed line represents the upper 95% confidence interval of the negative control. Dots show the mean of n = 3 measurement replicates, crosses show the individual measurements, and error bars represent the s.d. e, Early disease detection using FND LFAs was demonstrated by a seroconversion panel (ZeptoMetrix Corporation, Panel Donor No. 73698), taken from a single donor over a period of six weeks spanning the early stages of an HIV-1 infection. The thirteen samples of the panel were measured on FND LFAs (n = 1–2 experimental replicates, n = 3 measurement replicates). The measured values are plotted along with positive and negative non-amplification controls. They are colour-coded for RT–PCR results, and labelled with sample numbers, dates, and copy numbers in brackets. The blank cutoff is defined as the upper 95% interval of the negative control. The results show that the RNA was detectable on FND LFAs as early as RT–PCR, and six out of seven RT–PCR-positive samples were detected on FND LFAs, while six out of six RT–PCR-negative samples were negative.

Extended Data Fig. 11 Detection of HIV-1 capsid protein on using 600 nm FNDs.

A serial dilution of the capsid protein was detected on streptavidin-modified LFAs using a sandwich of a biotinylated capture nanobody and antibody-modified FNDs. The results are plotted (n = 3–4 experimental replicates, n = 3 measurement replicates), normalized to the blanks for each sample set with dots showing means and error bars showing s.d. These data were then fitted to a Langmuir curve (equation (6), shown as a line with shaded area denoting the 95% confidence interval of the fit). This gives a LOD of 120 fM, and a lowest concentration that is significantly different from the blank (at the 95% confidence level) of 3 pM, marked with *. Full experimental details are shown in Supplementary Information section 4.

Extended Data Fig. 12 Effect of lateral flow test strip drying on lock-in amplitude of FND assay.

a, Positive and negative lateral flow test strips were measured over time after running was complete (time = 0), showing a small increase in the positive strip lock-in amplitude as the strip dries (the initial lock-in amplitude is around 70% of the final value); however, no increase is seen in the negative control. The shaded areas show the standard deviation between repeats (n = 3 technical replicates). b, The resulting signal-to-blank ratio variation over time. The shaded areas show the standard deviation between repeats (n = 3 technical replicates), showing that the effect of drying is quite small compared to strip-to-strip variation.

Supplementary information

Supplementary Information

This file contains Supplementary Methods for Extended Data Figures: quantification of antibody binding constants, modelling of FND concentration, simulation of FND size distributions, antigen detection, qRT-PCR; and three Supplementary Tables summarizing estimated costs; comparing FNDs with other fluorescent reporters; and total assay time taken and outlook towards a point-of-care test.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Miller, B.S., Bezinge, L., Gliddon, H.D. et al. Spin-enhanced nanodiamond biosensing for ultrasensitive diagnostics. Nature 587, 588–593 (2020).

Download citation

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.


Quick links