Lettershttps://doi.org/10.1038/s41565-018-0123-1

1Nano-Optics and Biophotonics Group, Experimentelle Physik 5, Physikalisches Institut, Wilhelm-Conrad-Röntgen-Center for Complex Material Systems, Universität Würzburg, Würzburg, Germany. 2B CUBE – Center for Molecular Bioengineering, Technische Universität Dresden, Dresden, Germany. 3cfaed – Center for Advancing Electronics Dresden, Technische Universität Dresden, Dresden, Germany. 4Max Planck Institute of Molecular Cell Biology and Genetics, Dresden, Germany. 5Present address: Rudolf Virchow Center for Experimental Biomedicine, Universität Würzburg, Würzburg, Germany. 6Present address: Kurfürst-Moritz-Schule, Moritzburg, Germany. 7These authors contributed equally: Heiko Groß, Hannah S. Heil. *e-mail: hecht@physik.uni-wuerzburg.de; stefan.diez@tu-dresden.de

In the vicinity of metallic nanostructures, absorption and emission rates of optical emitters can be modulated by sev-eral orders of magnitude1,2. Control of such near-field light–matter interaction is essential for applications in biosensing3, light harvesting4 and quantum communication5,6 and requires precise mapping of optical near-field interactions, for which single-emitter probes are promising candidates7–11. However, currently available techniques are limited in terms of through-put, resolution and/or non-invasiveness. Here, we present an approach for the parallel mapping of optical near-field inter-actions with a resolution of < 5 nm using surface-bound motor proteins to transport microtubules carrying single emitters (quantum dots). The deterministic motion of the quantum dots allows for the interpolation of their tracked positions, resulting in an increased spatial resolution and a suppression of localization artefacts. We apply this method to map the near-field distribution of nanoslits engraved into gold layers and find an excellent agreement with finite-difference time-domain simulations. Our technique can be readily applied to a variety of surfaces for scalable, nanometre-resolved and artefact-free near-field mapping using conventional wide-field microscopes.

The challenge of probing the local optical near-field of a nano-structure is converting its evanescent near-field into modes that can propagate to the far-field. In this regard, a single emitter constitutes a superior probe. It responds to the near-field modes of a nano-structure via both the local intensity and the zero-point fluctuations described by the local density of optical states. The position-depen-dent coupling of the emitter to the nanostructure thereby leads to changes in the excitation rate as well as in the excited-state lifetime and the angular emission pattern. Moreover, its small oscillator strength ensures a negligible impact of the emitter on the resonant field distribution of the nanostructure.

Previously, diffusing or adsorbed single molecules in aqueous solution have been used as probes for spatially resolved local field mapping including fluorescence enhancement or surface-enhanced Raman scattering10,12–15. However, recent studies16,17 clarified that such an endeavour is limited by the uncertainty in the exact loca-tion of the emitter, which cannot be estimated by single-particle tracking methods due to coupling with the nanostructure (here-inafter referred to as localization artefacts). More deterministic results can be obtained by scanning a single molecule18 or single

nitrogen-vacancy centre19 attached to a scanning tip. Such measure-ments, however, are inherently slow and suffer from the influence of the scanning tip on the local near-field. An elegant approach to avoid such perturbations consists in controlling the emitter position by microfluidic flow11, but this approach is limited in terms of scal-ability and accuracy of the position control.

In our approach, we use kinesin-1 motor proteins and microtu-bules—the biomolecular machinery responsible for cargo transport inside living cells—to scan quantum dots (QDs) over nanostructured gold surfaces. This approach offers a multi-probe method for paral-lel mapping of optical near-fields using individual dipole emitters in a virtually artefact-free fashion. In our experiment the sample of interest is decorated with kinesin-1 motor proteins (Fig. 1a). When fuelled with adenosine triphosphate (ATP), the surface-bound motors transport the QD-labelled microtubules across the substrate at controlled uniform velocity, an assay well-established in the field of motor protein biophysics20 and bio-nanotechnology21. Such glid-ing assays are readily prepared following a simple, straightforward protocol (see Methods). The samples are illuminated from below by collimated laser light and imaged from above by a fluorescence microscope equipped with an electron-multiplying charge-coupled device (EMCCD) camera (Fig. 1). This allows parallel recording of the fluorescence emission of up to ~100 individual QDs in a field of view of ~100 × 100 µ m2 (see Fig. 1c and Methods). Importantly, the impact of our ‘scanning apparatus’ on the near-field interaction is negligible due to the small scattering cross-section of the nm-sized proteinaceous transport components (globular size of kinesin-1 motor-domain, ~5 nm; diameter of microtubules, ~25 nm; refrac-tive index, ~1.38). Motor-propelled QDs are thus well-suited scan-ning probes to investigate near-field interactions.

The potential of our approach is demonstrated by mapping the near-field interactions of individual QDs with nanoslits of 110 to 240 nm width engraved in a 25-nm-thick gold layer (Fig. 1b). Compared to low fluorescence intensities when the QDs were transported over unstructured regions (due to decreased illumina-tion intensities ‘behind’ the gold), we observed higher, position-dependent fluorescence intensities when the QDs crossed a nanoslit (Figs. 1c and 2 and Supplementary Video 1). In contrast, when imaged on bare glass, the fluorescence intensities were independent of the QD positions (Supplementary Fig. 1 and Supplementary Video 2).

To quantitatively analyse the near-field interactions between the QD and the nanoslit, we recorded hundreds of QD trajectories by

Parallel mapping of optical near-field interactions by molecular motor-driven quantum dotsHeiko Groß   1,7, Hannah S. Heil   2,5,7, Jens Ehrig2, Friedrich W. Schwarz2,3,6, Bert Hecht   1* and Stefan Diez2,3,4*

NaturE NaNotEcHNoloGy | VOL 13 | AUGUST 2018 | 691–695 | www.nature.com/naturenanotechnology 691

Letters Nature NaNotechNology

tracking individual QDs using the image analysis software FIESTA22 (Fig. 2a). The precision of the QD position in each frame obtained by this conventional localization analysis varied due to QD blinking and different inherent QD brightnesses and was typically around 30 nm. However, we were able to substantially increase the localiza-tion precision by exploiting the deterministic path and the uniform velocity of the microtubule-attached QDs. For each trajectory, the tracked positions around the nanoslit regions were fitted with a third-degree polynomial function. As exemplified in Fig. 2a, this is a very good approximation for the paths that gliding microtubules take across the surface due to their high persistence length. Owing to the uniform gliding velocity of the microtubules, the QD posi-tions can then be reassigned as equidistantly distributed along the path by minimizing the sum of squared distances to the originally tracked positions. Following this procedure we were able to improve the QD localization precision by one order of magnitude (to < 5 nm, Fig. 2b–d). Slow gliding velocities (~300 nm s–1 at 40 µ M ATP) in our assays and a fast acquisition rate (20 frames per second) ensure that the impact of fluctuations due to changes in the attachment geometry of the motors23 remains below the localization precision. Importantly, the described interpolation strategy efficiently sup-pressed localization artefacts that we indeed found to be present in our conventional tracking results near the edges of the nanoslits (Supplementary Fig. 2).

For each trajectory, the experimentally detected fluorescence intensity was then registered as a function of the distance of the QD to the nanoslit centreline and normalized to account for varia-tions in the intrinsic QD brightness (Supplementary Methods, Supplementary Figs. 3–5 and Supplementary Section 1). By aver-aging the normalized QD intensities for each nanoslit width, we minimized the intensity variations caused by the characteristic QD blinking. Ultimately, we obtained nanometre-resolved, averaged fluorescence intensity profiles for all nanoslits under investigation (Fig. 2e–f). This resolution substantially surpasses the ones previ-ously reported in experiments on similar structures10,11,15,18,19,24.

The measured fluorescence intensity profiles are supported by numerical finite-difference time-domain simulations (Fig. 3 and Supplementary Section 2). Because the wavelengths for QD excitation

(488 nm) and emission (655 nm) are well separated, the detected fluorescence intensity can be described as resulting from two dis-tinct effects: (i) the excitation of the QD in the inhomogeneous near-field distribution of the illuminated nanostructure and (ii) the resonant near-field energy transfer between the excited QD and the same nanostructure, which typically leads to changes of the excited-state lifetime (via radiative and non-radiative decay channels) and the angular emission pattern.

The excitation intensity distribution I x y( , )ex of the near-field above the gold layer was simulated for the illumination wavelength of 488 nm (Fig. 3a, upper panels). For all slit widths, Iex (which determines the QD excitation rates) is highest directly above the nanoslits and small elsewhere due to absorption and reflection by the gold layer.

To quantify the impact of resonant near-field energy transfer between the QD and the nanoslit sample we introduce the emis-sion efficiency η x y( , ) which is the normalized power emitted by a dipole (directly and via surface plasmons) that is captured by the microscope objective. We simulated η by placing an electric dipole emitting at 655 nm at distinct positions above the sample sur-face and integrating the radiated power over the collection angle (Supplementary Fig. 6). Since Iex remained below saturation condi-tions in our experiments, the power captured for each dipole posi-tion was normalized to the power captured from positions away from the slit regions (Fig. 3, lower panels).

Our simulations include effects of quenching as well as excita-tion of radiative surface plasmon polaritons. The random QD orientation is accounted for by averaging over all three dipole ori-entations (Supplementary Section 2). We found the emission effi-ciency η to be strongly position dependent both as function of the horizontal distance x to the slit centreline and the vertical distance y to the gold surface (Fig. 3a, lower panels).

For excitation rates well below saturation, the average QD fluo-rescence intensity Φ detected in the far-field, is given by

Φ η= × ⟨ ×x c I x y x y( ) ( , ) ( , ) (1)yex

5 µm

240 nm 220 nm 200 nm 175 nm 160 nm 140 nm 125 nm 110 nm

Kinesin

MicrotubuleQD

Gold

Camera

488 nmlaser

Det

ectio

nIll

umin

atio

n

Polarizer

Filt

er

Objective

cb

a

Fig. 1 | using biomolecular motors and microtubules to scan individual QDs across gold nanoslits. a, Schematic diagram of the set-up: QDs are randomly attached to microtubules, which are propelled across the sample by surface-bound kinesin-1 motors. The sample is illuminated by collimated laser light (488 nm, polarization along the long axis of the nanoslit) from below and the fluorescence signal is collected from above by a fluorescence microscope equipped with an EMCCD camera. Changes in the near-field interaction of the QDs with the gold nanoslits are detected in the far-field as spatially varying fluorescence intensities. b, Scanning electron microscopy (SEM) image of the investigated nanoslits with indicated widths in a 25-nm-thick gold layer. c, Maximum-intensity projection (pixel width 129 nm) of QDs transported across the nanostructured sample depicted in b (see also Supplementary Videos 1 and 3). The slits are situated in the centre of the regions indicated by the grey dashed lines.

NaturE NaNotEcHNoloGy | VOL 13 | AUGUST 2018 | 691–695 | www.nature.com/naturenanotechnology692

LettersNature NaNotechNology

where c is a scaling factor that accounts for the quantum efficiency of the optical set-up. Due to a random distribution of QD bind-ing sites on the microtubule surface and the fact that microtubules rotate around their longitudinal axis during transport25, the QDs in our experiments were scanned across the surface at vertical dis-tances between 15 and 55 nm (see Supplementary Section 2 and ref. 26). This is accounted for in our simulations by averaging the product of Iex and η over the specified height range. When com-paring Φ to the mean normalized fluorescence intensities detected in the experiments, we find excellent agreement for all slit widths (Fig. 3b).

Our results lay the foundation for future studies of more com-plex nanostructures: whereas the electromagnetic field behind an illuminated nanostructure can often be calculated, straightforward quantification of the position-dependent near-field energy transfer requires advanced methods and easily becomes intractable for com-plex geometries. Our method provides a solution to this problem as it is capable of deducing emission efficiency maps by deconvolving the measured fluorescence intensity with the calculated electromag-netic field. We note that the emission efficiency accommodates all relevant effects of near-field energy transfer, making it a valuable experimental parameter when studying light–matter interactions, in particular for plasmonic nanostructures. For conditions well below saturation (as in our experiments), the emission efficiency describes changes in quenching and changes in the angular emis-sion pattern. In saturation, changes of the emitter’s excited-state life-time contribute to additional variations in the emission efficiency. In the future, it will be interesting to explore possibilities to decom-pose the emission efficiency into its constituting factors and what type of measurements are required to do so.

With the rapid evolution of plasmonic nanostructures self- assembled in aqueous environment, for example, by arranging metal-lic nanoparticles on DNA origami27,28, there is an increased need for compatible near-field scanning methods. Whereas traditional

Position along short slit axis (nm)

Nor

mal

ized

QD

inte

nsity

(a.

u.)

–750 –500 –250 0 250 500 750

240 nm

0

2

4

220 nm

0

2

4

200 nm

0

2

4

175 nm

0

2

4

140 nm

0

2

4

125 nm

0

2

4

Nor

mal

ized

QD

inte

nsity

(a.

u.)

Position along short slit axis (nm)

0

2

4

–600 –400 –200 0 200 400 600

160 nm

Positions alongfitted path

0

2

4

Positions fromtracking

0

2

4

500 nm

0

4

23

1

Φ

100 nm

110 nm

0

2

4f

e

d

c

ba

Fig. 2 | Quantitative mapping of QD–nanoslit near-field interactions. a, Maximum-intensity projection (greyscale image, pixel width 129 nm) of a time-lapse video showing a microtubule-bound QD being transported across a nanoslit (160 nm width) along with the QD positions in the individual frames as obtained from conventional particle tracking (dots, ‘tracked positions’ colour-coded with the normalized fluorescence intensity Φ). The tracked positions were fitted with a third-degree polynomial (red line) to approximate the path of the QD. b, Magnified segment of the QD path. Each tracked position is represented by a blue dot along with a semi-transparent circle with a radius of twice the mean fitting error of the tracking (that is, representing the area in which the QD was located with 95% probability). To improve the localization precision, the QD positions were reassigned as uniformly distributed points along the fitted path (red dots, connected to the tracked positions by blue lines), which effectively decreased the uncertainty in determining the QD position by one order of magnitude as evident from the comparison of the 95% prediction (red lines) and confidence interval (blue lines) of the polynomial fit. c, Normalized QD intensity plotted against the tracked position perpendicular to the nanoslit (blue dots in b). d, Same as c but with reassigned positions (red dots in b). e, Averaged normalized QD intensity from all measured trajectories using the reassigned positions (dots, 10 nm bins) along with all individual normalized intensity profiles (lines) of the 160-nm-wide slit. f, Same as e for all remaining slit widths. In plots a,c–f the nanoslit edges are indicated by vertical dashed lines.

0

2

4

0.5

1.0

Position (nm)

0

2

4

0.5

1.0

0

2

4

0.5

1.0

–750 –500 –250 0 250 500 750

–750 –500 –250 0 250 500 750

–750 –500 –250 0 250 500 750

Position (nm)

Hei

ght (

nm)

100

50

0

103050

–750 –500 –250 0 250 500 750

110 nm

Hei

ght (

nm)

100

50

0

103050

–750 –500 –250 0 250 500 750

160 nm

Hei

ght (

nm)

100

50

0

103050

–750 –500 –250 0 250 500 750

240 nm

Iex(a.u.)

Iex

Iex

0

η (a.u.)

η ⟨η⟩

η

Iex

η

2 4 6 8

1.41.10.80.5

ba

Ф⟨η

⟩Ф

⟨η⟩

Ф

Fig. 3 | Numerical simulation of the QD–nanoslit near-field interactions. a, Upper panels show the near-field intensity distributions Iex above the gold nanoslits on plane wave excitation with 488 nm from below. Lower panels show the far-field emission efficiencies η of a dipole emitter with an emission wavelength of 655 nm as a function of its position and height above the gold surface. Slit widths as indicated in the upper panels. b, Normalized excitation intensity Iex (blue dashed lines) and emission efficiency η (red lines) averaged over the experimentally relevant height range (15–55 nm above the surface, see text and Supplementary Section 2) as indicated by the horizontal dashed lines in the corresponding panels in a. Only the product of these two quantities η×Iex , namely the normalized QD fluorescence intensity Φ (orange lines), agrees well with the normalized experimentally measured QD fluorescence intensity (black dots).

NaturE NaNotEcHNoloGy | VOL 13 | AUGUST 2018 | 691–695 | www.nature.com/naturenanotechnology 693

Letters Nature NaNotechNology

methods fall short in that respect because the sensitivity of cantile-ver- or tuning-fork-based position control is decreased in water, our approach is well adapted to these requirements.

Scanning a single emitter by a biomolecular transport machin-ery directly integrated on the sample not only leads to an improved resolution of near-field mapping itself but also provides a number of crucial technological advantages: (i) the negligible influence of the proteinaceous transport machinery on the local refractive index minimizes any distortions of the QD–nanostructure near-field interactions; (ii) the inherent mechanical stability makes the system invulnerable to external influences (drift and vibration) and thus, near-field mapping can be carried out using a standard fluorescence microscope; (iii) the fact that we obtain a full intensity profile from each individual QD makes it possible to account for inhomogene-ities in sample illumination and inherent QD brightnesses by nor-malizing each intensity profile individually; (iv) the deterministic motion of the QDs allows for the interpolation of their tracked posi-tions in the areas of interest, resulting in a largely increased spatial resolution and a suppression of localization artefacts inherent to previous approaches using single-molecule localization microscopy near photonic nanostructures. Taken together, our method thus combines the precision of scanning probe approaches with the scal-ability of multi-emitter techniques.

With regard to the resolution demonstrated in our current work, we believe that it can be further improved by increasing the number of recorded data points per probe trajectory as well as by employ-ing non-blinking probes with higher brightness. At the same time, a reduction in the total measurement time is conceivable by imple-menting topographical and chemical guiding strategies29 to direct the gliding microtubules to the regions of interest.

Further improvements provided by our method include the additional tracking of the emitter’s distance to the sample surface30. Together with the fact that microtubules rotate around their longi-tudinal axis during gliding25, and by that dynamically change the emitter–sample distance, our approach thus holds the potential to evolve into a 3D measurement technique. Alternatively, embed-ding the emitters inside the lumen of the microtubules would allow adjustment of the constant emitter-to-sample distances when using differently sized linker molecules or even smart polymer systems with in situ adjustable conformational states. In both cases, our improved localization strategy based on positional interpolation is applicable due to the deterministic motility behaviour of the glid-ing microtubules. Notwithstanding, the inherent properties of our approach to employ multiple QD probes in parallel and to perform measurements without the necessity of any electronic feedback control promises applications of near-field scanning on large-scale devices, such as sensor chips and other metasurfaces, where current technologies fail.

MethodsMethods, including statements of data availability and any asso-ciated accession codes and references, are available at https://doi.org/10.1038/s41565-018-0123-1.

Received: 18 October 2016; Accepted: 22 March 2018; Published online: 30 April 2018

references 1. Kinkhabwala, A. et al. Large single-molecule fluorescence enhancements

produced by a bowtie nanoantenna. Nat. Photon. 3, 654–657 (2009). 2. Hoang, T. B. et al. Ultrafast spontaneous emission source using plasmonic

nanoantennas. Nat. Commun. 6, 7788 (2015). 3. Fabrizio, E. D. et al. Roadmap on biosensing and photonics with advanced

nano-optical methods. J. Opt. 18, 063003 (2016). 4. Knight, M. W., Sobhani, H., Nordlander, P. & Halas, N. J. Photodetection

with active optical antennas. Science 332, 702–704 (2011). 5. Beugnon, J. et al. Quantum interference between two single photons emitted

by independently trapped atoms. Nature 440, 779–782 (2006).

6. Reiserer, A., Kalb, N., Rempe, G. & Ritter, S. A quantum gate between a flying optical photon and a single trapped atom. Nature 508, 237–240 (2014).

7. Farahani, J. N., Pohl, D. W., Eisler, H.-J. & Hecht, B. Single quantum dot coupled to a scanning optical antenna: a tunable superemitter. Phys. Rev. Lett. 95, 017402 (2005).

8. Anger, P., Bharadwaj, P. & Novotny, L. Enhancement and quenching of single-molecule fluorescence. Phys. Rev. Lett. 96, 113002 (2006).

9. Kühn, S., Håkanson, U., Rogobete, L. & Sandoghdar, V. Enhancement of single-molecule fluorescence using a gold nanoparticle as an optical nanoantenna. Phys. Rev. Lett. 97, 017402 (2006).

10. Cang, H. et al. Probing the electromagnetic field of a 15-nanometre hotspot by single molecule imaging. Nature 469, 385–388 (2011).

11. Ropp, C. et al. Nanoscale imaging and spontaneous emission control with a single nano-positioned quantum dot. Nat. Commun. 4, 1447 (2013).

12. Wu, D., Liu, Z., Sun, C. & Zhang, X. Super-resolution imaging by random adsorbed molecule probes. Nano Lett. 8, 1159–1162 (2008).

13. Stranahan, S. M. & Willets, K. A. Super-resolution optical imaging of single-molecule SERS hot spots. Nano Lett. 10, 3777–3784 (2010).

14. Willets, K. A. Super-resolution imaging of SERS hot spots. Chem. Soc. Rev. 43, 3854–3864 (2014).

15. Wertz, E., Isaacoff, B. P., Flynn, J. D. & Biteen, J. S. Single-molecule super-resolution microscopy reveals how light couples to a plasmonic nanoantenna on the nanometer scale. Nano Lett. 15, 2662–2670 (2015).

16. Raab, M., Vietz, C., Stefani, F. D., Acuna, G. P. & Tinnefeld, P. Shifting molecular localization by plasmonic coupling in a single-molecule mirage. Nat. Commun. 8, 13966 (2017).

17. Fu, B., Isaacoff, B. P. & Biteen, J. S. Super-resolving the actual position of single fluorescent molecules coupled to a plasmonic nanoantenna. ACS Nano 11, 8978–8987 (2017).

18. Michaelis, J., Hettich, C., Mlynek, J. & Sandoghdar, V. Optical microscopy using a single-molecule light source. Nature 405, 325–328 (2000).

19. Kühn, S., Hettich, C., Schmitt, C., Poizat, J.-P. & Sandoghdar, V. Diamond colour centres as a nanoscopic light source for scanning near-field optical microscopy. J. Microsc. 202, 2–6 (2001).

20. Nitzsche, B. et al. in Methods in Cell Biology Vol. 95 (eds. Wilson, L. & Correia, J. J.) Ch. 14, 247–271 (Academic Press, Cambridge, 2010).

21. Hess, H., Clemmens, J., Howard, J. & Vogel, V. Surface imaging by self-propelled nanoscale probes. Nano Lett. 2, 113–116 (2002).

22. Ruhnow, F., Zwicker, D. & Diez, S. Tracking single particles and elongated filaments with nanometer precision. Biophys. J. 100, 2820–2828 (2011).

23. Palacci, H. et al. Velocity fluctuations in kinesin-1 gliding motility assays originate in motor attachment geometry variations. Langmuir 32, 7943–7950 (2016).

24. Anikeyev, V., Temnov, V. V., Woggon, U., Devaux, E. & Ebbesen, T. W. Propagation oscillations in the near-field response of traveling surface waves launched by metallic nanoapertures. Appl. Phys. B 93, 171–176 (2008).

25. Nitzsche, B., Ruhnow, F. & Diez, S. Quantum-dot-assisted characterization of microtubule rotations during cargo transport. Nat. Nanotech. 3, 552–556 (2008).

26. Kerssemakers, J., Howard, J., Hess, H. & Diez, S. The distance that kinesin-1 holds its cargo from the microtubule surface measured by fluorescence interference contrast microscopy. Proc. Natl Acad. Sci. USA 103, 15812–15817 (2006).

27. Acuna, G. P. et al. Fluorescence enhancement at docking sites of DNA-directed self-assembled nanoantennas. Science 338, 506–510 (2012).

28. Gür, F. N., Schwarz, F. W., Ye, J., Diez, S. & Schmidt, T. L. Toward self-assembled plasmonic devices: high-yield arrangement of gold nanoparticles on DNA origami templates. ACS Nano 10, 5374–5382 (2016).

29. van den Heuvel, M. G. L., Butcher, C. T., Smeets, R. M. M., Diez, S. & Dekker, C. High rectifying efficiencies of microtubule motility on kinesin-coated gold nanostructures. Nano Lett. 5, 1117–1122 (2005).

30. Sun, Y., McKenna, J. D., Murray, J. M., Ostap, E. M. & Goldman, Y. E. Parallax: high accuracy three-dimensional single molecule tracking using split images. Nano Lett. 9, 2676–2682 (2009).

acknowledgementsWe thank M. Braun, R. Heintzman, A. Mitra, C. Reuther and T. Korten for fruitful discussions as well as C. Bräuer and T. Korten for supplying the kinesin-1 enzyme and technical support. This work was financially supported by the German Research Foundation (DFG) through the Center for Advancing Electronics Dresden (cfaed), the Heisenberg programme (DI 1226/4-1 to S.D.) and the European Social Funds (ESF) (contract 100111059, MindNano). H.G. and B.H. acknowledge financial support from the DFG via grant He5618/1-1 and a Reinhart Koselleck project.

NaturE NaNotEcHNoloGy | VOL 13 | AUGUST 2018 | 691–695 | www.nature.com/naturenanotechnology694

LettersNature NaNotechNology

author contributionsH.S.H., F.W.S., B.H. and S.D. conceived and designed the experiments. H.S.H. and F.W.S. performed the experiments. H.S.H., J.E. and F.W.S. analysed the data. H.G. performed the numerical simulations. All authors contributed to writing the paper.

competing interestsThe authors declare no competing interests.

additional informationSupplementary information is available for this paper at https://doi.org/10.1038/s41565-018-0123-1.

Reprints and permissions information is available at www.nature.com/reprints.

Correspondence and requests for materials should be addressed to B.H. or S.D.

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

NaturE NaNotEcHNoloGy | VOL 13 | AUGUST 2018 | 691–695 | www.nature.com/naturenanotechnology 695

Letters Nature NaNotechNology

MethodsChemicals. Casein, d-glucose, ATP, catalase, paclitaxel, NH4OH and H2O2 were purchased from Sigma-Aldrich. Glucose oxidase was purchased from SERVA. GMPCPP was purchased from Jena Bioscience. Ultrapure water was produced in a Milli-Q water purification system from Merck Millipore. Ethanol absolute was purchased from VWR. mPEG-SH was purchased from Nanocs. Hellmanex III solution was purchased from Omnilab. Mucasol was purchased from Merz Hygiene GmbH. Anti-His antibody was purchased from QIAGEN GmbH. Tubulin was purified in-house31. Biotinylated tubulin was purchased from Cytoskeleton. QD655 streptavidin conjugate, DTT, and Protein A/G were purchased from Thermo Fisher Scientific.

QD-labelled microtubules. Microtubules were polymerized by incubating 20 µ g of a 1/100 mixture of biotinylated tubulin and tubulin in 160 µ l polymerization buffer (BRB80 with 1 mM GMPCPP) at a temperature of 37 °C for 2 h. After spinning down the microtubules in a tabletop microcentrifuge (Heraeus Fresco 17, Thermo Fisher Scientific) at 17,000g and 23 °C for 20 min, the remaining pellet was resuspended in 85 μ l BRB80 with 1 μ M paclitaxel. Subsequently, the microtubules were labelled ‘in tube’ with QDs by incubating them for at least 10 min in BRB80 with 1 μ M paclitaxel, 1 mM ATP, 0.1 mg ml–1 casein, 10 mM DTT, 20 mM d-glucose, 55 µ g ml–1 glucose oxidase, 11 µ g ml–1 catalase and 1 nM QD655 solution.

Motor proteins. The purification of full length Drosophila melanogaster kinesin heavy chain (Dm KHC) kinesin-1 expressed in insect cells is described in full detail in ref. 32. The kinesin motor solution used for the gliding assays consisted of 12 μ g ml–1 full length Dm KHC 6xHis, 10 mM DTT, 1 μ M ATP, and 0.2 mg ml–1 casein in BRB80.

Sample fabrication. The nanoslit sample was fabricated on a 130-µ m-thick glass coverslip (Gerhard Menzel GmbH). A 3 nm chromium adhesion layer was deposited by electron beam evaporation followed by a 25-nm-thick gold layer. Nanoslits were etched into the sample by focused ion beam milling (Helios Nanolab 600 DualBeam) with an acceleration voltage of 30 kV and an ion current of 9.7 pA. SEM images of the nano-slit samples were obtained using a scanning electron microscope (Helios Nanolab 600 DualBeam) at 10 keV.

Sample preparation. The sample was cleaned in 2% Hellmanex III solution by incubation at 23 °C for 5 min. After rinsing the sample for 1 min with ultrapure water an RCA cleaning33 was performed. For this the sample was incubated at 60 °C for 5 min in an RCA solution (50 ml ultrapure water and 10 ml of 28% NH4OH heated to 60 °C, then 10 ml of 30% H2O2 were added). The clean surface was then coated with a self-assembling PEG layer. For this the sample was rinsed for 1 min with ultrapure water, dried with N2 and then incubated for 12 h in a passivation solution (1 mM 350 Da mPEG-SH in ethanol absolute). Finally, the sample was washed in ethanol absolute to remove the unbound PEG and dried with N2.

QD scanning assay on nanostructured gold samples. Flow cells were constructed by thermally bonding the sample to an ‘easy clean’ coverslip (22 × 50 mm #1; Corning) via spacer strips of parafilm at 65 °C for 1 min (with the gold-coated sample surface inside the resulting flow cell). The ‘easy clean’ procedure includes the following steps: firstly, the coverslips were sonicated for 15 min in Mucasol diluted in a 1:20 ratio with deionized water. After rinsing them for 2 min with deionized water they were sonicated for 10 min in ethanol absolute. In the next step the coverslips were again rinsed for 2 min with deionized water and then another 2 min with ultrapure water. Finally, they were dried with N2. The flow cells were filled with 0.1 mg ml–1 Protein A/G in BRB80. After 1 min of incubation, the solution was exchanged with 0.5 mg ml–1 casein and in a second step with 2 µ m ml–1 anti-His antibody in BRB80. After another 1 min of incubation and flushing out the anti-His antibody with at least two times the channel volume of 0.5 mg ml–1 casein in BRB80, 20 µ l kinesin motor solution was flushed into the channel. After another 1 min of incubation the solution was exchanged with 20 µ l motility solution containing 1 mM ATP (0.22 mg ml–1 casein, 22 mM d-glucose, 10 mM DTT, 1 mM ATP, 0.011 mg ml–1 catalase, 0.11 mg ml–1 glucose oxidase in BRB80T (1 μ M paclitaxel in BRB80)). Finally, 20 μ l of QD-labelled microtubules were added and incubated for 5 min, followed by a flush with 20 μ l motility solution containing 40 μ M ATP. During the measurement the last two steps were repeated every half an

hour to replace microtubules that had detached and to ensure the supply with fresh oxygen scavengers and buffer solution.

QD scanning assay on bare glass. The flow cell was constructed by thermally bonding two easy clean coverslips (18 × 18 mm and 22 × 22 mm, #1.5, Gerhard Menzel GmbH) via three parafilm spacers. At both ends of each channel, reservoirs were constructed with a two-component silicon glue (picodent twinsil 22, picodent). The flow cell was flushed with 0.5 mg ml–1 casein in BRB80. After 5 min the solution in the cell was replaced by a kinesin motor solution (20 µ g ml–1), 0.2 mg ml–1 casein, 1 mM ATP and 10 mM DTT in BRB80), which was washed out with a motility solution containing 1 mM ATP (0.22 mg ml–1 casein, 22 mM d-glucose, 10 mM DTT, 1 mM ATP, 0.011 mg ml–1 catalase, 0.11 mg ml–1 glucose oxidase in BRB80T (1 µ M paclitaxel in BRB80)) after 5 min incubation time. Finally, 20 µ l of QD-labelled microtubules were added and incubated for 5 min, followed by a flush with 20 µ l motility solution containing 1 mM ATP.

Optical imaging. The optical set-up is shown schematically in Fig. 1. QDs were imaged using an inverted light microscope (Axio Observer Z3, Carl Zeiss AG) equipped with an EMCCD camera (iXon ultra DU-888U3, Andor), a C-APOCHROMAT 1.2 NA, × 63 water objective (Carl Zeiss AG) and an optovar with 1.6-fold magnification (resulting in a pixel width of 129 nm). The objective was held at a constant temperature of 20 °C using a self-build water-cooled ring together with a water bath (FC25; Julabo GmbH). Illumination of the sample was performed with close to parallel illumination (NA < 0.05) in transmission mode using a homebuilt collimator (micro bench system components, LINOS) and a polarization-maintaining glass fibre coupled continuous-wave diode laser (Vortran Laser Technology) operated at 488 nm and 15 mW emission power (5 mW for reference measurements on bare glass (transmittance of the gold layer ~30%)).

The excitation light was cleaned up with a 488/10 nm band pass filter (ZET 488/10; Chroma Technology GmbH) and the polarization was adjusted parallel to the long axis of the nano-slit structures. The QD655 emission light was filtered using a 655/40 nm band pass filter (655/40 Bright Line HC; Semrock). Image streams were taken with 50 ms exposure time (100 ms for reference measurements on bare glass) in frame transfer mode, with the preamp setting 1 (which for our camera corresponds to 15.9 electrons per A/D count and 23.8 electrons of RMS readout noise), an EM-gain of 200 and 16 bit A/D-Conversion at a pixel rate of 1 MHz. From these values we calculated that on average about 335 ± 93 photons were collected from a single QD in each frame.

Data drift correction. The drift was determined using stuck QDs as fiducial markers, which were identified as follows. First, all fluorescent objects were localized in all frames using the ImageJ plugin ThunderSTORM34. The resulting position data were density-filtered to keep only positions occurring in a local radius of 60 nm for at least 15 consecutive frames. Subsequently, the fiducial markers were identified from those residual data by the following criteria: maximum travelling distance of 40 nm between frames and a minimum marker visibility of 15% in all frames. The mean displacement of all fiducial markers in reference to the first frame was used as drift correction.

Data availability. The data that support the findings of this study are available from the corresponding authors upon reasonable request.

references 31. Castoldi, M. & Popov, A. V. Purification of brain tubulin through two cycles

of polymerization–depolymerization in a high-molarity buffer. Protein Expr. Purif. 32, 83–88 (2003).

32. Korten, T., Chaudhuri, S., Tavkin, E., Braun, M. & Diez, S. Kinesin-1 expressed in insect cells improves microtubule in vitro gliding performance, long-term stability and guiding efficiency in nanostructures. IEEE Trans. NanoBioscience 15, 62–69 (2016).

33. Kern, W. The evolution of silicon wafer cleaning technology. J. Electrochem. Soc. 137, 1887–1892 (1990).

34. Ovesný, M., Křížek, P., Borkovec, J., Švindrych, Z. & Hagen, G. M. ThunderSTORM: a comprehensive ImageJ plug-in for PALM and STORM data analysis and super-resolution imaging. Bioinformatics 30, 2389–2390 (2014).

NaturE NaNotEcHNoloGy | www.nature.com/naturenanotechnology