ARTICLES NATURE PHOTONICS DOI: 10.1038/NPHOTON.2017.

55

a b 47 nm e

Intensity (norm.)
Diffraction dSTORM
limited ×60 NA 1.2 1.0 50 µm

0.5

0.0
0 50 100 150
2 µm Position (nm)

1 µm

f dSTORM
c Objective-based Waveguide chip- ×20 NA 0.45
TIRF dSTORM based dSTORM

5 µm Diffraction
50 nm 50 nm limited
Intensity (norm.)

Intensity (norm.)

d 49 nm g 138 nm
1.0 1.0

0.5 0.5

0.0 0.0 Diffraction dSTORM

−50 0 50 100 0 200 400 limited ×20 NA 0.45
Position (nm) Position (nm)

Figure 2 | Demonstration of chip-based dSTORM. a, Diffraction-limited and dSTORM imaging of immunostained tubulin in liver sinusoidal endothelial cells
(LSECs). b, Measuring a lateral profile of 540 nm width along a straight microtubule (magenta marking in the inset of the dSTORM image in a) reveals its
hollow structure. c, The resolution capability is further investigated by imaging DNA-origami nanorulers of (50 ± 5) nm specified length, which can be clearly
resolved with waveguide-based dSTORM, similar to objective-based TIRF dSTORM. d, Analysing their line profiles, a mean nanoruler length of 49 nm is
found in both cases, confirming that the chip-based implementation shows comparable performance to a conventional inverted dSTORM set-up.
e, Waveguide chip-based illumination also allows for using a low-magnification/low-NA (×20/NA 0.45) objective lens for dSTORM imaging over a
FOV of 0.5 mm × 0.5 mm. f, A detail from the white box in e. g, The profile over adjacent tubulin filaments reveals their separation by 138 nm.

this problem, a piezo stage is used to oscillate the coupling objective in contrast to the dSTORM acquisition procedure where single-
lens or fibre back and forth along the input facet of the waveguide molecule switching is required. While continuously changing the
during the measurement. This maintains continuous coupling but excitation pattern in this manner, about 200 frames are recorded.
shifts the mode pattern to obtain an average distribution that The acquired data are used as the input for the ESI reconstruction
shows significantly less modulation over larger length scales than algorithm to generate a super-resolved image (Fig. 3a), which we
in conventional TIRF set-ups (Supplementary Figs 8 and 9 and demonstrate by imaging tubulin in LSECs (Fig. 3b,c).
Supplementary Movie 2). Consequently, the distinct modes are Following the ESI acquisition, the same sample is again imaged
not visible in the reconstructed dSTORM images. using dSTORM by increasing the illumination intensity. This
allows for a direct comparison of the ESI image with the
Chip-based fluctuation imaging. In contrast, fluctuations induced dSTORM image of higher resolution but requiring a much higher
by the shifted mode pattern are desired in the case of chip-based number of input frames (Supplementary Table 1). Hence,
ESI. Although original implementations of the fluctuation-based dSTORM serves as a control reference for assessing the performance
approaches SOFI and ESI use intrinsic quantum dot or of chip-based ESI. This verifies that the resolution of the ESI image
fluorophore temporal intensity fluctuations (for example, due to is on the order of 110 nm, as adjacent microtubules at 106 nm
blinking and bleaching), it has recently been shown that speckle distance are still resolved and simultaneously observed in both
pattern illumination38 can also invoke temporal emission super-resolved images (Fig. 3c,d). This resolution is also confirmed
fluctuations, allowing for super-resolved fluctuation imaging39. by a FWHM of 104 nm of a single tubule in the ESI image
Intrinsic intensity fluctuations originate from single emitters and (Supplementary Fig. 12). Thus, chip-based ESI using spatial
therefore spatially tightly confined sources. In contrast, the spatial excitation pattern fluctuations readily achieves a resolution
frequencies of the waveguide illumination pattern define the enhancement of about a factor of 2, using an NA 1.2 objective
length scales on which the fluctuations occur and, hence, the lens for fluorescence detection (Supplementary Note 1).
obtainable resolution (Supplementary Fig. 10 and Supplementary Chip-based dSTORM pushes the resolution even further, admittedly
Note 1). Using dSTORM, we have measured fringes in the at the cost of longer acquisition times.
multimode interference pattern of a waveguide. The FWHM of
the structure sizes is as small as 140 nm for a vacuum laser Scalable super-resolution imaging. As the evanescent field generation
excitation wavelength of 660 nm (Supplementary Fig. 11). in waveguide chip-based nanoscopy does not depend on the objective
To use the illumination intensity fluctuations induced by the lens used for fluorescence detection, the presented approaches can be
waveguide for ESI analysis, we oscillate the coupling along the applied for successive image acquisitions at different magnifications,
input facet such that the mode distribution changes from image allowing for scalable FOV imaging. To obtain an overview image
to image. As random intensity fluctuations are desired, there is no with a large FOV, low-magnification/low-NA lenses can be used.
need to further control the illumination pattern besides changing If higher resolution is desired, specific regions of interest (ROIs)
it from frame to frame. Low input power is used, keeping the inten- can be imaged at superior resolution afterwards by switching to a
sity under the threshold of undesired single-molecule switching, high-magnification/high-NA objective lens.

324 NATURE PHOTONICS | VOL 11 | MAY 2017 | www.nature.com/naturephotonics

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.