Deep3DSIM: Super-resolution imaging of thick tissue using 3D structured illumination with adaptive optics
- Department of Biochemistry, University of Oxford, United Kingdom
- Department of Engineering, University of Oxford, United Kingdom
Figures
Figure 1
Simplified overview of the Deep3DSIM setup.
(A) Optical arrangement of Deep3DSIM and the conceptual use of micromanipulators for applications such as microinjection and electrophysiology. The excitation light hits the spatial light modulator (SLM), which is conjugated to the object/image plane, and then reflects off a deformable mirror (DM) before being focused on the sample with an objective lens. In the imaging path, the reflected fluorescence is coupled off via a dichroic beam splitter and then collected by separate cameras for each channel. (B) Different imaging modes are enabled in parallel on Deep3DSIM. Deep imaging without adaptive optics (AO) usually leads to an aberrated point spread function (PSF) due to refractive index (RI) mismatch and sample inhomogeneity. Sensorless AO correction compensates for sample-induced aberrations. Remote focusing enables fast focusing at various depths without moving the specimen/objective. Combination of AO and remote focusing produces aberration-corrected imaging at different depths, without mechanical movement. (C) Deep3DSIM is controlled by Cockpit, which consists of three packages: python-microscope is responsible for the control of all hardware devices; microscope-cockpit provides a user-friendly GUI; microscope-aotools provides the AO functionality. The system uses a real-time controller, in this case, the Red Pitaya STEMlab 125–14, to coordinate different instruments for image acquisition with structured or widefield illumination.
Figure 2
Resolution estimations of the Deep3DSIM system.
(A) Full width at half-maximum (FWHM) measurements of 100 nm green fluorescent beads, deposited on a glass cover slip and imaged in water-immersion configuration. The images show lateral (XY) and axial (XZ) views of one bead from each sample, indicated with a cross on the distribution plots. SD: standard deviation. Scale bar: 1 µm. (B) Frequency space analysis of images of fixed COS-7 cells, with labelled microtubules in the green channel and endoplasmic reticulum in the red channel. The images were acquired in water-immersion configuration, imaging through a cover slip. Scale bars: 10 µm. The XY images are average projections along Z, and the XZ images are orthogonal projections along the middle of the Y axis. The power spectra are in logarithmic scale and centred on the zero frequency; they show lateral kxyview (top panel) and an axial kxzview (bottom panel) of the 3D DFT, both taken along the middle of the corresponding third axis. Scale bar: 3 µm–1. The power spectra were thresholded to remove noise. The power plots show the attenuation of the frequency response with increasing resolution. The lateral power plot was created by radial averaging of the frequency amplitudes in the kxy view, while the axial plot shows the frequency amplitudes of the kxz view after averaging the two directions 0 → ± kz and then averaging along the kx axis. The resolution thresholds (dashed lines) were chosen as the points at which the normalised logarithmic power reached 0.01 (i.e. 1%). The resolution thresholds were converted to the spatial domain by taking the inverse of the spatial frequency.
Figure 3 with 1 supplement
Adaptive optics (AO) correction removes artefacts and improves the contrast and resolution of 3D structured illumination microscopy (3D-SIM) in tissue imaging.
(A, B) 3D-SIM images of neuromuscular junction in a fixed L3 Drosophila larva expressing Nrx-IV::GFP (green) and membrane labelled with anti-HRP antibodies (magenta) acquired without (A) and with AO (B), respectively. The data was acquired in water-immersion mode, imaging through a cover slip. The XZ views on the left are maximum projections along the Y axis. The boxed regions are shown scaled up on the right, with their own orthogonal XZ views along the middle of the Y axis. The arrow in (A) shows an example of ghosting artefacts in the red channel. The scale bars are 10 μm. (C) Intensity profiles along the dashed lines in the XZ views on the right of (A) and (B).
Figure 3—figure supplement 1
Frequency-space analysis and aberration correction.
(A) Power spectra of the green (Nrx-IV) and red (HRP) channels, and their associated plots. Scale bar: 3 µm–1. (B) Example of aberration correction mode amplitudes for this type of sample.
Figure 4 with 1 supplement
3D structured illumination microscopy (3D-SIM) in deep tissue samples enabled by adaptive optics (AO) aberration correction.
(A) 3D-SIM image stack of fixed Drosophila L3 larval brain expressing Cno::YFP and imaged from top downwards at a depth of ~3 μm from the cover slip surface without AO (left) and with AO (right). Both the lateral views (XY) and orthogonal views (XZ) show maximum projections. Scale bar: 10 μm. (B) 3D-SIM images of the same fixed Drosophila L3 larval brain, acquired through the entire volume of a single brain lobe at ~130 μm from the top surface, with 100 nm diameter red fluorescent beads attached to the surface of the glass slide. Images are displayed as in (A) but magnified. The insets show further 3× magnification of the regions indicated with dashed lines. Scale bar: 10 μm. (C) Intensity profiles of the red channel along the lines in the insets. (D) Schematic illustration of the specimen mounting and imaging. The larval brain was mounted in PBS between a glass cover slip (top) and a glass slide (bottom); the slide was coated with red fluorescent beads at medium density. Image stacks in (A) and (B) were recorded at the proximal (P) and distal (D) sites indicated with arrowheads on the right, respectively.
Figure 4—figure supplement 1
Frequency-space analysis and aberration correction.
(A) Power spectra and plots for the proximal site images. Scale bar: 3 μm–1. (B) Power spectra and plots for the distal site images. Left column is the green channel, right column is the red channel. The lateral resolution is not improved with adaptive optics (AO) in the green channel, because the prominent reconstruction artefacts in bypass (BP) mode misrepresent the actual optical performance of the system. Scale bar: 3 μm–1. (C) Example of aberration correction mode amplitudes obtained at the distal site.
Figure 5 with 1 supplement
Remote focusing for fast live 3D structured illumination microscopy (3D-SIM) time-lapse imaging and for large volume imaging using multi-position aberration correction (MPAC).
(A) Mitotic embryonic divisions in Drosophila syncytial blastoderm embryos, expressing transgenic Jupiter::GFP labelling microtubules (green) and transgenic Histone H2A::RFP labelling chromosomes (magenta). The imaging was carried out in water-dipping configuration, without the use of a cover slip. Images show maximum projection of volumes with about 1 µm thickness. Brightness and contrast of each image were adjusted independently. Scale bar: 10 µm. (B) COS-7 cell in metaphase with microtubules immunostained with AF 488, visualising the mitotic spindle. The volumes were acquired with remote focusing, imaging through a cover slip in water-immersion configuration. The lateral (XY) views are maximum projections. Aberration correction was performed at two imaging planes, indicated by arrows in the adaptive optics (AO) YZ view; dynamic correction was applied for all other planes. Insets show magnifications of the regions indicated with dashed boxes. Scale bars: 10 μm in the lateral bypass (BP) view and 0.5 μm in the insets. (C) Intensity plots in the lateral and axial directions, along the dashed lines in the insets, showing increased resolution in the AO (MPAC) case.
Figure 5—figure supplement 1
Multi-position aberration correction (MPAC) mode amplitudes.
The two correction planes, used for the MPAC, are indicated with arrows in the inset image. The scale bar is 5 μm.
Appendix 1—figure 1
Structured illumination microscopy (SIM) pattern generation.
(A) Example of the structured illumination (SI) pattern at one of the angle orientations and phase shifts, visualised by imaging a monolayer of 100 nm diameter fluorescent beads in the green (left) and in the red (middle) channels. The power spectrum (right) shows combined frequency response in all three angles of the green channel, with the centre masked with a black circle to create a better contrast. Scale bars: 5 μm. (B) Fluorescence signal modulated with structured illumination, taken from the centre of a single red fluorescent bead. The plot shows three 50 μm scans through the bead, one for each of the SI angles.
Appendix 2—figure 1
Effect of rescaling on remote focusing.
(A) Comparison of remote focusing to conventional refocusing with piezo stage. (B) Demonstration of how pixel rescaling restores the axial profile of bead images acquired with remote focusing. (C) Displacement comparison between piezo (left) and remote focusing (right). Scale bars: 1 μm.
Appendix 4—figure 1
Optical arrangement.
405 nm, 488 nm, 561 nm, and 640 nm: laser sources. S: mechanical shutter. BX: beam expander. F1 to F2: optical filters. M: mirrors. λ/2: half-wave plates. L1 to L13: lenses, whose focal lengths are given in the table at the top of the figure. P: polariser. SLM: spatial light modulator. A1 to A3: apertures. FM1 to FM6: motorised flip mirrors. PR: polarisation rotator. BS: beamsplitter. FBS: 50/50 beamsplitter on flip mount. D1 to D5: dichroic beamsplitters. DM: deformable mirror. C1 to C2 are EMCCD cameras, and C3 is a CCD camera for the wavefront sensing interferometer. 10× and 60× are objective lenses. ST: assembly of stages (X, Y, and Z movement). LL: LED light used for brightfield imaging.
Appendix 6—figure 1
Deformable mirror (DM) drift and temperature characterisation.
Wavefront phase root mean square (RMS) error (RMS error, RMSE) measurements for (A) long-term drift without warm-up and for (B) short-term drift with warm-up (4 hr of 1 rad defocus), following the sequences listed in Appendix 6—table 1. The two plots in (B) show shapes held for 20 min (top) and the final shape held for 60 min (bottom).
Appendix 6—figure 2
Actuator control signal patterns used in the deformable mirror (DM) exercise procedures.
(A) Checkerboard. (B) Refocusing. (C) Individual poking.
Tables
Appendix 6—table 1
Experimental conditions.
| Shape | Zernike mode (Noll index) | Holding time [min] |
|---|---|---|
| (defocus) | 240 | |
| (primary oblique astigmatism) | 20 | |
| (primary vertical astigmatism) | 20 | |
| (primary vertical coma) | 20 | |
| (primary horizontal coma) | 20 | |
| (primary spherical) | 20 | |
| (secondary spherical) | 60 |
Additional files
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Deep3DSIM: Super-resolution imaging of thick tissue using 3D structured illumination with adaptive optics
eLife 14:e102144.
https://doi.org/10.7554/eLife.102144
