We designed an epi-illumination SPIM system that uses a single objective and has a sample interface identical to that of an inverted fluorescence microscope with no additional reflection elements. It achieves subcellular resolution and single-molecule sensitivity, and is compatible with common biological sample holders, including multi-well plates. We demonstrated multicolor fast volumetric imaging, single-molecule localization microscopy, parallel imaging of 16 cell lines and parallel recording of cellular responses to perturbations.
Access optionsAccess options
Subscribe to Journal
Get full journal access for 1 year
only $20.17 per issue
All prices are NET prices.
VAT will be added later in the checkout.
Rent or Buy article
Get time limited or full article access on ReadCube.
All prices are NET prices.
The data that support the findings of this study are available from the corresponding author upon reasonable request.
Custom-written code is available from the corresponding author upon reasonable request.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Power, R. M. & Huisken, J. Nat. Methods 14, 360–373 (2017).
Huisken, J., Swoger, J., Bene, F. D., Wittbrodt, J. & Stelzer, E. H. K. Science 305, 1007–1009 (2004).
Tomer, R., Khairy, K., Amat, F. & Keller, P. J. Nat. Methods 9, 755 (2012).
Wu, Y. et al. Proc. Natl Acad. Sci. USA 108, 17708–17713 (2011).
Holekamp, T. F., Turaga, D. & Holy, T. E. Neuron 57, 661–672 (2008).
Chen, B.-C. et al. Science 346, 1257998 (2014).
Dunsby, C. Opt. Express 16, 20306–20316 (2008).
Li, T. et al. Sci. Rep. 4, 7253 (2014).
McGorty, R. et al. Opt. Express 23, 16142 (2015).
Bouchard, M. B. et al. Nat. Photonics 9, 113–119 (2015).
Strnad, P. et al. Nat. Methods 13, 139–142 (2016).
Maioli, V. et al. Sci. Rep. 6, 37777 (2016).
Galland, R. et al. Nat. Methods 12, 641–644 (2015).
Meddens, M. B. M. et al. Biomed. Opt. Express 7, 2219–2236 (2016).
Zhao, T. et al. Sci. Rep. 6, 26159 (2016).
Sharonov, A. & Hochstrasser, R. M. Proc. Natl Acad. Sci. USA 103, 18911–18916 (2006).
Beghin, A. et al. Nat. Methods 14, 1184–1190 (2017).
Dean, K. M. et al. Optica 4, 263–271 (2017).
Royer, L. A. et al. Nat. Biotechnol. 34, 1267–1278 (2016).
Mcgorty, R., Xie, D. & Huang, B. Opt. Express 25, 17798–17810 (2017).
Botcherby, E. J., Juškaitis, R., Booth, M. J. & Wilson, T. Opt. Commun. 281, 880–887 (2008).
Kumar, M. & Kozorovitskiy, Y. Opt. Lett. 44, 1706–1709 (2019).
Edelstein, A. D. et al. J. Biol. Methods 1, e10 (2014).
Goddard, T. D. et al. Protein Sci. 27, 14–25 (2018).
Feng, S. et al. Nat. Commun. 8, 370 (2017).
Leonetti, M. D., Sekine, S., Kamiyama, D., Weissman, J. S. & Huang, B. Proc. Natl Acad. Sci. USA 113, E3501–E3508 (2016).
Huang, B., Wang, W., Bates, M. & Zhuang, X. Science 319, 810–813 (2008).
Wang, Y. et al. Opt. Express 22, 15982–15991 (2014).
We acknowledge T. Goddard and T. Ferrin from UCSF for help in using ChimeraX. We thank E. Hamid from Nikon for help in providing the technical information of the objective lenses. We thank Tuscen Photonics Co. Ltd for generously loaning the 400BSI sCMOS camera to us. This project is supported by National Institutes of Health (grant nos. R33EB019784, R21EB022798 and R01GM124334 to B.H.; R35GM118119 to R.D.M.), UCSF Program for Breakthroughs in Biomedical Research and the Byers Award in Basic Science (to B.H.). V.P. was supported by a predoctoral fellowship from the American Heart Association. X.C. was supported by an international exchange fellowship from the Chinese Scholar Council. B.H. is a Chan Zuckerberg Biohub Investigator. R.D.M. is supported by the Howard Hughes Medical Institute. M.D.L. is supported by the Chan Zuckerberg Biohub.
Integrated supplementary information
O1-O3: objective lenses; M1-M8: mirrors; L1-L9: achromatic plano-concave lenses; CL1-CL3: cylindrical lenses; DM1-DM3: dichroic mirrors; TS1-TS5: translation stages; AOM: acoustic-optical modulator; EF: emission filter. Focal lengths of all lenses: 200 mm (L1), 75 mm (L2), 125 mm (L3), 150 mm (L4), 125 mm (L5), 100 mm (L6), 150 mm (L7), 45 mm (L8), 100 mm (L1), 100 mm (CL1), 200 mm (CL2) and 50 mm (CL3). The diameter of the pinhole is 30 µm. In the case of the Bessel light-sheet, the slit is replaced with a photomask.
The epi-fluorescence microscope part is on the left underneath the motorized stage.
a, A complete presentation of the microscope. b, Zoomed-in view of the remote imaging module, which contains O2, O3 and the water chamber. The chamber is filled with water (shown in green) and has a glass coverslip glued to its surface to allow an air–water interface between O2 and O3.
Blue shades indicate the light cone collected by each objective; black arcs indicate the extent of the NA (full-aperture angle 2θ) of each objective; red lines represent the light rays with the highest incident angle; black dashed lines show the optical axis. The intermediate image at the focal space of O2 is magnified by 1.33 (equal to NA1/NA2) along both the lateral and axial directions to minimize the aberration of the intermediate image. The blue dashed line indicates the position of the coverslip where the air–water interface is. All the light passing though O2 is refracted at this interface and enters O3, with the exception of only a small portion of the light being cropped by the coverslip (left side between the blue and red dashed lines). The effective detection NA along the x'-axis is ~ 1.33 × 1.33 × sin (73.8°/2) ≈ 1.06. The effective NA along the y axis can be obtained straightforwardly, since light is not cropped along this direction. We have therefore NAy ~ 0.9 × 1.33 ≈ 1.20, e.g., NA2 multiplying the magnification of the intermediate image. The collection solid angle of the system can be estimated as ~ 3.34, calculated by subtracting the cropped light at the air–water interface from the solid angle of O2. Assuming isotropic fluorescence emission, the resulting collection efficiency is then ~ 28.2%, equivalent to that of a water-immersion objective of NA ≈ 1.17. As an alternative way of understanding the remote imaging module and the reason for the high detection NA, O3 can be also seen as an air objective with NA = 1 where the collection solid angle is 2π. All light transmitted by O2 can then be collected by O3, resulting a high collection efficiency.
At the top is the optical layout of the detection path of the eSPIM simulated with Zemax. O1–O3: objective lenses—respectively, water immersion 60×/1.27-NA, air 100×/0.9-NA and water immersion 60×/1.0-NA. The objectives are modeled using parameters that were inferred from patent literature of similar objectives from the same producer: 60×/1.27-NA water-immersion objective from patent US8199408, 100×/0.9-NA air objective from patent US9341832, and 60×/1.0-NA water-immersion objective from patent US7133212. L1–L5: achromatic doublet lenses, with focal lenses as 200 mm, 75 mm, 125 mm, 150 mm and 125 mm. The parameters of those lenses are obtained from the Thorlabs website. At the bottom, the Huygens PSFs (image in the camera/detector plane) are given for five different position (i–v) in the focal space of O1. The FWHMs along the x- and y-axes are, respectively, 255 ± 2 nm and 355 ± 5 nm (average of the measurement from the five PSFs, mean ± s.d.).
a,b, The simulated intensity maps of the oblique Gaussian and Bessel light sheet within the x'z' plane. Insets of a and b show the corresponding pupil plane, with the red circles indicating the full extent of the pupil of the objective, and the white lines the effective illumination laser pattern (with a uniform intensity) at the pupil plane. The pupil function to generate the Gaussian light sheet comprises light with incident angles ranging from 47.28° to 72.72° (i.e., 60° ± 12.72°). The pupil function to generate the Bessel light sheet is segmented into two parts, one with incident light from 47.28° to 52.72°, the other from 67.28° to 72.72°. c,d, The simulated cross-section profiles along the z'- and x'-axes crossing the peak intensities of the intensity maps in a and b. The FWHMs (of the central peak) along the z'-axis are 730 nm and 554 nm for the Gaussian and Bessel beam correspondingly, while the FWHMs along the x'-axis are 12.8 μm and 18.6 μm.
Maximal intensity projections in three different planes are shown. Five experiments were repeated independently, with similar results.
a–c, The pupil function of the microscope is obtained using the iterative phase-retrieval method1 on the PSF of the microscope (see Fig. 1b). The amplitude map of the pupil function is shown in a. The light is compressed and shifted toward the left, which agrees with the analysis in Supplementary Fig. 4. The phase map of the pupil function is first fitted with the first 15 Zernike polynomials and then, as shown in b, with the first four Zernike polynomials removed. The coefficients of the 5–15 Zernike polynomials are shown in c. The Strehl ratio of the phase map shown in b is 0.90, which indicates that our microscope has a performance close to the diffraction limit. Three experiments were repeated independently, with similar results.
a,b, Cells are transfected with EGFP-α-tubulin (a) and mRuby2-TOMM20-N-10 (b). The maximum intensity projection (MIP) in the x–y plane is shown together with two representative x–z and y–z slices along the dashed lines. The regions outlined by the dashed boxes are also shown as magnified images. The magnified region in b is shown for four different time points; images are taken at 1.6 s/volume. The white triangles indicate the place where the cell undergoes structural changes. Three experiments were repeated independently, with similar results.
The HEK 293T cells have endogenous lamin A/C labeled by mNeonGreen211 knock-in (green) and lysosomes stained by LysoTracker Deep Red dye (red). The maximum intensity projection in the x–y plane and representative x–z and y–z slices along the dashed lines are shown. Three experiments were repeated independently, with similar results.
Supplementary Figure 11 Comparison of photobleaching observed with a spinning-disk confocal microscope and with the eSPIM system.
For both microscopes, volumetric fluorescence images of HEK 293T cells with endogenous clathrin light chain A labeled by mNeoGreen211 knock-in were recorded. The exposure conditions were adjusted so that similar signal levels were obtained on both microscopes. When setting up the parameters of the spinning-disk confocal microscope so that the SNR is comparable to that of the eSPIM, we focused on the bottom of the cell where the features are the sharpest. The pixel intensity of all slices (36 slices for spinning disk and 51 for eSPIM, adjacent slices separated by 400 nm in both cases) within each volumetric dataset is summed, normalized to the first volume and then plotted as a function of the volume number. The data are fitted with a double exponential decay function. Inset shows the plot of the first 150 data points. The first volumetric images in both modes have approximately the same signal-to-noise ratio. The objective used in the confocal experiment is Nikon CFI Plan Apo VC 100× Oil (NA = 1.40). The pixel sizes are, respectively, 127.6 nm and 133.0 nm for the confocal and SPIM images. Three experiments were repeated independently, with similar results.
A Drosophila S2 cell with lysosomes labeled with LysoTracker Deep Red dye. Four maximum intensity projection images at different time points are shown. The triangles highlight two noticeable events. Images were acquired at 15 volumes per second (volume size 35 µm × 35 µm × 7 µm, 34 slices per volume and camera running at 500 frames per second). At this imaging speed, we could reliably perform 3D tracking of lysosome movement dynamics. See also Supplementary Movies 8–10. Three experiments were repeated independently, with similar results.
Supplementary Figure 13 Super-resolution stochastic optical reconstruction microscopy (STORM) imaging using the eSPIM system.
S2 cells have microtubules immunostained with Alexa Fluor 647. a, STORM images of six adjacent planes in the x'y plane. The advantage of light-sheet illumination in minimizing out-of-focus bleaching is evident from the similar image qualities from earlier- to later-acquired images (all using identical acquisition parameters). b, The projection of the images in a onto the xy plane. c,d, Diffraction-limited (c) and super-resolved (d) images of the area in the dashed rectangle in a. e, The cross-sectional profile along the dashed line in d. The super-resolved images clearly show the ultra-structure of the microtubules with a resolution (FWHM) down to 49.1 nm. f, The statistical distribution of the photon number of single molecules (data from the cell shown in a (z' = 0 µm)). The average photon number per photoswitching cycle is 2,568 ± 109 (average of 5 different datasets; the error is the s.d.). As a comparison, we imaged the same sample using another home-built STORM microscope. Under near-total internal reflection (TIR) conditions using a 60×/1.40-NA objective, the average photon number is 9,184 ± 287 (average of 5 different datasets, mean ± s.d.), 3.57 times of that collected with eSPIM. This difference was mostly the result of the higher collection efficiency of an oil-immersion objective, and the performance of an oil-immersion objective degrades quickly when imaging away from the coverslip surface because of refractive index mismatch between glass and water. On the other hand, using the same 1.27-NA water-immersion objective under high-inclined illumination conditions (TIR is impossible in this case) did not produce single-molecule data of sufficient quality to be reliably analyzed owing to a high background. Therefore, the oil-immersion objective with near-TIR illumination is better for flat samples near the surface, whereas the eSPIM system is more advantageous for thicker samples away from the coverslip surface. Five experiments were repeated independently, with similar results.
Supplementary Figure 14 Scheme of the photomask used to generate the oblique Gaussian and Bessel light sheets.
The mask contains patterns to generate both the Gaussian and Bessel light sheets. The area surrounded by red rectangles can transmit light, while the rest of the mask blocks the light. The patterns are organized in pairs, with that for Bessel light-sheet generation on the left and that for Gaussian light-sheet generation on the right. There are in total 13 × 13 patterns for each mode. The width of the patterns ranges from 0.7 mm to 1.0 mm from left to right. The patterns for Bessel beam generation have gaps (light blocking area) in the center, with the width of this gap ranging from 0.2 mm to 0.5 mm from top to bottom. The photomask was designed so that we could have Gaussian light sheets and especially Bessel light sheets with different characteristics. The pattern with a width of 0.825 mm and a gap of 0.30 mm was used for the Bessel beam generation in this study.
The region of interest of the sCMOS camera is 640 × 300 pixels. The camera works in rolling shutter mode where the readout is sequential for each row. In continuous mode, the laser illumination (blue rectangles) and readout (blue dashed lines) are simultaneous with texposure ≥ treadout, whereas in global exposure mode, the illumination and readout are sequential with the effective imaging time equaling texposure + treadout. treadout is 1.46 ms for a region of interest of 640 × 300 pixels. The two fluorescent bead images at the bottom are the raw images of one representative slice within an eSPIM stack, taken respectively in the two modes. The colored triangles highlight where the images of the beads appear ‘doublet’ in the continuous mode and normal in the global mode. Three experiments were repeated independently, with similar results.
Supplementary Figs. 1–15 and Supplementary Notes 1–4
Three-dimensional design file for the water container serving as the objective interface in the remote focusing module (.stl format for three-dimensional printing).
Three-dimensional volume rendering of HeLa cell with transient over-expression of EGFP-α-tubulin. See also Supplementary Fig. 9a. Three experiments were repeated independently, with similar results.
Four-dimensional dynamic volume rendering of HeLa cell with transient over-expression of mRuby2-TOM20. See also Supplementary Fig. 9b. Three experiments were repeated independently, with similar results.
Four-dimensional dynamic volume rendering (top view) of HEK 293T cells (mNeonGreen2 (ref. 11) knock-in for CLTA gene). See also Fig. 1d. At least five experiments were repeated independently, with similar results.
Four-dimensional dynamic volume rendering (side view) of HEK 293T cells (mNeonGreen2 (ref. 11) knock-in for CLTA gene). See also Fig. 1d. At least five experiments were repeated independently, with similar results.
Four-dimensional dynamic volume rendering (top view) of two-color images of nuclear lamina and lysosomes in HEK 293T cells (mNeonGreen2 (ref. 11) knock-in for LMNA gene and LysoTracker Deep Red staining, respectively). See also Supplementary Fig. 10. Three experiments were repeated independently, with similar results.
Four-dimensional dynamic volume rendering (side view) of two-color images of nuclear lamina and lysosomes in HEK 293T cells (mNeonGreen2 (ref. 11) knock-in for LMNA gene and LysoTracker Deep Red staining, respectively). See also Supplementary Fig. 10. Three experiments were repeated independently, with similar results.
Four-dimensional dynamic volume rendering of B16F1 mouse cells stably expressing membrane marker CAAX-EGFP. Three experiments were repeated independently, with similar results.
Three-dimensional volume rendering of Drosophila S2 cells with lysosomes labeled with LysoTracker Deep Red. See also Supplementary Fig. 12. Three experiments were repeated independently, with similar results.
Four-dimensional dynamic volume rendering (top view) of Drosophila S2 cells with lysosomes labeled with LysoTracker Deep Red. See also Supplementary Fig. 12. Three experiments were repeated independently, with similar results.
Four-dimensional dynamic volume rendering (side view) of Drosophila S2 cells with lysosomes labeled with LysoTracker Deep Red. See also Supplementary Fig. 12. Three experiments were repeated independently, with similar results.
Parallel long-term imaging of HEK 293T cells in 16-well plates. See also Fig. 2a. Three experiments were repeated independently, with similar results.
Parallel long-term imaging of HEK 293T cells in 16-well plates (zoom-in view). See also Fig. 2a. Three experiments were repeated independently, with similar results.
Parallel imaging of S2 cells in 96-well plates with drug treatment. Three experiments were repeated independently, with similar results.