Open Access
Add to BibSonomy
Abstract
A new fluorescence microscopy technique for optical sectioning was investigated. This technique combined Spinning Disk microscopy (SD) with Structured Illumination Microscopy (SIM), resulting in more background removal than either method. Spinning Disk Structured Illumination Microscopy (SD-SIM) resulted in higher signal-to-background ratios. The method detected and quantified a dendritic spine neck that was impossible to detect with either SIM or SD alone.
© 2023 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Optical sectioning is an essential part of fluorescence microscopy in which out-of-focus light is removed. Optical sectioning is especially important in thick or densely labeled samples such as when imaging the brain or other tissues [1–4]. Techniques include confocal microscopy, structured illumination microscopy (SIM), and spinning disk (SD) microscopy [5–11]. Both SD and confocal microscopy use pinholes to block out-of-focus light from reaching the detector. Confocal microscopes are limited to scanning a single illumination beam across the focal plane to collect an image. Spinning disk microscopes use a disk, spun at high speeds, with pinholes to collect an optically sectioned image within a single exposure. SIM is also a widefield imaging method based on a camera detector, but requires multiple exposures. The excitation illumination is patterned, typically using spatial light modulators (SLMs) or diffraction gratings, with the pattern being shifted until the entire sample has had equal exposure. An image is acquired for each pattern position and then a single image of the focal plane is reconstructed through various methods from those images.
SIM is also capable of super resolution (SR-SIM) imaging, which can be accomplished using a number of methods [12–15]. SD microscopes have also been shown to be capable of surpassing the diffraction limit using photon reassignment methods [16,17].
In this work, a setup for Spinning Disk Structured Illumination Microscopy (SD-SIM) is characterized with special attention focused on optical sectioning and the removal of background light in comparison to SD or SIM alone. Recently, a new method of SIM optical sectioning using a phase-modulated spinning disk (PMSD) with confocal pinholes was reported [18]. While this method and SD-SIM both use a spinning disk, the methods have a number of key differences including pattern formation and image reconstruction. As far as the authors are aware, these spinning disk and structured illumination techniques have not been previously combined in this way into a single imaging setup.
2. Microscope design
The SD-SIM microscope illuminates a spinning disk with structured illumination to provide optical sectioning that exceeds both individual methods. The setup is shown in Fig. 1 and is based on an IX83 microscope (Olympus, Tokyo, Japan). Excitation illumination is provided by inexpensive 445 nm and 532 nm lasers. Laser illumination passes through a speckle reducer (Optotune) and is collected by a multi-mode optical fiber, 1 mm in diameter (Thor Labs). After passing through the fiber, the light is collimated by a 50 mm FL lens (Thor Labs) and polarized by a linear polarizer (Thor Labs). The polarized light reflects off a polarizing beam splitter (PBS) cube (Thor Labs) and onto a Liquid Crystal On Silicon (LCOS) microsdisplay (Fourth Dimension Displays) which displays the desired SIM pattern. Light that hits an on-pixel has its polarization rotated by $\sim 90^{\circ }$ and is passed by the PBS, thereby structuring the illumination. The structured light is imaged onto the spinning disk by a series of lenses, during which it reflects off a double-band dichroic mirror (ZT442/532 RPC, Chroma). After passing through the disk, the patterned illumination is finally imaged onto the sample through the tube lens and objective. This configuration results in an incoherent illumination pattern meaning that the incoherent optical transfer function applies [19].
Emitted fluorescence was then collected by the objective, passed back through the tube lens, disk, and dichroic. The disk removed out-of-focus light through the confocal effect. Emission light was then filtered and imaged onto the camera (Andor). Emission filtering was done using two filters, a dual band fluorescence emission filter matched with the dichroic mirror, with additional GFP (used for mouse brain images), RFP (used for the nanobeads and thin film), or Cy3 (used for the rat epididymis) filters located in the filter wheel. The additional filtering was required to completely remove excitation light that reflected off of the spinning disk. It should be noted that available materials and space required the lenses surrounding the dichroic mirror to be spaced such that the image on the camera has an additional 1.096x magnification. This results in a total magnification of 109.6x for the system when using a 100x objective.
The disk was designed using the steps and Mathematica code provided by Halpern et al. [20]. The disk was designed for a 100x/1.4 NA oil immersion objective (Olympus) with 50 $\mu$m diameter pinholes arranged in Archimedean spirals. The pinholes were spaced with a constant arclength spacing of 500 $\mu$m apart resulting in a 1% fill factor. This cut down on total illumination but reduced cross-talk between pinholes and is more suitable for thick samples [21]. Mounting and timing holes for a SR540 optical chopper (Stanford Research Systems, Sunnyvale, CA) were added to the disk’s design in FreeCAD 0.19 (https://www.freecad.org/). The disk was obtained from Arizona Micro Inc. (Las Vegas, NV). Rotation speed was limited to 400 Hz to ensure disk stability.
2.1 SIM reconstruction
SIM images are formed by combining multiple images together, each with a different position (or phase) of the illuminating pattern. Generally, each position is the same pattern offset by a phase shift. When all the phases are added together, the resulting illumination should be homogeneous [22]. The pattern used in this study was a series of lines and reconstruction was done using two methods, averaging and Homodyne. An example of the patterned SD-SIM illumination is shown in Supplement 1. The simplest method of image reconstruction averages all individual images with patterned illumination, $I_n$, together
where a pattern with $N$ total iterations is used. This method is equivalent to a wide field image, and does not provide optical sectioning. This reconstruction method was used for all wide field (WF) images used in this work.Optically sectioned SIM images can be obtained through square law reconstruction,
where $I_1$, $I_2$, and $I_3$ are images containing the patterned illumination with a phase difference of 0, $2 \pi /3$, and $4 \pi /3$ between the patterns. An alternative method, Homodyne detection, given by where $I_n$ is an image with the $n^{th}$ phase of $N$ total phases, can also be used [23,24]. Homodyne detection was used for SIM and SD-SIM images shown in this work.2.2 Data collection
Data was collected in Andor IQ3 software and processed in SIMToolbox, ImageJ, Napari, MatLab (MathWorks, Natick, MA), and SRMeasure [24,25]. Lateral and axial resolution was measured using FluoSpheres Carboxylate-Modified Microspheres (Thermofisher, F8800) and optical sectioning was measured using a thin film of rhodamine in polyvinyl alcohol polymer. Biological samples used were an optically cleared, coronal slice of a mouse brain expressing green fluorescent protein (GFP) in a subset of neurons (SUNJin Lab, Hsinchu City, Taiwan) approximately 150 $\mu$m thick. A second sample was a rat epididymis thin slice stained with hematoxylin and eosin.
Spinning disk images were acquired using an illumination pattern with all pixels turned on and required no reconstruction. SIM and SD-SIM images used a pattern of parallel lines made by two rows of pixels turned on then 10 rows of pixels turned off. Finer, higher frequency patterns tend to result in higher resolution images, but the disk blocks out a substantial amount of illumination necessitating the thicker lines to maximize illumination reaching the sample [11]. Even so the two-pixel-on pattern results in nearly diffraction-limited lines widths of $\sim$272 nm using the 100x objective (microdisplay pixel size 13.62 $\mu$m). The larger distance between the illuminating lines also helped minimize background in thick samples such as the mouse brain [26].
3. Results and discussion
Optical sectioning strength was measured using a thin fluorescent film. A $z$-stack image was taken using a $z$-step size of 50 nm. The average intensity was calculated for each $z$-slice. The $z$-axis intensity profile was then fitted to a Gaussian distribution using MatLab (See Supplement 1). The full width at half max (FWHM) was calculated from the fitting parameters and gave the axial resolution for each method. The resulting measured thicknesses can be found in Table 1.
Table 1. Measurements of axial and lateral resolution for each imaging method obtained from measuring nanobeads.
Each imaging method was checked for optical sectioning homogeneity to ensure even optical sectioning across the field of view by examining optical sectioning heat maps. These maps were obtained in the same way the average lateral resolution was, except each image was broken into an 8-by-8 grid. Each region of the grid was then used to create a $z$-axis intensity profile and Gaussian fit. These maps are shown in Fig. 2. A map was not able to be obtained for WF since the amount of out-of-focus light was too large. All optical sectioning methods used show stronger optical sectioning near the center with a slight offset towards the top. This is due to slightly uneven illumination on the microsdisplay from the optical fiber. The set-up with the spinning disk in place also shows some vignetting in the lower right corner since the pinholes do not perfectly cover the entire image. It should be noted that all images shown in this work have been cropped to remove vignetting. The vignetting can be further reduced by increasing the pinhole sector size on the disk. The obtained optical sectioning thickness values are similar to other SIM and SD microscopes setups which have reported values around 500 nm [20]. However, the combination of spinning disk and SIM resulted in an improvement of optical sectioning by 13% over SD alone and 25% over SIM alone. Thus further optimisation of the setup, including smaller pinholes on the disk, should improve SD-SIM’s optical sectioning.
Fig. 2. Optical sectioning heat maps for SD, SIM, and SD-SIM. Each grid section covers a 256 pixel x 256 pixel area on the camera (15.2 $\mu$m by 15.2 $\mu$m in the sample). The optical sectioning power for a section was obtained from a $z$-stack image of a thin fluorescent film with a $z$-step of 50 nm. The $z$-axis intensity profile for each section was fitted to a Gaussian distribution whose FWHM corresponds to the thickness of the thin film.
Lateral and axial resolution were measured using 100 nm fluorescent nanobeads. In-focus frames and orthogonal views of the beads can be seen in Fig. 3 and values for the resolutions can be found in Table 1. Lateral resolution was analyzed by visually selecting the frame with the most in-focus beads and then processed using ThunderSTORM [27]. These results include a 50 nm minimum standard deviation to remove outliers such as hot pixels. All lateral resolutions are above the diffraction limit as expected and fall within each margin of error. Axial resolution was measured using the Napari PSF analysis plugin on a minimum of 12 randomly chosen beads for each method (https://github.com/fmi-faim/napari-psf-analysis/tree/main). The final value for axial resolution was then obtained by a weighted average of the FWHMs. SD-SIM improved axial resolution by $\sim$150 nm over SIM and $\sim$100 nm over SD. Additionally, Fig. 3 shows SIM has a lower total background than SD, while the side profiles for SD are tighter. SD-SIM has both low background and a tighter point spread function (PSF) showing improvement over both individual methods.
Fig. 3. Nanobeads imaged with each method, with $xz$- and $yz$-plane projections. Both SD and SIM remove a significant amount of out-of-focus light, but SD-SIM is able to remove more as can be seen by its small profile. A $z$-step size of 50 nm was used. Intensity was scaled from zero to the maximum intensity for each individual image.
Two biological samples were imaged to test SD-SIM. The rat epididymis in Fig. 4 is representative of thin samples and has easily visualized cross sections. Of the optical sectioning methods, SD has the largest background and SD-SIM has the lowest. The orthogonal views highlight the improved axial resolution of SD-SIM. Neighboring sperm tails are more distinct from one another since more out-of-focus light has been removed. The resolution and signal-to-noise ratio (SNR) in maximum intensity projection images were measured with SRMeasure and are shown in Table 2. It should be noted that SD-SIM image has improved lateral resolution despite being diffraction limited like the other methods. The improvement to resolution comes from the removal of out-of-focus light that obscures fine details, as well as the effect from the confocal pinholes. The WF SNR is boosted by out-of-focus light, while SD-SIM has the second highest. The pinholes in the spinning disk let about 1% of excitation light through, resulting in a much lower signal level. A 2 O.D. neutral destiny filter was used to decrease excitation illumination when the disk was not used. This allowed imaging with a consistent exposure time between all methods without saturation. However, SIM still has a lower SNR than SD-SIM, even with higher signal intensity. The signal-to-background ratio was measured by finding the mean counts in three randomly chosen areas of background, and in three randomly chosen areas of signal in the maximum intensity projection images. Again, SD-SIM shows the largest SBR, closely followed by SIM.
Fig. 4. Rat epididymis imaged with WF, SD, SIM, and SD-SIM. The maximum intensity projections (a) and orthogonal $yz$-views (b) are shown. All images have been scaled from 0 to the maximum intensity in the given image. An exposure time of 1 second was used for each method to ensure background and noise was comparable. SD-SIM has the lowest background level of all methods, as well as the highest SBR. The zoomed in $yz$-orthogonal views (c) show the best distinctions between neighboring sperm tails in SD-SIM over the other methods.
Table 2. Image statistics for the rat epididymis images in Fig. 4.
A mouse neuron was also imaged using SD-SIM as shown in Fig. 5. The mouse brain sample was approximately 150 $\mu$m thick, making it a good demonstration of SD-SIM’s ability to image in thick, densely-labeled samples. The optical sectioning power of SD-SIM is more obvious here, all images were scaled from 0 to an intensity chosen to be best for visual clarity. WF had to have the intensity scale limited to a range similar to the other methods since the cell body was so bright. SD-SIM has the fewest average background counts, with less than half the background of of the next lowest. As previously mentioned, SD-SIM did have lower signal counts due to less excitation light reaching the sample. However the lowered background more than made up for the decrease, resulting in the highest signal-to-background ratio of the methods considered in this study as seen in Table 3. It should noted that the cell body as avoided when choosing 170 areas for signal and background due to large amounts of out-of-focus light in the WF 171 image. SD-SIM was able to cut out enough background counts to measure the neck of the dendritic spine seen in Fig. 5(c). Only SD-SIM has a clear profile, Fig. 5(d), that could be fit to a Gaussian distribution with a FWHM of 431 $\pm$ 103 nm. All other profiles had no clear distinction between the neck and background and any attempt at fitting failed.
Fig. 5. Mouse neuron imaged using various methods (a). Images were all scaled from zero to clearly demonstrate the background count decrease with SD-SIM. For every image besides WF, the maximum intensity displayed was determined on a per-image basis for visual clarity. WF was scaled in a similar range as the other images since the cell body was significantly brighter due to out-of-focus light. All images were acquired with an exposure time of 1 second. A zoomed in image of the cell body (b) in WF and SD-SIM. Detail in the cell body, as well as surrounding dendrites, are obscured by out-of-focus light in WF and are easily seen in SD-SIM. A zoomed in view of the other boxed area (c) showing a denditic spine. Color scaling was again done on a per-image basis to match background color between images and highlight SD-SIMs higher SBR. The profile of the dendritic spine’s neck, indicated by the dotted line is shown in (d) along with a Gaussian fit of the profile. The width of the spine was measured at 431 $\pm$ 103 nm using the FWHM of the fitted curve. The intensity profile was averaged over a line 8 pixels thick for each of the imaging methods, but WF, SD, and SIM had backgrounds that washed out the spine’s neck. The other profiles are shown in the Supplement 1.
Table 3. Image statistics for the mouse neuron images in Fig. 5.
4. Conclusion
A microscope setup combining spinning disk and structured illumination microscopy techniques was investigated with a focus on optical sectioning. Both SD and SIM naturally block illumination from reaching the sample and thus the combination of the two leads to a decrease in signal strength. However, the combined optical sectioning strength and improved axial resolution resulted in improved signal-to-background ratios. The optical sectioning strength made a dendritic spine detectable when it was not with SIM or SD. Further improvements could be done to the system, including synchronizing timing between the disk and camera, optimization of pinhole size, and better removal of laser speckle. Testing with super-resolution SIM patterns and reconstruction methods could also be investigated.
Funding
BioFrontiers Institute, University of Colorado Colorado Springs; National Institutes of Health (2R15GM128166-02).
Disclosures
The authors declare no conflicts of interest.
Data Availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
Supplemental document
See Supplement 1 for supporting content.
References
1. Z. Li, Q. Zhang, S. W. Chou, Z. Newman, R. Turcotte, R. Natan, Q. Dai, E. Y. Isacoff, and N. Ji, “Fast widefield imaging of neuronal structure and function with optical sectioning in vivo,” Sci. Adv. 6(19), eaaz3870 (2020).10.1126/sciadv.aaz3870 [CrossRef]
2. O. D. Supekar, A. Sias, S. R. Hansen, G. Martinez, G. C. Peet, X. Peng, V. M. Bright, E. G. Hughes, D. Restrepo, D. P. Shepherd, C. G. Welle, J. T. Gopinath, and E. A. Gibson, “Miniature structured illumination microscope for in vivo 3D imaging of brain structures with optical sectioning,” Biomed. Opt. Express 13(4), 2530–2541 (2022). [CrossRef]
3. T. C. Schlichenmeyer, M. Wang, K. N. Elfer, J. Q. Brown, A. O. Yossepowitch, J. A. Bjartell, M. Eastham, B. D. Graefen, P. I. Guillonneau, R. Karakiewicz, and F. Montironi, “Video-rate structured illumination microscopy for high-throughput imaging of large tissue areas,” Biomed. Opt. Express 5(2), 366–377 (2014). [CrossRef]
4. A. Nakano, “Spinning-disk Confocal Microscopy — A Cutting-Edge Tool for Imaging of Membrane Traffic,” Cell Struct. Funct. 27(5), 349–355 (2002). [CrossRef]
5. R. Heintzmann and C. G. Cremer, “Laterally modulated excitation microscopy: improvement of resolution by using a diffraction grating,” Proc. SPIE 3568, 185–196 (1999). [CrossRef]
6. A. C. Boccara, F. Chasles, and B. Dubertret, “Optimization and characterization of a structured illumination microscope,” Opt. Express 15(24), 16130–16140 (2007). [CrossRef]
7. J. Oreopoulos, R. Berman, and M. Browne, “Spinning-disk confocal microscopy: present technology and future trends,” Methods Cell Biol. 123, 153–175 (2014). [CrossRef]
8. T. Wilson, “Optical sectioning in fluorescence microscopy,” J. Microsc. 242(2), 111–116 (2011). [CrossRef]
9. M. Hadravský, M. Petrán, M. D. Egger, and R. Galambos, “Tandem-Scanning Reflected-Light Microscope,” J. Opt. Soc. Am. 58(5), 661–664 (1968). [CrossRef]
10. K. A. Johnson, D. Noble, R. Machado, T. C. Paul, and G. M. Hagen, “Flexible Multiplane Structured Illumination Microscope with a Four-Camera Detector,” Photonics 9(7), 501 (2022). [CrossRef]
11. M. G. Gustafsson, “Surpassing the lateral resolution limit by a factor of two using structured illumination microscopy,” J. Microsc. 198(2), 82–87 (2000). [CrossRef]
12. R. Heintzmann, “Structured illumination methods,” Handbook of Biological Confocal Microscopy: Third Edition pp. 265–279 (2006).
13. J. Pospíšil, T. Lukeš, J. Bendesky, K. Fliegel, K. Spendier, and G. M. Hagen, “Imaging tissues and cells beyond the diffraction limit with structured illumination microscopy and Bayesian image reconstruction,” GigaScience 8(1), 1–12 (2019). [CrossRef]
14. M. G. Gustafsson, L. Shao, P. M. Carlton, C. J. Wang, I. N. Golubovskaya, W. Z. Cande, D. A. Agard, and J. W. Sedat, “Three-Dimensional Resolution Doubling in Wide-Field Fluorescence Microscopy by Structured Illumination,” Biophys. J. 94(12), 4957–4970 (2008). [CrossRef]
15. F. Orieux, E. Sepulveda, V. Loriette, B. Dubertret, and J. C. Olivo-Marin, “Bayesian estimation for optimized structured illumination microscopy,” IEEE Trans. on Image Process. 21(2), 601–614 (2012). [CrossRef]
16. T. Azuma and T. Kei, “Super-resolution spinning-disk confocal microscopy using optical photon reassignment,” Opt. Express 23(11), 15003–15011 (2015). [CrossRef]
17. O. Schulz, C. Pieper, M. Clever, J. Pfaff, A. Ruhlandt, R. H. Kehlenbach, F. S. Wouters, J. Großhans, G. Bunt, and J. Enderlein, “Resolution doubling in fluorescence microscopy with confocal spinning-disk image scanning microscopy,” Proc. Natl. Acad. Sci. U. S. A. 110(52), 21000–21005 (2013). [CrossRef]
18. Y. Zhang, P. Asghari, D. R. L. Scriven, E. D. W. Moore, and K. C. Chou, “Structured illumination microscopy with a phase-modulated spinning disk for optical sectioning,” Opt. Lett. 48(15), 3933–3936 (2023). [CrossRef]
19. J. W. Goodman, Introduction to Fourier Optics in Electrical and Computer Engineering, vol. 8 (McGraw-Hill Series, 1996).
20. A. R. Halpern, M. Y. Lee, M. D. Howard, M. A. Woodworth, P. R. Nicovich, and J. C. Vaughan, “Versatile, do-it-yourself, low-cost spinning disk confocal microscope,” Biomed. Opt. Express 13(2), 1102–1120 (2022). [CrossRef]
21. J. B. Pawley, “Handbook of biological confocal microscopy: Third edition,” Handbook of Biological Confocal Microscopy: Third Edition pp. 1–985 (2006).
22. K. A. Johnson and G. M. Hagen, “Artifact-free whole-slide imaging with structured illumination microscopy and Bayesian image reconstruction,” GigaScience 9(4), giaa035 (2020). [CrossRef]
23. M. A. A. Neil, R. Juškaitis, and T. Wilson, “Method of obtaining optical sectioning by using structured light in a conventional microscope,” Opt. Lett. 22(24), 1905–1907 (1997). [CrossRef]
24. P. Křížek, T. Lukeš, M. Ovesný, K. Fliegel, and G. M. Hagen, “SIMToolbox: a MATLAB toolbox for structured illumination fluorescence microscopy,” Bioinformatics 32(2), 318–320 (2016). [CrossRef]
25. J. Pospíšil, K. Fliegel, and M. Klíma, “Assessing resolution in live cell structured illumination microscopy,” Proc. SPIE 10603, 106030D (2017). [CrossRef]
26. O. Mandula, M. Kielhorn, K. Wicker, G. Krampert, I. Kleppe, and R. Heintzmann, “Line scan - structured illumination microscopy super-resolution imaging in thick fluorescent samples,” Opt. Express 20(22), 24167 (2012). [CrossRef]
27. M. Ovesný, P. Křížek, J. Borkovec, Z. Švindrych, and G. M. Hagen, “ThunderSTORM: a comprehensive ImageJ plug-in for PALM and STORM data analysis and super-resolution imaging,” Bioinformatics 30(16), 2389–2390 (2014). [CrossRef]
