Open Access
Add to BibSonomyAbstract
Spinning-disk confocal microscopy is a proven technology for investigating 3D structures of biological specimens. Here we report a super-resolution method based on spinning-disk confocal microscopy that optically improves lateral resolution by a factor of 1.37 with a single exposure. Moreover, deconvolution yields twofold improvement over the diffraction limit. With the help of newly modified Nipkow disk which comprises pinholes and micro-lenses on the front and back respectively, emitted photons from specimen can be optically reassigned to the most probable locations they originate from. Consequently, the improvement in resolution is achieved preserving inherent sectioning capabilities of confocal microscopy. This extremely simple implementation will enable reliable observations at super high resolution in biomedical routine research.
© 2015 Optical Society of America
1. Introduction
Super-resolution fluorescence microscopy that allows the imaging of details beyond the diffraction limit has become an invaluable tool for biologists. Increased spatial resolution can be achieved with a variety of advanced methods such as stimulated emission depletion (STED) microscopy [1], structured illumination microscopy (SIM) [2] and single-molecule localization microscopy (SLM) including photoactivated localization microscopy (PALM) [3] and stochastic optical reconstruction microscopy (STORM) [4]. However, drawbacks of existing methods such as technical complexity, slow imaging speed and constraints of dye choice restrict the extensive use in routine research.
Confocal fluorescence microscopy, on the other hand, is widely used for its optical sectioning ability, simplicity of use and versatility despite the diffraction-limited resolution under practical conditions [5,6]. Besides, spinning-disk confocal microscopy has the advantages of fast imaging speed and low bleaching rate [5]. In addition, some spinning-disk based super-resolution implementations of SLM, SIM and Fresnel incoherent correlation holography (FINCH) fluorescence microscopy have been presented [7–9].
In a theoretical study in 1988, Sheppard [10] pointed out that confocal microscopy can surpass the diffraction limit without sacrificing signal-to-noise ratio. In this technique, referred to as image scanning microscopy (ISM) by Müller and Enderlein [11], the confocal pinhole is replaced by an imaging detector and each image from the detector is recorded as the illumination laser spot scans over the entire field. If we assume an imaging system with unity magnification for simplicity, the recorded image at scan position is given by
where is the position on the imaging detector, is the detection point spread function (PSF), is the illumination PSF, and is the specimen function describing the distribution of emitters. The illumination PSF, , has its peak at a position in the specimen plane, whereas the peak of the detection PSF, , is located at . Since the peak of the detection PSF is shifted by an amount of from the peak of illumination PSF, the effective PSF, , has a peak at the midpoint between the peaks of illumination and detection PSFs, namely, in the specimen plane if we neglect the Stokes shift so that . Therefore, the most probable origin of the recorded signal, , is not the detection position, but in the specimen plane. The idea of ISM is to shift the recorded signals, back to the most probable locations they originate from by an amount of , and then integrate over to construct a specimen image with enhanced resolution. This computational pixel reassignment scheme can be done by contracting the recorded images by a factor of two and then integrating contracted images at the center position to the resultant specimen image. The specimen image of ISM is given bySubstituting leadsTherefore, the PSF of ISM is given byEquation (4) shows that the PSF of ISM is equal to the original detection PSF convolved with the illumination PSF and then rescaled to half size. If we neglect the Stokes shift so thatand assume a Gaussian shaped PSF, where is standard deviation, the integral can be solved analytically:Therefore, we obtain a lateral resolution improvement by a factor of √2 what we would expect for confocal imaging with a closed pinhole. Moreover, Sheppard et al. [12] presented the effects of the size of the imaging detector of ISM which corresponds to pinhole size of confocal imaging. In case of infinitely small detector which is equivalent to confocal imaging with a closed pinhole, the lateral resolution improves by a factor of 1.39 over conventional microscope, but at the cost of losing light to form a visible image. However, the resolution of ISM slightly improves monotonically with the increase of the detector size in contrast to confocal imaging whose resolution deteriorates with the increase of pinhole size. In case of infinitely large detector, lateral resolution improves by a factor of 1.53, but at the cost of losing optical sectioning ability. Therefore, using 1 Airy unit diameter detector is good choice to retain optical sectioning while achieving substantial improvement of resolution by a factor of 1.49 over conventional widefield imaging. In this instance, since the emitted photons concentrate more precise locations they originate from, the peak intensity of PSF also improves by a factor of 1.63 over conventional widefield imaging. Hence, the resolution improvement is achieved without sacrificing optical sectioning ability and signal-to-noise ratio. Furthermore, Fourier transform of Eq. (4) yields optical transfer function (OTF) of ISMwhere is spatial frequency vector and a tilde denotes Fourier transform of the corresponding function. Equation (6) shows that the ISM image, , contains frequency component twice as high as conventional widefield image with an OTF, , but high frequency component is weaker than ideally resolution doubled image with an OTF, . Therefore, deconvolution can be applied to the ISM image so as to recover the image that has a full factor of 2 resolution improvement over the diffraction limit.While the first implementation of ISM [11] offered better lateral resolution in a simple manner, the considerable drawback was the slowness of the imaging speed. Recording images from the imaging detector at each scanning position limited the imaging speed too slow for investigation of living specimens. In recent years, various improvements of ISM have been realized. York et al. [13] demonstrated multifocal structured illumination microscopy (MSIM) which scanned with multiple excitation foci simultaneously using digital micro-mirror device. With the aid of parallelized scanning, MSIM achieved the imaging speed up to 1 Hz. Another implementation of multifocal excitation scheme based on a spinning-disk confocal microscope was also demonstrated by Schulz et al. [14].
Instead of computational pixel reassignment, equivalent methods that perform the reassignment by hardware have been proposed. Roth et al. [15] demonstrated optical photon reassignment microscopy (OPRA) that employed an intermediate beam expansion in the detection path to optically contract the focus on the image plane. Another implementation by Luca et al. [16] employed a “re-scanner” to change the scanning magnification maintaining the size of scanning spot. These methods have the advantage of reduced read-noise as well as fast imaging speed since super-resolution image is acquired in only one exposure frame.
The hardware approach can also be parallelized. York et al. [17] demonstrated instant structured illumination microscopy (ISIM) using a micro-lens array to generate a multifocal excitation, a matched pinhole array to reject out-of-focus emissions, another matched micro-lens array to optically contract each pinholed emission and a galvanometric mirror to scan the excitation and emission foci. Although ISIM offers excellent improvement in resolution and imaging speed preserving virtues of confocal fluorescence microscopy, the drawback is complexity of configuration. Considerable expertise is required to utilize the optical system adequately, because small misalignments of optical elements arranged separately cause substantial striping artifacts in the images as well as non-uniformity of illumination. Furthermore, a large number of optical elements in the detection path reduce signal levels.
Here, we present a simple super-resolution method based on spinning-disk confocal microscopy using multifocal excitation and optical photon reassignment schemes that optically improves lateral resolution by a factor of 1.37 with a single exposure. Moreover, deconvolution results twofold improvement over the diffraction limit. We term the method as SD-OPR. SD-OPR involves minimal modifications to an existing spinning-disk confocal microscope (Yokogawa®, CSU-X1).
2. Setup
A schematic diagram of the optical setup of SD-OPR is shown in Fig. 1. The upper disk of the scanning unit is a micro-lens array (MLA) disk which produces a multifocal excitation from collimated excitation laser (Coherent®, Sapphire LP 488nm). The lower disk is a micro-lens added pinhole array (MLPHA) disk on which pinholes and micro-lenses are regularly disposed on the upper and lower surfaces respectively. The individual micro-lens on the MLA and the MLPHA is located concentrically with the corresponding pinhole on the MLPHA. By rotating disks via a spindle motor, the multifocal excitation passing through the MLPHA scans specimen through a tube lens (Edmund Optics®, #49-287, f = 250 mm) and an objective lens. On the detection side, micro-lenses on the MLPHA locally contract each emission focus twofold (As shown in the inset of Fig. 1, each micro-lens on the lower surface of the MLPHA focuses the fluorescence emission from the specimen to a smaller pinhole. The details are described in the next paragraph.), and pinholes reject out-of-focus emissions to achieve resolution enhancement preserving confocal sectioning. The locally contracted multifocal emission reflected by a dichroic mirror (Semrock®, Di01-T405/488/561) is imaged by a camera through a 2 × magnifying relay optics (Yokogawa®, CSUX1PRT/P057) and an emission filter (Semrock®, FF03-525/50-25). We use a 100 × oil immersion objective lens (Olympus®, UPLSAPO100XO, NA 1.4), which translates to 278 × total magnification of the optical system. We also use a scientific CMOS camera (Andor®, Neo sCMOS®) with 6.5 μm pixels, which translates to 23.4 nm pixel size and 60 × 50 μm field of view in specimen space.
Fig. 1 Simplified schematic diagram of the optical setup. A micro-lens array (MLA) disk focuses the excitation light onto pinholes on a micro-lens added pinhole array (MLPHA) disk through a dichroic mirror (DM). The excitation light passing through the micro-lenses on the lower surface of the MLPHA is focused onto specimen by an objective lens. The emission collected by the same objective lens is contracted and pinholed by the MLPHA and imaged by the camera through an emission filter (EM). The scanning unit of SD-OPR is implemented on a CSU-X1 spinning-disk confocal microscope by substituting the internal spinning-disk assembly. The scanning unit is attached to an inverted microscope (Olympus®, IX71) using custom-made joint including a tube lens (f = 250 mm). In this arrangement, the original tube lens of the microscope is not on the optical path. The specimen is held on a motorized microscope stage (OptoSigma®, BIOS-225T). A piezo objective scanner (PI®, P-721.10) is used to perform axial scanning. The optical system is built on a vibration isolation table (HERZ, TDI-1585LAB). Although we didn’t use autofocus and environmental chamber in our setup, the use of them may be essential for thermal stability to perform long-term live-cell imaging.
To implement the optical photon reassignment scheme to the original scanning unit, we modified the Nipkow disk, namely, the MLPHA disk. The inset of Fig. 1 shows the function of the MLPHA disk with enlarged schematic diagram particularly showing a pair of the pinhole and the micro-lens on the MLPHA. The emission beam from the downside of the Fig. 1 focuses to the MLPHA at a cone angle of θ. While passing through the MLPHA, the beam is refracted by the micro-lens and the upper surface of the MLPHA, and diverges from the pinhole at a cone angle of 2θ. Since the size of focus is approximately inversely proportional to the cone angle if the angle is small, the micro-lens contracts the focus twofold while maintaining the orientation of the focus. This focus contraction leads to the displacement of emitted photons to the most probable locations they originate from.
On the other hand, the micro-lens on the MLPHA halves the cone angle of the excitation beam input from the upside of the figure. Consequently, the excitation light under-fills the objective pupil resulting in degradation in resolution. To overcome this degradation by over-illuminating the objective pupil, the diameter of the micro-lens on the MLA which is 250 μm for the normal CSU-X1, was increased to 580 μm. Moreover, the focal length of the tube lens which is normally 180 mm for Olympus® objective lenses, was also increased to 250 mm. In conjunction with the widening of the micro-lens on the MLA, the spacing between pinholes on the MLPHA was increased to 580 μm and the pinhole diameter was changed to 25 μm which approximately correspond to the first dark ring of the Airy disk of contracted emission focus.
3. Characterization of SD-OPR performance
To confirm the resolution improvement, subdiffractive fluorescent beads were imaged with the described SD-OPR setup. The excitation laser power on the specimen was 10 W/cm2 and the exposure time was 200 ms. Deconvolved images were obtained by applying 3D deconvolution (Wiener filter preconditioned Landweber algorithm) to the raw PSF images using Piotr Wendykier’s parallel iterative deconvolution plugin for ImageJ (available at http://sites.google.com/site/piotrwendykier/software/deconvolution/paralleliterativedeconvolution). We used the following parameters: Method, WPL; Boundary, Reflexive; Resizing, Auto; Output, Float; Precision, Double; Max number of iterations, 40. For comparison, a conventional widefield microscope (Olympus®, IX71) and a normal spinning-disk confocal microscope (Yokogawa®, CSU-X1) with the same objective lens were also evaluated as controls. The CSU-X1 was attached to the standard side port of IX71 using the original tube lens of the microscope. To determine the full width at half maximum (FWHM), 50 beads were individually fitted with a Gaussian function. The results are shown in Fig. 2 and Table 1. The first and the second image columns in Fig. 2 show X-Y and X-Z cross sectional views of point spread functions respectively. The scale bars represent 0.5 μm. The estimated mean values and standard deviations of lateral and axial FWHMs are shown in Table 1. As is clear from the results, the raw SD-OPR provided a 1.37-fold lateral resolution improvement over the raw widefield, while the resolution of the raw CSU-X1 spinning disk (SD) was still restricted by diffraction of light. The deviation between theoretical lateral resolution improvement of 1.49 and the results of actual measurement is probably due to residual optical aberration and Stokes shift [12]. Deconvolution improved the lateral resolution approximately 1.53-fold over the corresponding raw images, and consequently the deconvolved SD-OPR provided a 2.11-fold lateral resolution improvement over the raw widefield. We also found the axial resolution improvement over raw widefield by a factor of 1.26 for the raw SD-OPR and 1.92 for the deconvolved SD-OPR. These improvements by the deconvolution seem better than expected, and they may be due to over-enhancement of the deconvolution since the fluorescent beads are considerably simpler than practical specimens. Therefore, similar resolution evaluation of biological specimen with complex structure is essential to justify the resolution improvement further.
Fig. 2 Comparison of point spread functions. Subdiffractive fluorescent beads (Thermo Fisher Scientific®, FluoSpheres® F8803, 0.1 μm diameter, yellow-green stained) were imaged with widefield, normal Yokogawa® CSU-X1 spinning-disk (SD) and SD-OPR. Deconvolved images were obtained by applying 3D deconvolution (Wiener filter preconditioned Landweber algorithm) using Piotr Wendykier’s parallel iterative deconvolution plugin for ImageJ. The first and the second image columns show X-Y and X-Z cross sectional views of PSFs respectively. The scale bars represent 0.5 μm.
Table 1. FWHM measurements of point spread functions.
To investigate the resolution improvement in biological specimen, we imaged microtubules in fixed bovine pulmonary artery endothelial cells (Thermo Fisher Scientific®, FluoCells® prepared slide #2 F14781). The microtubules are labeled with mouse monoclonal anti-α-tubulin antibodies in conjunction with goat anti-mouse IgG conjugated with green-fluorescent BODIPY® FL. The excitation laser power on the specimen was 10 W/cm2 and the exposure time was 500 ms. Deconvolved images were obtained using Huygens® software (Scientific Volume Imaging). We used the following parameters: Method, Classic maximum likelihood estimation; Max number of iterations, 100; Signal to noise ratio, 60; Quality threshold, 0.0001; Iteration mode, Optimized; Brick layout, Auto. The results are shown in Fig. 3. Figures 3(a)–3(d) represent images of raw widefield, deconvolved widefield, raw SD-OPR and deconvolved SD-OPR respectively. The scale bars in Figs. 3(a)–3(d) represent 10 μm. Figures 3(e)–3(h) show magnified images of a smaller area corresponding to the rectangles in the Figs. 3(a)–3(d) respectively. The scale bars in Figs. 3(e)–3(h) indicate 1 μm. As shown in these figures, SD-OPR obtained improved resolution in the biological specimen as well. Although the deconvolved widefield image also seems to achieve resolution enhancement, it could not extract the actual structures which could be seen in the raw SD-OPR image as shown in Figs. 3(f), 3(g) and 3(i). A pair of microtubules corresponding to the yellow lines in the Figs. 3(e)–3(h) was not separated in the widefield images regardless of deconvolution as shown in Fig. 3(i). However, the microtubules were clearly separated in the deconvolved SD-OPR image as shown in Figs. 3(h) and 3(i). Figure 3(j) indicates the minimum separation distance of crossing microtubules in the deconvolved SD-OPR image. The sectional plot shown in Fig. 3(j) corresponds to the red line in the Fig. 3(h). The minimum separation distance calculated by fittings of two Gaussian functions was 113 nm. These results indicate almost twofold resolution improvement over the diffraction limit was achieved in the biological specimen.
Fig. 3 Imaging of microtubules in fixed bovine pulmonary artery endothelial cells. (a) Raw widefield image. (b) Deconvolved widefield image. (c) Raw SD-OPR image. (d) Deconvolved SD-OPR image. The scale bars in the panels a, b, c and d represent 10 μm. Deconvolved images were obtained by applying 3D deconvolution using Huygens® (Scientific Volume Imaging). (e–h) Magnified views of smaller areas corresponding to the rectangles in the panels a, b, c and d. The scale bars in the panels e, f, g and h represent 1 μm. (i) Sectional plots of pixel intensities across the yellow lines in the panels e, f, g and h. (j) Sectional plot of pixel intensities across the red line in the panel h. Fittings of two Gaussian functions are also shown with broken lines, and the distance between the peaks of Gaussian functions is 113 nm.
Finally, we evaluated chromatic aberration of the optical system. Since the micro-lenses on the MLA and MLPHA are not achromatic, chromatic aberration of the optical system would be a matter of concern. Because the micro-lenses affect axial chromatic aberration rather than lateral, the major concern is co-registration of dyes of different excitation and emission wavelengths in axial direction. Thus, we evaluated the axial chromatic shift of the optical system by an optical simulation. Figure 4 shows the axial shift of excitation focus in the specimen. As shown in the result, the axial focal shift between violet and red was 75 nm, and was sufficiently smaller than axial resolution of SD-OPR.
Fig. 4 Simulation of excitation focal shift vs. wavelength. Wavelength characteristic of axial shift of excitation focus in the specimen was calculated using Zemax® (Zemax LLC) optical simulation software. To calculate the focal shift caused only by the micro-lenses, aberration-free microscope model was used in the simulation. The axial shift of excitation focus was evaluated in the wavelength range of 405 to 640 nm.
4. Conclusions
We demonstrated that a simple spinning-disk based super-resolution method, SD-OPR, dramatically improves resolution with a single exposure. The abilities to observe biological specimens with ease and no limitation of dye are significant practical advantages inherited from confocal fluorescence microscopy. Another advantage of SD-OPR is straightforwardness of implementation. The configuration that involves dual Nipkow disks has been used in commercial spinning-disk confocal microscope for over a decade and has proven to bear with routine research. Therefore, we anticipate the method to be capable of reliable super high resolution observations in biomedical routine research. Furthermore, multi-photon excitation technique will be able to be applied to SD-OPR. Application of multi-photon excitation to SD-OPR will improve penetration depth of super-resolution imaging in thick scattering specimens [18–20].
Acknowledgments
We would like to thank H. Hirukawa and H. Sangu for helpful discussions. We also would like to thank T. Iwahara for feedbacks on the manuscript. This work has no associated award funding. T. A. and T. K. are employees of Yokogawa Electric Corporation, which has patents pending regarding this method.
References and links
1. S. W. Hell and J. Wichmann, “Breaking the diffraction resolution limit by stimulated emission: stimulated-emission-depletion fluorescence microscopy,” Opt. Lett. 19(11), 780–782 (1994). [CrossRef] [PubMed]
2. M. G. L. Gustafsson, “Surpassing the lateral resolution limit by a factor of two using structured illumination microscopy,” J. Microsc. 198(2), 82–87 (2000). [CrossRef] [PubMed]
3. E. Betzig, G. H. Patterson, R. Sougrat, O. W. Lindwasser, S. Olenych, J. S. Bonifacino, M. W. Davidson, J. Lippincott-Schwartz, and H. F. Hess, “Imaging intracellular fluorescent proteins at nanometer resolution,” Science 313(5793), 1642–1645 (2006). [CrossRef] [PubMed]
4. M. J. Rust, M. Bates, and X. Zhuang, “Sub-diffraction-limit imaging by stochastic optical reconstruction microscopy (STORM),” Nat. Methods 3(10), 793–796 (2006). [CrossRef] [PubMed]
5. J. B. Pawley, Handbook of Biological Confocal Microscopy, 3rd ed. (Plenum, 2006).
6. G. Cox and C. J. R. Sheppard, “Practical limits of resolution in confocal and non-linear microscopy,” Microsc. Res. Tech. 63(1), 18–22 (2004). [CrossRef] [PubMed]
7. N. A. Hosny, M. Song, J. T. Connelly, S. Ameer-Beg, M. M. Knight, and A. P. Wheeler, “Super-resolution imaging strategies for cell biologists using a spinning disk microscope,” PLoS ONE 8(10), e74604 (2013). [CrossRef] [PubMed]
8. N. Siegel and G. Brooker, “Improved axial resolution of FINCH fluorescence microscopy when combined with spinning disk confocal microscopy,” Opt. Express 22(19), 22298–22307 (2014). [CrossRef] [PubMed]
9. S. Hayashi and Y. Okada, “Ultrafast superresolution fluorescence imaging with spinning disk confocal microscope optics,” Mol. Biol. Cell 26(9), 1743–1751 (2015). [CrossRef] [PubMed]
10. C. J. R. Sheppard, “Super-resolution in confocal imaging,” Optik (Stuttg.) 80(2), 53–54 (1988).
11. C. B. Müller and J. Enderlein, “Image scanning microscopy,” Phys. Rev. Lett. 104(19), 198101 (2010). [CrossRef] [PubMed]
12. C. J. R. Sheppard, S. B. Mehta, and R. Heintzmann, “Superresolution by image scanning microscopy using pixel reassignment,” Opt. Lett. 38(15), 2889–2892 (2013). [CrossRef] [PubMed]
13. A. G. York, S. H. Parekh, D. D. Nogare, R. S. Fischer, K. Temprine, M. Mione, A. B. Chitnis, C. A. Combs, and H. Shroff, “Resolution doubling in live, multicellular organisms via multifocal structured illumination microscopy,” Nat. Methods 9(7), 749–754 (2012). [CrossRef] [PubMed]
14. 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] [PubMed]
15. S. Roth, C. J. R. Sheppard, K. Wicker, and R. Heintzmann, “Optical photon reassignment microscopy (OPRA),” Opt. Nanosc. 2(1), 5 (2013). [CrossRef]
16. G. M. De Luca, R. M. Breedijk, R. A. Brandt, C. H. Zeelenberg, B. E. de Jong, W. Timmermans, L. N. Azar, R. A. Hoebe, S. Stallinga, and E. M. Manders, “Re-scan confocal microscopy: scanning twice for better resolution,” Biomed. Opt. Express 4(11), 2644–2656 (2013). [CrossRef] [PubMed]
17. A. G. York, P. Chandris, D. D. Nogare, J. Head, P. Wawrzusin, R. S. Fischer, A. Chitnis, and H. Shroff, “Instant super-resolution imaging in live cells and embryos via analog image processing,” Nat. Methods 10(11), 1122–1126 (2013). [CrossRef] [PubMed]
18. T. Shimozawa, K. Yamagata, T. Kondo, S. Hayashi, A. Shitamukai, D. Konno, F. Matsuzaki, J. Takayama, S. Onami, H. Nakayama, Y. Kosugi, T. M. Watanabe, K. Fujita, and Y. Mimori-Kiyosue, “Improving spinning disk confocal microscopy by preventing pinhole cross-talk for intravital imaging,” Proc. Natl. Acad. Sci. U.S.A. 110(9), 3399–3404 (2013). [CrossRef] [PubMed]
19. M. Ingaramo, A. G. York, P. Wawrzusin, O. Milberg, A. Hong, R. Weigert, H. Shroff, and G. H. Patterson, “Two-photon excitation improves multifocal structured illumination microscopy in thick scattering tissue,” Proc. Natl. Acad. Sci. U.S.A. 111(14), 5254–5259 (2014). [CrossRef] [PubMed]
20. P. W. Winter, A. G. York, D. D. Nogare, M. Ingaramo, R. Christensen, A. Chitnis, G. H. Patterson, and H. Shroff, “Two-photon instant structured illumination microscopy improves the depth penetration of super-resolution imaging in thick scattering samples,” Optica 1(3), 181–191 (2014). [CrossRef] [PubMed]
