Downloaded from http://cshprotocols.cshlp.org/ at TULANE UNIV on September 8, 2014 - Published by Cold Spring Harbor Laboratory Press
Grating Imager Systems for Fluorescence Optical-Sectioning Microscopy Frederick Lanni Cold Spring Harb Protoc; doi: 10.1101/pdb.top083493 Email Alerting Service Subject Categories
Receive free email alerts when new articles cite this article - click here. Browse articles on similar topics from Cold Spring Harbor Protocols. Confocal Microscopy (92 articles) Fluorescence (416 articles) Fluorescence, general (284 articles) Imaging/Microscopy, general (542 articles)
To subscribe to Cold Spring Harbor Protocols go to:
http://cshprotocols.cshlp.org/subscriptions
© 2014 Cold Spring Harbor Laboratory Press
Downloaded from http://cshprotocols.cshlp.org/ at TULANE UNIV on September 8, 2014 - Published by Cold Spring Harbor Laboratory Press
Topic Introduction
Grating Imager Systems for Fluorescence Optical-Sectioning Microscopy Frederick Lanni
In fluorescence microscopy, optical sectioning is defined as the attenuation or removal of out-of-focus features from an image, and it is a prerequisite for quantitative analysis of three-dimensional structure or function within the specimen. Optical sectioning is most commonly performed by confocal scanning fluorescence microscopy or two-photon scanning fluorescence microscopy. However, structured illumination can be used in conventional fluorescence microscopes to obtain optical sectioning performance, and, in advanced systems, 3D superresolution. The simplest structured-illumination system uses a Ronchi grating as a mask to project parallel stripes within the sharp depth-of-focus of the objective to encode in-focus specimen features differently from out-of-focus features. By shifting the grating, the in-focus image component can be discriminated and separated by elementary image processing operations. This implementation of structured illumination, the fluorescence grating imager, uses a conventional light source, is compatible with all high-quality fluorescence filter sets, and provides high optical-sectioning performance when used to image specimens in which (1) the outof-focus image component is not much brighter than the in-focus features and (2) there is no significant movement in the specimen during the grating shift and image capture process.
OPTICAL SECTIONING
A single optical section is a view of a slice within the specimen centered axially on the plane of focus. A through-focus stack of optical sections therefore constitutes a 3D representation of the specimen. For the past 25 yr, optical sectioning has been driven by the development of the laser confocal scanning fluorescence microscope (CSFM), the two-photon scanning fluorescence microscope, and deconvolution algorithms. In 1997, Wilson and colleagues described both a principle and an optical system to achieve nearly direct optical sectioning in a conventionally illuminated reflectance microscope (Wilson et al. 1997, 1998a,b,c; Neil et al. 1997; Juskaitis et al. 1998). The essential idea was the use of “structured light” or striped illumination projected with sharply defined depth of focus to achieve axial selectivity. The extension to fluorescence was straightforward (Neil et al. 1998, 2000; Wilson et al. 1998b; Ben-Levy and Peleg 1999; Lanni and Wilson 2000; Lagerholm et al. 2003; Vanni et al. 2003), and there are now instruments available from Carl Zeiss (ApoTome; www.zeiss.com/) and from Qioptiq LINOS (OptiGrid; www.qioptiqlinos.com/) that incorporate the concept. In addition, much more complex coherent structured-illumination microscopy systems for optical sectioning and superresolution have been developed which are beyond the scope of this introduction (Lanni 1986; Lanni et al. 1986; Bailey et al. 1993; Frohn et al. 2000; Gustafsson 2000; Gustafsson et al. 2008; Shao et al. 2008; Schermelleh et al. 2008). For the purposes of the present discussion, we refer to the basic instrument as a fluorescence grating imager (FGI). The attractive characteristics of the FGI are its versatility and performance, given its simplicity and relatively low cost. Adapted from Imaging: A Laboratory Manual (ed. Yuste). CSHL Press, Cold Spring Harbor, NY, USA, 2011. © 2014 Cold Spring Harbor Laboratory Press Cite this introduction as Cold Spring Harb Protoc; doi:10.1101/pdb.top083493
923
Downloaded from http://cshprotocols.cshlp.org/ at TULANE UNIV on September 8, 2014 - Published by Cold Spring Harbor Laboratory Press
F. Lanni
OPERATING PRINCIPLE AND INSTRUMENTATION
In a grating imager, only a single essential modification is made to a conventional research-grade fluorescence microscope. A movable Ronchi grating mask is placed in the field iris plane of the incident-light illuminator, thus producing in the specimen striped illumination that is coincident with the geometric focus plane and within the depth of focus of the objective (Fig. 1). The actual focus range of the stripes can be made very sharp if a fine grating is used with an objective of high numerical aperture (NA). Therefore, the image formed by the microscope will consist of striped in-focus features superposed with uniformly illuminated out-of-focus features (background). The period of the projected stripes is the actual grating period divided by the demagnification ratio between the illuminator field iris plane and the specimen, which is objective dependent (and generally not equal to the specified objective magnification). When the grating is shifted so as to move the projected stripes transversely across the object, the fluorescence detected in any pixel element in the image will consist of (1) a steady or DC component owing to the out-of-focus background, plus (2) an oscillating (AC) component owing to the shifting of the stripes across in-focus structures, and (3) shot noise owing to both the DC and AC signal components. In brief, the amplitude of the AC component is the in-focus signal level for that pixel. By shifting the grating between a minimum of three defined positions to obtain three images, a very simple digital image processing operation can subtract away the background and demodulate the AC component to produce an optical section.
A
B
Cooled CCD camera
Field iris Field plane Grating shift
EM
DR Adjustment of grating focus
Specimen
EX
Lamp
Ronchi grating
FIGURE 1. Schematic of fluorescence grating imager. (A) Optics layout showing movable grating mask inserted at location of illuminator field iris. Ronchi gratings (equal-width square-wave pattern) in the range of 8–50 line pairs per mm (LP/mm) are used, depending on the demagnification ratio determined by the objective and the focal length of the illuminator tube lens. A means for bringing the grating to a sharp focus in the plane of focus of the objective is necessary, here shown as an axial adjustment of the illuminator tube lens. To minimize aberration in the grating image, the excitation filter (EX) is shown, removed from its usual position to precede the grating. This is conveniently performed by the use of an external filter wheel between lamp collector lens and microscope stand. Additionally, the dichroic reflector (DR) must be flat and strain-free. Actuators for grating movement can be piezoelectric or electromagnetic, or the grating may be fixed but refractively shifted by a separate optic. Shift increments are generally one-third or one-quarter of the grating period, so they will be in the range 5.00–42.0 µm. Computer control synchronizes discrete grating movement and CCD (charge-coupled device) readout between periods of light exposure. A fast light shutter (not shown) is essential for precision exposure timing and minimization of unnecessary exposure. EM, emission filter. (B) Schematic showing finite grating depth-of-focus within a 3D specimen. Geometrically, it can be seen that both grating spatial frequency and objective NA will affect the depth-of-focus of the pattern.
924
Cite this introduction as Cold Spring Harb Protoc; doi:10.1101/pdb.top083493
Downloaded from http://cshprotocols.cshlp.org/ at TULANE UNIV on September 8, 2014 - Published by Cold Spring Harbor Laboratory Press
Grating Imager Systems
Microscope optical system quality has a definite effect on FGI performance. First, proper immersion is essential to avoid spherical aberration in both the projected grating and the fluorescence image. Indirect or direct water-immersion must be used when optically sectioning living cells, or other lowindex specimens. Oil immersion must be used for specimens mounted in high-index medium. Second, the illuminator optics must provide diffraction-limited imaging of the field iris plane onto the specimen, so that the grating is projected at the full resolution of the objective. Although excitation filters are generally of excellent quality in terms of pass band and blocking, they may not be optically flat, thus introducing aberration into the projected grating pattern. Mounting the excitation filter between the lamp and the grating mask (as shown in Fig. 1) obviates this problem. For the same reason, the DR must be strain-free and flat. Third, the full illumination NA (INA) of the incident light optics must be used to minimize the depth of focus of the projected grating. Fourth, if the instrument is used over a significant range in the UV–visible spectrum (e.g., with DAPI and Cy3 dyes), chromatic shift of focus could necessitate axial adjustment of the grating when filter sets are interchanged. This is analogous to the differential adjustment of pinhole focus in a confocal scanner, and is minimized by the use of apochromatic optics. In the ApoTome and Optigrid, this correction can be set and automated. Lastly, the grating shift mechanism must be precise and accurate, and must hold the mask motionless during each image exposure. COMPUTATION OF THE OPTICAL SECTION
The simplest projected grating image is a sine wave pattern, which, when translated across the specimen, produces a sinusoidal modulation of the fluorescence. Translation of the grating is denoted by the phase shift (φ), where, for example, a shift of 90˚ is equivalent to movement equal to one-quarter of the grating period. In any pixel, the fluorescence signal can be expressed as a periodic function of φ, plus a constant term due to out-of-focus background: f (f) = DC term + AC term = a0 /2 + (a1 cos f + b1 sin f).
(1)
The in-focus part of the total fluorescence signal in that pixel is the AC amplitude, which is given by 2 the Pythagorean sum of the sine and cosine coefficients: a1 + b21 . It can be shown generally that the AC amplitude can be computed on a pixel-by-pixel basis from three images made with the grating shifted between three distinct positions. Three images are necessary because there are three unknown coefficients in Equation 1. The most commonly used shift sequences are given in Table 1, along with corresponding formulas for computing the optical section. The out-of-focus image component, which is the same in all images in the set, is removed by computing differences between image pairs (Fig. 2). The resulting bipolar, striped images are then demodulated by algebraic operations that effectively form the Pythagorean sum. A real projected grating will include spatial harmonics. For a perfect Ronchi (square wave) grating, only odd harmonics occur: f (φ) = DC + (a1 cos φ + b1 sin φ) + (a3 cos 3φ + b3 sin 3φ) + (a5 cos 5φ + b5 sin 5φ) + ··· . The amplitude of the AC term is still the in-focus part of the total fluorescence signal in that pixel. However, now it can be seen that the simple image processing formulas will not give 2 a1 + b21 exactly. The harmonics occur as error terms, and may appear as fine stripes in the optical sections. Two factors can minimize or eliminate this problem. (1) The shift sequence and TABLE 1. FGI image processing formulas Grating shift sequence
Computation of optical section
One-third period, three images 0˚–120˚–240˚ One-quarter period, four images 0˚–90˚–180˚–270˚ One-quarter period, three images 0˚–90˚–180˚
(2 /3)[(i0 − i120) + (i120 − i240) + (i240 − i0) ] (1/2)[(i0 − i180)2 + (i90 − i270)2]1/2 (2−1/2)[(i0 – i90)2 + (i90 – i180)2]1/2
Cite this introduction as Cold Spring Harb Protoc; doi:10.1101/pdb.top083493
1/2
2
2
Uniform exposure 2 1/2
Yes Yes No
925
Downloaded from http://cshprotocols.cshlp.org/ at TULANE UNIV on September 8, 2014 - Published by Cold Spring Harbor Laboratory Press
F. Lanni
FIGURE 2. Optical section computation from image data showing the actin cytoskeleton in a 3T3 fibroblast. Cells were grown on a cover glass under standard incubator conditions, fixed, permeabilized, and stained with rhodaminephalloidin to show F-actin. (A,B,C) i0, i90, and i180 with focus set close to the adherent basal region of the cell. (D, E) difference images (A–B) and (B–C) as in Table 1, formula 3. In D and E, the gray scale is bipolar with 0 at midrange gray. (F ) Optical section computed by Pythagorean summation of D and E. Field of view, 39 µm.
algorithm used can suppress one or more harmonics. As originally pointed out by Wilson et al. (1997), the one-third-period shift sequence exactly compensates the third harmonic, therefore the first error term is due to the fifth harmonic. Because, for a Ronchi grating, the amplitudes drop off as 1/n, this error is small. (2) Because the incoherent modulation transfer function of the microscope decreases with spatial frequency (to zero at the inverse of Abbe’s resolution limit), the microscope attenuates the higher harmonics that are projected into the specimen—further reducing the error terms. In principle, harmonic error can be eliminated altogether by choosing the Ronchi so that its fifth harmonic matches or exceeds Abbe’s resolution limit for the objective. This is not a severe restriction. With the period of the fifth harmonic set equal to Abbe’s resolution limit (λ/2NA), the fundamental period will equal 5 × (l/2NA). This is 2.5× more coarse than the optimal period (λ/NA; see Equation 3 below), but it is projected with high contrast. Alternatively, the grating could be chosen so that the third harmonic was equal to Abbe’s limit, in which case the fundamental period would be 3 × (λ/ 2NA). This is close to the optimal period (λ/NA), but it will have lower contrast than with fifthharmonic suppression. In general, a wide variety of grating shift sequences can be formulated, all of which, in principle, provide the basis for computation of an optical section. Sequences of more than three images per focus plane can be used to compute the contribution of higher harmonics, or, for example, to eliminate second-harmonic error. However, this is a trade-off in terms of speed and light exposure. The particular shift sequence used, the number of images recorded, and the processing formula all affect various performance characteristics of the instrument: For example, as noted, the 0˚– 120˚–240˚ shift sequence eliminates third-harmonic error. It also uniformly exposes the plane of focus, which minimizes patterned photobleaching (Table 1). In general, light exposure of the specimen in the grating imager is comparable with conventional fluorescence microscopy. Because the Ronchi passes 50% of the incident light, the three exposures required for one optical section deliver 1.5× the light dose needed for a conventional image in which the in-focus features are equally bright (but superposed with out-of-focus features). 926
Cite this introduction as Cold Spring Harb Protoc; doi:10.1101/pdb.top083493
Downloaded from http://cshprotocols.cshlp.org/ at TULANE UNIV on September 8, 2014 - Published by Cold Spring Harbor Laboratory Press
Grating Imager Systems
SHARPNESS OF THE OPTICAL SECTION
Based on an analysis of the optical transfer function of the FGI (Lagerholm et al. 2003), it can be shown that the optical section thickness, δ, is related to the NA and projected grating period, L, by the formula d = (l/2)/{[n2 − (NA − l/L)2
1/2 2 − n − NA2 ]1/2 }
for
L . l/2NA.
(2)
The combination of a high-NA objective and a fine grating can produce sharp sections that, in principle, match the optical-sectioning performance of a confocal scanner. Ideally, the sharpest sectioning is obtained when the projected grating period is twice Abbe’s resolution limit: dmin = (l/2)/{n − [n2 − NA2 ]1/2 }
for L = l/NA.
(3)
In practice, a twofold to fourfold coarser grating generally gives better performance owing to increased stripe contrast. As in any form of fluorescence microscopy, FGI signal per pixel is reduced as sectioning is sharpened. SIGNAL-TO-NOISE RATIO
Relative to a confocal scanner, performance of the FGI is limited ultimately by photon counting noise due to out-of-focus (background) features in the object. Unlike a confocal scanner, in which the pinhole blocks most background light before detection, background light in an FGI is detected in each image, and then removed by digital subtraction (Table 1 formulas). Both signal (S) and background (B) photocount levels can be treated as Poisson-distributed variables, with means NS and NB. The subtraction steps remove the background mean, NB, √ but leave a residual noise with a level roughly equal to the Poisson root-mean-square (RMS) value, example, if the background amounts √ B . For √N to 104 counts, and the in-focus signal is 103, then NB = 104 = 100, therefore a signal-to-noise ratio (SNR) of 103/100 = 10 is expected. If the in-focus signal were 100 in a background of 104, it could barely be distinguished from the noise. A more detailed analysis shows the RMS additivity of noise in the image processing steps leading to the optical section. The intrinsic FGI noise level can be estimated from the ideal case of removal of a perfectly defocused background in a “0˚–120˚–240˚” image set with S = 0 (blank specimen). By definition, N√ B is the same for all three images, and the RMS noise level in the resulting blank optical section is 1.155 NB , only 15% greater than the noise level in a single image. Therefore, a√ simple SNR the RMS sum of the Poisson of S to√ √ratio approximation is the noise levels in the signal S + 2B and the background B :SNR S/ S + 2B. SAMPLING IN THE IMAGE PLANE
What sampling is required for a camera to pick up all of the detail in the optical image field that is formed by a microscope? This is set by Nyquist’s sampling theorem, when noise is not a limiting factor: The density of sample points must exceed the inverse of the signal bandwidth. In the case of a microscope, Abbe’s resolution limit, λ/2NA, is the period of the finest grating that can be resolved, and sets the signal bandwidth. The spatial frequency of that grating is 2NA/λ (cycles/µm), and the bandwidth is defined as the frequency interval from −2NA/λ to 2NA/λ, or 4NA/λ. Therefore, the Nyquist sampling interval in nonmagnified coordinates is 1/bandwidth or λ/4NA; that is, the camera pixel spacing divided by total magnification must be ≤λ/4NA (Lanni and Baxter 1992). For typical high-NA microscopes, λ/4NA is generally close to 0.1 µm. When using a camera having, for example, 6.45-µm CCD elements, a total magnification ≥64.5× is required to achieve Nyquist sampling. In real cameras, finite pixel size also causes spatial averaging of the fine details in the image field Cite this introduction as Cold Spring Harb Protoc; doi:10.1101/pdb.top083493
927
Downloaded from http://cshprotocols.cshlp.org/ at TULANE UNIV on September 8, 2014 - Published by Cold Spring Harbor Laboratory Press
F. Lanni
(Lanni and Baxter 1992). The effect of finite CCD pixel size in the grating imager is a reduction in the percent modulation in an image of the grating projected on a uniformly fluorescent specimen. Clearly, if the grating period matched the pixel size, no grating would be seen. A graph of percent modulation versus number of camera pixels across one grating line pair (Fig. 3) shows that modulation is preserved to a high degree when at least 10 pixels span a projected line pair. As an example, a grating chosen to set L = 5λ/2NA (see Computation of the Optical Section) would also make L/10 = λ/ 4NA, the Nyquist interval. PERFORMANCE
Grating modulation (imaged)
A very basic measure of performance for any optical-sectioning fluorescence microscope is its axial response, defined as the graph of pixel brightness versus focus increment for a uniform planar specimen. In practice, a stable fluorescent film that is thin compared with the smaller of (1) the wave-optical depth of field (2nl/NA2) or (2) the optical-sectioning limit must be used to obtain a meaningful measurement. Figure 4A shows graphically the axial response of a grating imager based on a Zeiss Axiovert 200 M operated with a 1.30-NA objective (Zeiss 100× Plan-Neofluar) and 1.33-µm projected grating period, as measured using a 0.05-µm rhodamine-labeled thin film specimen. The optical section thickness, defined by the graphical full width at half-maximum (FWHM), was 0.65 µm. In comparison, Equation 2 gives a value δ = 0.623 µm for NA = 1.3, L = 1.33 µm, and λ = 570 nm (strictly speaking, Equation 2 should distinguish between the excitation and emission wavelengths, but these are usually close; in this example they are 540 and 605 nm, respectively). The experimental FWHM is comparable with a confocal scanner operated with the same objective. Furthermore, the FGI axial response could be further sharpened by use of a finer grating because, in this example, L exceeded λ/NA by a factor of 3 (see Equation 3). In Figure 4B showing the perinuclear actin cytoskeleton of a fibroblast, a focus increment of 0.6 µm (encoded in magenta/green pseudocolor) clearly discriminates axial structure in the FGI optical sections. In cell biological applications, where out-of-focus image features are generally less than an order of magnitude greater in brightness than in-focus features, grating imagers generally produce fluorescence optical sections comparable with high-quality confocal scans. Figure 5 shows the quantitative effect of optical sectioning in a specimen of chick embryo fibroblasts grown in a 3D collagen gel, then fixed and stained for F-actin with rhodamine phalloidin. The left panel in the figure shows a fluorescence image in the absence of optical sectioning, and the right panel shows the FGI optical section at the same plane of focus. A pixel-row profile across the images between the guidelines shows the large effect that out-of focus light would have on quantification of actin in the cytoskeleton, and the degree of improvement obtained with the grating imager. As in any form of high-resolution microscopy, a number of use-related factors have been identified that have a significant effect on performance. Table 2 lists those known to date. These can be classified generally as (1) reduction in aberration and grating defocus (which reduce projected grating contrast and image contrast), (2) optimization of the light source and minimization of light exposure, and (3) optimization of sampling. In comparison to confocal scanning, the FGI has a number of advantages and 1 0.8 0.6 0.4 0.2
5
10
15
N (#pixels/cycle)
928
20
25
FIGURE 3. Effect of finite camera CCD element size on image contrast. Graph shows the contrast of a projected sine wave grating versus number of contiguous pixels per cycle (N). The plotted function is |sin(π/N )/(π/N)|. As N is increased to more than 10 pixels per cycle, the contrast of the projected grating approaches 1 (100%).
Cite this introduction as Cold Spring Harb Protoc; doi:10.1101/pdb.top083493
Downloaded from http://cshprotocols.cshlp.org/ at TULANE UNIV on September 8, 2014 - Published by Cold Spring Harbor Laboratory Press
Grating Imager Systems
A 800 700 600
FGI OS ROI
500
400 300
200 100 0 –1.0
–0.8
–0.6
0.0 –0.4 –0.2 Focus increment (μm)
0.2
0.4
0.6
B
FIGURE 4. (A) Grating imager axial response. The graph shows optical section signal level versus focus drive movement in a small region-of-interest in the center of an image of a thin film. For the test reported here, poly-[methyl methacrylate] covalently labeled with rhodamine (Rh-PMMA) was spin-cast from chlorobenzene onto a coverglass to form a thin film. A simple reflectance interferometer was used to determine the film thickness (50 nm) based on fringe shift and a refractive index equal to 1.49 for PMMA. The coverglass was then mounted with the film in contact with optical cement (Norland Optical Adhesive #60, index = 1.56) on a slide, and UV light was used to polymerize the cement. It was previously determined that Rh-PMMA is insoluble in NOA#60. The advantage of this type of standard specimen is its simple refractive structure and similar refractive indexes (glass 1.52, PMMA 1.49, NOA 1.56). The FWHM defines the sharpness of the optical section, here 0.65 µm. Instrument: Zeiss Axiovert 200 M with Queensgate piezoelectric grating shifter, 100× 1.30 NA oil-immersion objective, 50 LP/mm Ronchi grating projected to 1.33-µm period, Hamamatsu Orca-ER CCD camera with 6.45-µm pixels, and QED imaging in vivo instrument control software. (B) Pseudocolor overlay showing F-actin structures in the perinuclear region of a fibroblast. Field-of-view width, 10.0 µm. The basal plane of focus (magenta) shows small actin fibers below the nucleus. A 0.6-µm focus increment brings stress fibers into view (green). In the FGI optical sections (right), the structures are readily discriminated. Overlay of the corresponding pseudocolor conventional images (left) shows that most structures are not discriminated by focus alone.
disadvantages, which are listed in Table 3. An outstanding advantage is that the instrument is immediately compatible with almost any standard high-quality fluorescence filter set. The only wavelength range limitation so far encountered is longitudinal chromatic aberration when using UV excitation. Finally, although the grating imager has been developed as a fast and compact means to obtain optical sections, it is possible to apply 3D deconvolution methods to the unprocessed serial-focus image sets. Although this is computationally much more demanding, such processing can be performed offline while still obtaining fast optical sections using one of the image-processing formulas in Table 1.
FUTURE OUTLOOKS
In the past several years, a number of multiplex optical sectioning systems have come into general use in addition to FGIs. These would include the Yokogawa-type pinhole array scanning confocal Cite this introduction as Cold Spring Harb Protoc; doi:10.1101/pdb.top083493
929
Downloaded from http://cshprotocols.cshlp.org/ at TULANE UNIV on September 8, 2014 - Published by Cold Spring Harbor Laboratory Press
F. Lanni
FIGURE 5. Quantification of actin cytoskeletal structures in a grating imager optical section (upper right) versus the corresponding conventional image (upper left). (Lower panels) Pixel-row profile between guidelines shows fluorescence from out-of-focus features mixed with in-focus features and how attenuation of the background in the optical section makes quantification possible. Gray scale in optical section profile (lower right) is boosted threefold. The field of view (width and height) of each square image is 165 µm.
microscope and various slit-scan confocal systems. Most recently, a pinhole aperture-correlation confocal scanner (Juskaitis et al. 1996; Wilson et al. 1996) with highly efficient use of the light source has become available to biologists. The FGI, however, remains the simplest accessory modification to a fluorescence microscope giving confocal-quality optical sectioning.
TABLE 2. Use factors affecting fluorescence grating imager performance Minimization of aberration Proper immersion/dial-in spherical aberration correction Proper parfocalization of CCD (at ±0 diopter ocular setting) Imaging-grade excitation filters (or use external filter wheel) Strain-free DR Alignment of light source Use of full INA Stability of light source Precision of exposure timing Shift sequence: Cancellation of harmonics and uniformity of summed exposure in plane of focus Grating parfocus Axial color correction Speed versus movement in specimen Nyquist sampling of image field in camera (pixel spacing ≤λ/4NA) Oversampling of grating image (N > 10 pixels/line pair) Camera linearity, >10-bit digitization Minimization of photobleaching
930
Cite this introduction as Cold Spring Harb Protoc; doi:10.1101/pdb.top083493
Downloaded from http://cshprotocols.cshlp.org/ at TULANE UNIV on September 8, 2014 - Published by Cold Spring Harbor Laboratory Press
Grating Imager Systems
TABLE 3. Comparison of FGI to other optical-sectioning methods Advantages Simple accessory to microscope—all other modes unaffected Fast compared with point-scan CSFM Low light exposure (3 × 50%) Obtain optical section from single focus plane—image stack not required Deconvolution not required—processing is noniterative and fast Conventional light source Standard filter sets Axial bandwidth matches CSFM Low cost Disadvantages Slower than Yokogawa-type spinning-disk multipoint CSFM Background rejection not as high as CSFM Photobleaching or specimen movement gives stripe artifact
ACKNOWLEDGMENTS
The author gratefully acknowledges past support from the National Science Foundation IID and STC Programs, and from Carl Zeiss Inc., Thornwood, New York. The author thanks David Pane (Carnegie Mellon University) and Michel Nederlof (QED Imaging) for much good advice and for the development of instrument control software. Finally, the author especially thanks Winfried Denk for communicating his deep insight into the problem of optical-sectioning microscopy, which inspired much of the past and current work. REFERENCES Bailey B, Farkas DL, Taylor DL, Lanni F. 1993. Enhancement of axial resolution in fluorescence microscopy by standing-wave excitation. Nature 366: 44–48. Ben-Levy M, Peleg E. 1999. Imaging measurement system. U.S. Patent #5,867,604. Frohn JT, Knapp HF, Stemmer A. 2000. True optical resolution beyond the Rayleigh limit achieved by standing wave illumination. Proc Natl Acad Sci 97: 7232–7236. Gustafsson MGL. 2000. Surpassing the lateral resolution limit by a factor of two using structured illumination microscopy. J Microsc 198: 82–87. Gustafsson MG, Shao L, Carlton PM, Wang CJ, Golubovskaya IN, Cande WZ, Agard DA, Sedat JW. 2008. Three-dimensional resolution doubling in wide-field fluorescence microscopy by structured illumination. Biophys J 94: 4957–4970. Juskaitis R, Wilson T, Neil MA, Kozubek M. 1996. Efficient real-time confocal microscopy with white light sources. Nature 383: 804–806. Juskaitis R, Neil MAA, Wilson T. 1998. Microscopy using an interference pattern as illumination source. Proc SPIE 3261: 27–28. Lagerholm BC, Vanni S, Taylor DL, Lanni F. 2003. Cytomechanics applications of optical sectioning microscopy. Methods Enzymol 361: 175– 197. Lanni F. 1986. Standing wave fluorescence microscopy. In Applications of fluorescence in the biomedical sciences (ed. Taylor DL, et al.), pp. 505– 521. A.R. Liss, New York. Lanni F, Baxter GJ. 1992. Sampling theorem for square-pixel image data. Proc SPIE 1660: 140–147. Lanni F, Wilson T. 2000. Grating image systems for optical sectioning fluorescence microscopy of cells, tissues, and small organisms. In Imaging neurons: A laboratory manual (ed. Yuste R, et al.), pp. 8.1–8.9. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY.
Lanni F, Waggoner AS, Taylor DL. 1986. Standing-wave luminescence microscopy. U.S. Patent #4,621,911. Neil MAA, Juskaitis R, Wilson T. 1997. Method of obtaining optical sectioning by using structured light in a conventional microscope. Opt Lett 22: 1905–1907. Neil MAA, Juskaitis R, Wilson T. 1998. Real time 3D fluorescence microscopy by two beam interference illumination. Opt Commun 153: 1–4. Neil MAA, Squire A, Juskaitis R, Bastiaens PIH, Wilson T. 2000. Wide-field optically sectioning fluorescence microscopy with laser illumination. J Microsc 197: 1–4. Schermelleh L, Carlton PM, Haase S, Shao L, Winoto L, Kner P, Burke B, Cardoso MC, Agard DA, Gustafsson MGL, et al. 2008. Subdiffraction multicolor imaging of the nuclear periphery with 3D structured illumination microscopy. Science 320: 1332–1336. Shao L, Isaac B, Uzawa S, Agard DA, Sedat JW, Gustafsson MG. 2008. I5S: wide-field light microscopy with 100-nm-scale resolution in three dimensions. Biophys J 94: 4971–4983. Vanni S, Lagerholm BC, Otey C, Taylor DL, Lanni F. 2003. Internet-based image analysis quantifies contractile behavior of individual fibroblasts inside model tissue. Biophys J 84: 2715–2727. Wilson T, Juskaitis R, Neil MA, Kozubek M. 1996. Confocal microscopy by aperture correlation. Opt Lett 21: 1879–1981. Wilson T, Juskaitis R, Neil MAA. 1997. A new approach to three-dimensional imaging in microscopy. Cell Vis J Anal Morphol 4: 231. Wilson T, Neil MAA, Juskaitis R. 1998a. Real-time three-dimensional imaging of macroscopic structures. J Microsc 191: 116–118. Wilson T, Neil MAA, Juskaitis R. 1998b. Optically sectioned images in widefield fluorescence microscopy. Proc SPIE 3261: 4–6. Wilson T, Neil MAA, Juskaitis R. 1998c. Microscopy imaging apparatus and method. International Patent #WO 98/45745.
Cite this introduction as Cold Spring Harb Protoc; doi:10.1101/pdb.top083493
931
Grating Imager Systems for Fluorescence Optical-Sectioning Microscopy Frederick Lanni Cold Spring Harb Protoc; doi: 10.1101/pdb.top083493 Email Alerting Service Subject Categories
Receive free email alerts when new articles cite this article - click here. Browse articles on similar topics from Cold Spring Harbor Protocols. Confocal Microscopy (92 articles) Fluorescence (416 articles) Fluorescence, general (284 articles) Imaging/Microscopy, general (542 articles)
To subscribe to Cold Spring Harbor Protocols go to:
http://cshprotocols.cshlp.org/subscriptions
© 2014 Cold Spring Harbor Laboratory Press
Downloaded from http://cshprotocols.cshlp.org/ at TULANE UNIV on September 8, 2014 - Published by Cold Spring Harbor Laboratory Press
Topic Introduction
Grating Imager Systems for Fluorescence Optical-Sectioning Microscopy Frederick Lanni
In fluorescence microscopy, optical sectioning is defined as the attenuation or removal of out-of-focus features from an image, and it is a prerequisite for quantitative analysis of three-dimensional structure or function within the specimen. Optical sectioning is most commonly performed by confocal scanning fluorescence microscopy or two-photon scanning fluorescence microscopy. However, structured illumination can be used in conventional fluorescence microscopes to obtain optical sectioning performance, and, in advanced systems, 3D superresolution. The simplest structured-illumination system uses a Ronchi grating as a mask to project parallel stripes within the sharp depth-of-focus of the objective to encode in-focus specimen features differently from out-of-focus features. By shifting the grating, the in-focus image component can be discriminated and separated by elementary image processing operations. This implementation of structured illumination, the fluorescence grating imager, uses a conventional light source, is compatible with all high-quality fluorescence filter sets, and provides high optical-sectioning performance when used to image specimens in which (1) the outof-focus image component is not much brighter than the in-focus features and (2) there is no significant movement in the specimen during the grating shift and image capture process.
OPTICAL SECTIONING
A single optical section is a view of a slice within the specimen centered axially on the plane of focus. A through-focus stack of optical sections therefore constitutes a 3D representation of the specimen. For the past 25 yr, optical sectioning has been driven by the development of the laser confocal scanning fluorescence microscope (CSFM), the two-photon scanning fluorescence microscope, and deconvolution algorithms. In 1997, Wilson and colleagues described both a principle and an optical system to achieve nearly direct optical sectioning in a conventionally illuminated reflectance microscope (Wilson et al. 1997, 1998a,b,c; Neil et al. 1997; Juskaitis et al. 1998). The essential idea was the use of “structured light” or striped illumination projected with sharply defined depth of focus to achieve axial selectivity. The extension to fluorescence was straightforward (Neil et al. 1998, 2000; Wilson et al. 1998b; Ben-Levy and Peleg 1999; Lanni and Wilson 2000; Lagerholm et al. 2003; Vanni et al. 2003), and there are now instruments available from Carl Zeiss (ApoTome; www.zeiss.com/) and from Qioptiq LINOS (OptiGrid; www.qioptiqlinos.com/) that incorporate the concept. In addition, much more complex coherent structured-illumination microscopy systems for optical sectioning and superresolution have been developed which are beyond the scope of this introduction (Lanni 1986; Lanni et al. 1986; Bailey et al. 1993; Frohn et al. 2000; Gustafsson 2000; Gustafsson et al. 2008; Shao et al. 2008; Schermelleh et al. 2008). For the purposes of the present discussion, we refer to the basic instrument as a fluorescence grating imager (FGI). The attractive characteristics of the FGI are its versatility and performance, given its simplicity and relatively low cost. Adapted from Imaging: A Laboratory Manual (ed. Yuste). CSHL Press, Cold Spring Harbor, NY, USA, 2011. © 2014 Cold Spring Harbor Laboratory Press Cite this introduction as Cold Spring Harb Protoc; doi:10.1101/pdb.top083493
923
Downloaded from http://cshprotocols.cshlp.org/ at TULANE UNIV on September 8, 2014 - Published by Cold Spring Harbor Laboratory Press
F. Lanni
OPERATING PRINCIPLE AND INSTRUMENTATION
In a grating imager, only a single essential modification is made to a conventional research-grade fluorescence microscope. A movable Ronchi grating mask is placed in the field iris plane of the incident-light illuminator, thus producing in the specimen striped illumination that is coincident with the geometric focus plane and within the depth of focus of the objective (Fig. 1). The actual focus range of the stripes can be made very sharp if a fine grating is used with an objective of high numerical aperture (NA). Therefore, the image formed by the microscope will consist of striped in-focus features superposed with uniformly illuminated out-of-focus features (background). The period of the projected stripes is the actual grating period divided by the demagnification ratio between the illuminator field iris plane and the specimen, which is objective dependent (and generally not equal to the specified objective magnification). When the grating is shifted so as to move the projected stripes transversely across the object, the fluorescence detected in any pixel element in the image will consist of (1) a steady or DC component owing to the out-of-focus background, plus (2) an oscillating (AC) component owing to the shifting of the stripes across in-focus structures, and (3) shot noise owing to both the DC and AC signal components. In brief, the amplitude of the AC component is the in-focus signal level for that pixel. By shifting the grating between a minimum of three defined positions to obtain three images, a very simple digital image processing operation can subtract away the background and demodulate the AC component to produce an optical section.
A
B
Cooled CCD camera
Field iris Field plane Grating shift
EM
DR Adjustment of grating focus
Specimen
EX
Lamp
Ronchi grating
FIGURE 1. Schematic of fluorescence grating imager. (A) Optics layout showing movable grating mask inserted at location of illuminator field iris. Ronchi gratings (equal-width square-wave pattern) in the range of 8–50 line pairs per mm (LP/mm) are used, depending on the demagnification ratio determined by the objective and the focal length of the illuminator tube lens. A means for bringing the grating to a sharp focus in the plane of focus of the objective is necessary, here shown as an axial adjustment of the illuminator tube lens. To minimize aberration in the grating image, the excitation filter (EX) is shown, removed from its usual position to precede the grating. This is conveniently performed by the use of an external filter wheel between lamp collector lens and microscope stand. Additionally, the dichroic reflector (DR) must be flat and strain-free. Actuators for grating movement can be piezoelectric or electromagnetic, or the grating may be fixed but refractively shifted by a separate optic. Shift increments are generally one-third or one-quarter of the grating period, so they will be in the range 5.00–42.0 µm. Computer control synchronizes discrete grating movement and CCD (charge-coupled device) readout between periods of light exposure. A fast light shutter (not shown) is essential for precision exposure timing and minimization of unnecessary exposure. EM, emission filter. (B) Schematic showing finite grating depth-of-focus within a 3D specimen. Geometrically, it can be seen that both grating spatial frequency and objective NA will affect the depth-of-focus of the pattern.
924
Cite this introduction as Cold Spring Harb Protoc; doi:10.1101/pdb.top083493
Downloaded from http://cshprotocols.cshlp.org/ at TULANE UNIV on September 8, 2014 - Published by Cold Spring Harbor Laboratory Press
Grating Imager Systems
Microscope optical system quality has a definite effect on FGI performance. First, proper immersion is essential to avoid spherical aberration in both the projected grating and the fluorescence image. Indirect or direct water-immersion must be used when optically sectioning living cells, or other lowindex specimens. Oil immersion must be used for specimens mounted in high-index medium. Second, the illuminator optics must provide diffraction-limited imaging of the field iris plane onto the specimen, so that the grating is projected at the full resolution of the objective. Although excitation filters are generally of excellent quality in terms of pass band and blocking, they may not be optically flat, thus introducing aberration into the projected grating pattern. Mounting the excitation filter between the lamp and the grating mask (as shown in Fig. 1) obviates this problem. For the same reason, the DR must be strain-free and flat. Third, the full illumination NA (INA) of the incident light optics must be used to minimize the depth of focus of the projected grating. Fourth, if the instrument is used over a significant range in the UV–visible spectrum (e.g., with DAPI and Cy3 dyes), chromatic shift of focus could necessitate axial adjustment of the grating when filter sets are interchanged. This is analogous to the differential adjustment of pinhole focus in a confocal scanner, and is minimized by the use of apochromatic optics. In the ApoTome and Optigrid, this correction can be set and automated. Lastly, the grating shift mechanism must be precise and accurate, and must hold the mask motionless during each image exposure. COMPUTATION OF THE OPTICAL SECTION
The simplest projected grating image is a sine wave pattern, which, when translated across the specimen, produces a sinusoidal modulation of the fluorescence. Translation of the grating is denoted by the phase shift (φ), where, for example, a shift of 90˚ is equivalent to movement equal to one-quarter of the grating period. In any pixel, the fluorescence signal can be expressed as a periodic function of φ, plus a constant term due to out-of-focus background: f (f) = DC term + AC term = a0 /2 + (a1 cos f + b1 sin f).
(1)
The in-focus part of the total fluorescence signal in that pixel is the AC amplitude, which is given by 2 the Pythagorean sum of the sine and cosine coefficients: a1 + b21 . It can be shown generally that the AC amplitude can be computed on a pixel-by-pixel basis from three images made with the grating shifted between three distinct positions. Three images are necessary because there are three unknown coefficients in Equation 1. The most commonly used shift sequences are given in Table 1, along with corresponding formulas for computing the optical section. The out-of-focus image component, which is the same in all images in the set, is removed by computing differences between image pairs (Fig. 2). The resulting bipolar, striped images are then demodulated by algebraic operations that effectively form the Pythagorean sum. A real projected grating will include spatial harmonics. For a perfect Ronchi (square wave) grating, only odd harmonics occur: f (φ) = DC + (a1 cos φ + b1 sin φ) + (a3 cos 3φ + b3 sin 3φ) + (a5 cos 5φ + b5 sin 5φ) + ··· . The amplitude of the AC term is still the in-focus part of the total fluorescence signal in that pixel. However, now it can be seen that the simple image processing formulas will not give 2 a1 + b21 exactly. The harmonics occur as error terms, and may appear as fine stripes in the optical sections. Two factors can minimize or eliminate this problem. (1) The shift sequence and TABLE 1. FGI image processing formulas Grating shift sequence
Computation of optical section
One-third period, three images 0˚–120˚–240˚ One-quarter period, four images 0˚–90˚–180˚–270˚ One-quarter period, three images 0˚–90˚–180˚
(2 /3)[(i0 − i120) + (i120 − i240) + (i240 − i0) ] (1/2)[(i0 − i180)2 + (i90 − i270)2]1/2 (2−1/2)[(i0 – i90)2 + (i90 – i180)2]1/2
Cite this introduction as Cold Spring Harb Protoc; doi:10.1101/pdb.top083493
1/2
2
2
Uniform exposure 2 1/2
Yes Yes No
925
Downloaded from http://cshprotocols.cshlp.org/ at TULANE UNIV on September 8, 2014 - Published by Cold Spring Harbor Laboratory Press
F. Lanni
FIGURE 2. Optical section computation from image data showing the actin cytoskeleton in a 3T3 fibroblast. Cells were grown on a cover glass under standard incubator conditions, fixed, permeabilized, and stained with rhodaminephalloidin to show F-actin. (A,B,C) i0, i90, and i180 with focus set close to the adherent basal region of the cell. (D, E) difference images (A–B) and (B–C) as in Table 1, formula 3. In D and E, the gray scale is bipolar with 0 at midrange gray. (F ) Optical section computed by Pythagorean summation of D and E. Field of view, 39 µm.
algorithm used can suppress one or more harmonics. As originally pointed out by Wilson et al. (1997), the one-third-period shift sequence exactly compensates the third harmonic, therefore the first error term is due to the fifth harmonic. Because, for a Ronchi grating, the amplitudes drop off as 1/n, this error is small. (2) Because the incoherent modulation transfer function of the microscope decreases with spatial frequency (to zero at the inverse of Abbe’s resolution limit), the microscope attenuates the higher harmonics that are projected into the specimen—further reducing the error terms. In principle, harmonic error can be eliminated altogether by choosing the Ronchi so that its fifth harmonic matches or exceeds Abbe’s resolution limit for the objective. This is not a severe restriction. With the period of the fifth harmonic set equal to Abbe’s resolution limit (λ/2NA), the fundamental period will equal 5 × (l/2NA). This is 2.5× more coarse than the optimal period (λ/NA; see Equation 3 below), but it is projected with high contrast. Alternatively, the grating could be chosen so that the third harmonic was equal to Abbe’s limit, in which case the fundamental period would be 3 × (λ/ 2NA). This is close to the optimal period (λ/NA), but it will have lower contrast than with fifthharmonic suppression. In general, a wide variety of grating shift sequences can be formulated, all of which, in principle, provide the basis for computation of an optical section. Sequences of more than three images per focus plane can be used to compute the contribution of higher harmonics, or, for example, to eliminate second-harmonic error. However, this is a trade-off in terms of speed and light exposure. The particular shift sequence used, the number of images recorded, and the processing formula all affect various performance characteristics of the instrument: For example, as noted, the 0˚– 120˚–240˚ shift sequence eliminates third-harmonic error. It also uniformly exposes the plane of focus, which minimizes patterned photobleaching (Table 1). In general, light exposure of the specimen in the grating imager is comparable with conventional fluorescence microscopy. Because the Ronchi passes 50% of the incident light, the three exposures required for one optical section deliver 1.5× the light dose needed for a conventional image in which the in-focus features are equally bright (but superposed with out-of-focus features). 926
Cite this introduction as Cold Spring Harb Protoc; doi:10.1101/pdb.top083493
Downloaded from http://cshprotocols.cshlp.org/ at TULANE UNIV on September 8, 2014 - Published by Cold Spring Harbor Laboratory Press
Grating Imager Systems
SHARPNESS OF THE OPTICAL SECTION
Based on an analysis of the optical transfer function of the FGI (Lagerholm et al. 2003), it can be shown that the optical section thickness, δ, is related to the NA and projected grating period, L, by the formula d = (l/2)/{[n2 − (NA − l/L)2
1/2 2 − n − NA2 ]1/2 }
for
L . l/2NA.
(2)
The combination of a high-NA objective and a fine grating can produce sharp sections that, in principle, match the optical-sectioning performance of a confocal scanner. Ideally, the sharpest sectioning is obtained when the projected grating period is twice Abbe’s resolution limit: dmin = (l/2)/{n − [n2 − NA2 ]1/2 }
for L = l/NA.
(3)
In practice, a twofold to fourfold coarser grating generally gives better performance owing to increased stripe contrast. As in any form of fluorescence microscopy, FGI signal per pixel is reduced as sectioning is sharpened. SIGNAL-TO-NOISE RATIO
Relative to a confocal scanner, performance of the FGI is limited ultimately by photon counting noise due to out-of-focus (background) features in the object. Unlike a confocal scanner, in which the pinhole blocks most background light before detection, background light in an FGI is detected in each image, and then removed by digital subtraction (Table 1 formulas). Both signal (S) and background (B) photocount levels can be treated as Poisson-distributed variables, with means NS and NB. The subtraction steps remove the background mean, NB, √ but leave a residual noise with a level roughly equal to the Poisson root-mean-square (RMS) value, example, if the background amounts √ B . For √N to 104 counts, and the in-focus signal is 103, then NB = 104 = 100, therefore a signal-to-noise ratio (SNR) of 103/100 = 10 is expected. If the in-focus signal were 100 in a background of 104, it could barely be distinguished from the noise. A more detailed analysis shows the RMS additivity of noise in the image processing steps leading to the optical section. The intrinsic FGI noise level can be estimated from the ideal case of removal of a perfectly defocused background in a “0˚–120˚–240˚” image set with S = 0 (blank specimen). By definition, N√ B is the same for all three images, and the RMS noise level in the resulting blank optical section is 1.155 NB , only 15% greater than the noise level in a single image. Therefore, a√ simple SNR the RMS sum of the Poisson of S to√ √ratio approximation is the noise levels in the signal S + 2B and the background B :SNR S/ S + 2B. SAMPLING IN THE IMAGE PLANE
What sampling is required for a camera to pick up all of the detail in the optical image field that is formed by a microscope? This is set by Nyquist’s sampling theorem, when noise is not a limiting factor: The density of sample points must exceed the inverse of the signal bandwidth. In the case of a microscope, Abbe’s resolution limit, λ/2NA, is the period of the finest grating that can be resolved, and sets the signal bandwidth. The spatial frequency of that grating is 2NA/λ (cycles/µm), and the bandwidth is defined as the frequency interval from −2NA/λ to 2NA/λ, or 4NA/λ. Therefore, the Nyquist sampling interval in nonmagnified coordinates is 1/bandwidth or λ/4NA; that is, the camera pixel spacing divided by total magnification must be ≤λ/4NA (Lanni and Baxter 1992). For typical high-NA microscopes, λ/4NA is generally close to 0.1 µm. When using a camera having, for example, 6.45-µm CCD elements, a total magnification ≥64.5× is required to achieve Nyquist sampling. In real cameras, finite pixel size also causes spatial averaging of the fine details in the image field Cite this introduction as Cold Spring Harb Protoc; doi:10.1101/pdb.top083493
927
Downloaded from http://cshprotocols.cshlp.org/ at TULANE UNIV on September 8, 2014 - Published by Cold Spring Harbor Laboratory Press
F. Lanni
(Lanni and Baxter 1992). The effect of finite CCD pixel size in the grating imager is a reduction in the percent modulation in an image of the grating projected on a uniformly fluorescent specimen. Clearly, if the grating period matched the pixel size, no grating would be seen. A graph of percent modulation versus number of camera pixels across one grating line pair (Fig. 3) shows that modulation is preserved to a high degree when at least 10 pixels span a projected line pair. As an example, a grating chosen to set L = 5λ/2NA (see Computation of the Optical Section) would also make L/10 = λ/ 4NA, the Nyquist interval. PERFORMANCE
Grating modulation (imaged)
A very basic measure of performance for any optical-sectioning fluorescence microscope is its axial response, defined as the graph of pixel brightness versus focus increment for a uniform planar specimen. In practice, a stable fluorescent film that is thin compared with the smaller of (1) the wave-optical depth of field (2nl/NA2) or (2) the optical-sectioning limit must be used to obtain a meaningful measurement. Figure 4A shows graphically the axial response of a grating imager based on a Zeiss Axiovert 200 M operated with a 1.30-NA objective (Zeiss 100× Plan-Neofluar) and 1.33-µm projected grating period, as measured using a 0.05-µm rhodamine-labeled thin film specimen. The optical section thickness, defined by the graphical full width at half-maximum (FWHM), was 0.65 µm. In comparison, Equation 2 gives a value δ = 0.623 µm for NA = 1.3, L = 1.33 µm, and λ = 570 nm (strictly speaking, Equation 2 should distinguish between the excitation and emission wavelengths, but these are usually close; in this example they are 540 and 605 nm, respectively). The experimental FWHM is comparable with a confocal scanner operated with the same objective. Furthermore, the FGI axial response could be further sharpened by use of a finer grating because, in this example, L exceeded λ/NA by a factor of 3 (see Equation 3). In Figure 4B showing the perinuclear actin cytoskeleton of a fibroblast, a focus increment of 0.6 µm (encoded in magenta/green pseudocolor) clearly discriminates axial structure in the FGI optical sections. In cell biological applications, where out-of-focus image features are generally less than an order of magnitude greater in brightness than in-focus features, grating imagers generally produce fluorescence optical sections comparable with high-quality confocal scans. Figure 5 shows the quantitative effect of optical sectioning in a specimen of chick embryo fibroblasts grown in a 3D collagen gel, then fixed and stained for F-actin with rhodamine phalloidin. The left panel in the figure shows a fluorescence image in the absence of optical sectioning, and the right panel shows the FGI optical section at the same plane of focus. A pixel-row profile across the images between the guidelines shows the large effect that out-of focus light would have on quantification of actin in the cytoskeleton, and the degree of improvement obtained with the grating imager. As in any form of high-resolution microscopy, a number of use-related factors have been identified that have a significant effect on performance. Table 2 lists those known to date. These can be classified generally as (1) reduction in aberration and grating defocus (which reduce projected grating contrast and image contrast), (2) optimization of the light source and minimization of light exposure, and (3) optimization of sampling. In comparison to confocal scanning, the FGI has a number of advantages and 1 0.8 0.6 0.4 0.2
5
10
15
N (#pixels/cycle)
928
20
25
FIGURE 3. Effect of finite camera CCD element size on image contrast. Graph shows the contrast of a projected sine wave grating versus number of contiguous pixels per cycle (N). The plotted function is |sin(π/N )/(π/N)|. As N is increased to more than 10 pixels per cycle, the contrast of the projected grating approaches 1 (100%).
Cite this introduction as Cold Spring Harb Protoc; doi:10.1101/pdb.top083493
Downloaded from http://cshprotocols.cshlp.org/ at TULANE UNIV on September 8, 2014 - Published by Cold Spring Harbor Laboratory Press
Grating Imager Systems
A 800 700 600
FGI OS ROI
500
400 300
200 100 0 –1.0
–0.8
–0.6
0.0 –0.4 –0.2 Focus increment (μm)
0.2
0.4
0.6
B
FIGURE 4. (A) Grating imager axial response. The graph shows optical section signal level versus focus drive movement in a small region-of-interest in the center of an image of a thin film. For the test reported here, poly-[methyl methacrylate] covalently labeled with rhodamine (Rh-PMMA) was spin-cast from chlorobenzene onto a coverglass to form a thin film. A simple reflectance interferometer was used to determine the film thickness (50 nm) based on fringe shift and a refractive index equal to 1.49 for PMMA. The coverglass was then mounted with the film in contact with optical cement (Norland Optical Adhesive #60, index = 1.56) on a slide, and UV light was used to polymerize the cement. It was previously determined that Rh-PMMA is insoluble in NOA#60. The advantage of this type of standard specimen is its simple refractive structure and similar refractive indexes (glass 1.52, PMMA 1.49, NOA 1.56). The FWHM defines the sharpness of the optical section, here 0.65 µm. Instrument: Zeiss Axiovert 200 M with Queensgate piezoelectric grating shifter, 100× 1.30 NA oil-immersion objective, 50 LP/mm Ronchi grating projected to 1.33-µm period, Hamamatsu Orca-ER CCD camera with 6.45-µm pixels, and QED imaging in vivo instrument control software. (B) Pseudocolor overlay showing F-actin structures in the perinuclear region of a fibroblast. Field-of-view width, 10.0 µm. The basal plane of focus (magenta) shows small actin fibers below the nucleus. A 0.6-µm focus increment brings stress fibers into view (green). In the FGI optical sections (right), the structures are readily discriminated. Overlay of the corresponding pseudocolor conventional images (left) shows that most structures are not discriminated by focus alone.
disadvantages, which are listed in Table 3. An outstanding advantage is that the instrument is immediately compatible with almost any standard high-quality fluorescence filter set. The only wavelength range limitation so far encountered is longitudinal chromatic aberration when using UV excitation. Finally, although the grating imager has been developed as a fast and compact means to obtain optical sections, it is possible to apply 3D deconvolution methods to the unprocessed serial-focus image sets. Although this is computationally much more demanding, such processing can be performed offline while still obtaining fast optical sections using one of the image-processing formulas in Table 1.
FUTURE OUTLOOKS
In the past several years, a number of multiplex optical sectioning systems have come into general use in addition to FGIs. These would include the Yokogawa-type pinhole array scanning confocal Cite this introduction as Cold Spring Harb Protoc; doi:10.1101/pdb.top083493
929
Downloaded from http://cshprotocols.cshlp.org/ at TULANE UNIV on September 8, 2014 - Published by Cold Spring Harbor Laboratory Press
F. Lanni
FIGURE 5. Quantification of actin cytoskeletal structures in a grating imager optical section (upper right) versus the corresponding conventional image (upper left). (Lower panels) Pixel-row profile between guidelines shows fluorescence from out-of-focus features mixed with in-focus features and how attenuation of the background in the optical section makes quantification possible. Gray scale in optical section profile (lower right) is boosted threefold. The field of view (width and height) of each square image is 165 µm.
microscope and various slit-scan confocal systems. Most recently, a pinhole aperture-correlation confocal scanner (Juskaitis et al. 1996; Wilson et al. 1996) with highly efficient use of the light source has become available to biologists. The FGI, however, remains the simplest accessory modification to a fluorescence microscope giving confocal-quality optical sectioning.
TABLE 2. Use factors affecting fluorescence grating imager performance Minimization of aberration Proper immersion/dial-in spherical aberration correction Proper parfocalization of CCD (at ±0 diopter ocular setting) Imaging-grade excitation filters (or use external filter wheel) Strain-free DR Alignment of light source Use of full INA Stability of light source Precision of exposure timing Shift sequence: Cancellation of harmonics and uniformity of summed exposure in plane of focus Grating parfocus Axial color correction Speed versus movement in specimen Nyquist sampling of image field in camera (pixel spacing ≤λ/4NA) Oversampling of grating image (N > 10 pixels/line pair) Camera linearity, >10-bit digitization Minimization of photobleaching
930
Cite this introduction as Cold Spring Harb Protoc; doi:10.1101/pdb.top083493
Downloaded from http://cshprotocols.cshlp.org/ at TULANE UNIV on September 8, 2014 - Published by Cold Spring Harbor Laboratory Press
Grating Imager Systems
TABLE 3. Comparison of FGI to other optical-sectioning methods Advantages Simple accessory to microscope—all other modes unaffected Fast compared with point-scan CSFM Low light exposure (3 × 50%) Obtain optical section from single focus plane—image stack not required Deconvolution not required—processing is noniterative and fast Conventional light source Standard filter sets Axial bandwidth matches CSFM Low cost Disadvantages Slower than Yokogawa-type spinning-disk multipoint CSFM Background rejection not as high as CSFM Photobleaching or specimen movement gives stripe artifact
ACKNOWLEDGMENTS
The author gratefully acknowledges past support from the National Science Foundation IID and STC Programs, and from Carl Zeiss Inc., Thornwood, New York. The author thanks David Pane (Carnegie Mellon University) and Michel Nederlof (QED Imaging) for much good advice and for the development of instrument control software. Finally, the author especially thanks Winfried Denk for communicating his deep insight into the problem of optical-sectioning microscopy, which inspired much of the past and current work. REFERENCES Bailey B, Farkas DL, Taylor DL, Lanni F. 1993. Enhancement of axial resolution in fluorescence microscopy by standing-wave excitation. Nature 366: 44–48. Ben-Levy M, Peleg E. 1999. Imaging measurement system. U.S. Patent #5,867,604. Frohn JT, Knapp HF, Stemmer A. 2000. True optical resolution beyond the Rayleigh limit achieved by standing wave illumination. Proc Natl Acad Sci 97: 7232–7236. Gustafsson MGL. 2000. Surpassing the lateral resolution limit by a factor of two using structured illumination microscopy. J Microsc 198: 82–87. Gustafsson MG, Shao L, Carlton PM, Wang CJ, Golubovskaya IN, Cande WZ, Agard DA, Sedat JW. 2008. Three-dimensional resolution doubling in wide-field fluorescence microscopy by structured illumination. Biophys J 94: 4957–4970. Juskaitis R, Wilson T, Neil MA, Kozubek M. 1996. Efficient real-time confocal microscopy with white light sources. Nature 383: 804–806. Juskaitis R, Neil MAA, Wilson T. 1998. Microscopy using an interference pattern as illumination source. Proc SPIE 3261: 27–28. Lagerholm BC, Vanni S, Taylor DL, Lanni F. 2003. Cytomechanics applications of optical sectioning microscopy. Methods Enzymol 361: 175– 197. Lanni F. 1986. Standing wave fluorescence microscopy. In Applications of fluorescence in the biomedical sciences (ed. Taylor DL, et al.), pp. 505– 521. A.R. Liss, New York. Lanni F, Baxter GJ. 1992. Sampling theorem for square-pixel image data. Proc SPIE 1660: 140–147. Lanni F, Wilson T. 2000. Grating image systems for optical sectioning fluorescence microscopy of cells, tissues, and small organisms. In Imaging neurons: A laboratory manual (ed. Yuste R, et al.), pp. 8.1–8.9. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY.
Lanni F, Waggoner AS, Taylor DL. 1986. Standing-wave luminescence microscopy. U.S. Patent #4,621,911. Neil MAA, Juskaitis R, Wilson T. 1997. Method of obtaining optical sectioning by using structured light in a conventional microscope. Opt Lett 22: 1905–1907. Neil MAA, Juskaitis R, Wilson T. 1998. Real time 3D fluorescence microscopy by two beam interference illumination. Opt Commun 153: 1–4. Neil MAA, Squire A, Juskaitis R, Bastiaens PIH, Wilson T. 2000. Wide-field optically sectioning fluorescence microscopy with laser illumination. J Microsc 197: 1–4. Schermelleh L, Carlton PM, Haase S, Shao L, Winoto L, Kner P, Burke B, Cardoso MC, Agard DA, Gustafsson MGL, et al. 2008. Subdiffraction multicolor imaging of the nuclear periphery with 3D structured illumination microscopy. Science 320: 1332–1336. Shao L, Isaac B, Uzawa S, Agard DA, Sedat JW, Gustafsson MG. 2008. I5S: wide-field light microscopy with 100-nm-scale resolution in three dimensions. Biophys J 94: 4971–4983. Vanni S, Lagerholm BC, Otey C, Taylor DL, Lanni F. 2003. Internet-based image analysis quantifies contractile behavior of individual fibroblasts inside model tissue. Biophys J 84: 2715–2727. Wilson T, Juskaitis R, Neil MA, Kozubek M. 1996. Confocal microscopy by aperture correlation. Opt Lett 21: 1879–1981. Wilson T, Juskaitis R, Neil MAA. 1997. A new approach to three-dimensional imaging in microscopy. Cell Vis J Anal Morphol 4: 231. Wilson T, Neil MAA, Juskaitis R. 1998a. Real-time three-dimensional imaging of macroscopic structures. J Microsc 191: 116–118. Wilson T, Neil MAA, Juskaitis R. 1998b. Optically sectioned images in widefield fluorescence microscopy. Proc SPIE 3261: 4–6. Wilson T, Neil MAA, Juskaitis R. 1998c. Microscopy imaging apparatus and method. International Patent #WO 98/45745.
Cite this introduction as Cold Spring Harb Protoc; doi:10.1101/pdb.top083493
931
