We report on a simple yet powerful implementation of objective-type total internal reflection fluorescence (TIRF) and highly inclined and laminated optical sheet (HILO, a type of dark-field) illumination. Instead of focusing the illuminating laser beam to a single spot close to the edge of the microscope objective, we are scanning during the acquisition of a fluorescence image the focused spot in a circular orbit, thereby illuminating the sample from various directions. We measure parameters relevant for quantitative image analysis during fluorescence image acquisition by capturing an image of the excitation light distribution in an equivalent objective back-focal plane (BFP). Operating at scan rates above 1 MHz, our programmable light engine allows directional averaging by circular spinning the spot even for sub-millisecond exposure times. We show that restoring the symmetry of TIRF/HILO illumination reduces scattering and produces an evenly lit field-of-view that affords on-line analysis of evanescent-field excited fluorescence without pre-processing. Utilizing crossed acousto-optical deflectors, our device generates arbitrary intensity profiles in BFP, permitting variable-angle, multi-color illumination, or objective lenses to be rapidly exchanged.
©2008 Optical Society of America
Objective-type total internal reflection fluorescence (TIRF) (1) and highly inclined laminated optical sheet (HILO) illumination (2) rely on tightly focusing a laser spot in an eccentric position in the back focal plane (BFP) of a high numerical-aperture (NA) objective, Fig. 1(a). Common problems occurring with this type of illumination are interference fringes and uneven illumination (3–7). Scattering at refractive-index boundaries like cellular adhesion sites or intracellular organelles create further inhomogeneity. Biological tissue causes mostly forward scattering. The light confinement that is at the origin of the contrast enhancement observed with these wide-field fluorescence techniques is compromised, as fluorescence is generated by a superposition of the spatially confined and diffuse component of scattered excitation light that varies across the field (7). Non-linear excitation can be a remedy, because the scattered intensities are too weak to produce two-photon excited fluorescence (2PEF). However, a 2PEF TIRF setup involves high cost and complexity and, at least in its objective-type variant, imposes constraints on the excitation power. Inspired by the ring illumination used in the original work (1), a scanning-type of TIRF illuminator has recently been proposed (8,9) that restores radial symmetry by circularly spinning the illumination spot in the BFP. Averaging over different orientations of the electromagnetic wave vector, azimuthally spinning spatially redistributes and dilutes scattered photons. Hence, the relative contribution of diffuse excitation is reduced. However, in these devices scan speed has been limited, reducing frame rates to a few Hz.
We here report on a straightforward yet powerful implementation of fast scanning TIRF/-HILO microscopy. Acousto-optical deflectors (AODs) in conjunction with a high-quality scan lens and high-NA objective permit us to point a focused spot within µs to any point in the objective BFP. Unlike with rotating wedges (8) or kinematic mirrors (9), we scan a full circle at 1.2 kHz, allowing the acquisition of evenly lit and highly contrasted fluorescence images with millisecond exposure time. Circular scans at different radii allow for accommodating high-NA objectives of different magnification, and permit multi-color alternating excitation (ALEX) (11) TIRF while maintaining a constant penetration depth.
2. Experimental procedures
2.1 Eccentric-spot illumination
Focusing a laser beam in the objective aperture plane (or back focal plane, BFP) produces a thin collimated beam emerging from the objective. Its radial distance r from the optical axis determines the beam angle α. With Abbe’s sine condition, r=f·NA, and NA=n 2sinα NA, we have α=arcsin[M·r/(n 2·f TL)]. Here, f=f TL/M, f TL and M are the objective and tube lens focal lengths, and the objective transverse magnification, respectively, Fig. 1(b). TIR occurs above the critical angle αc=arcsin(n 1/n 2), i.e., for values of r≥r c=n 2·f obj and sets up an evanescent wave (EW) with a penetration depth ẉz≈λ(4π)-1 [(/n 2 2sin2 α)2-n21]-1/2, of the order of ~λ/5. λ n 1 and n 2 denote the wavelength of light and the sample and substrate refractive indices, respectively. For HILO, we set α just beneath α c to produce a highly inclined sheet of light that transverses the sample at an oblique angle and illuminates a thin optical section close to the cover slip (2). In either case, the objective back pupil Øpupil=2r max=2n 2sinα NA·f FL/M determines the maximal beam angle, α NA.
Upon TIR, the average energy flux across the refractive-index boundary is zero. As the dominant component of the wave vector in HILO illumination, the Poynting vector P=E×H of the evanescent wave is oriented parallel to the boundary and takes, for TE polarization of the incident beam, the form
Here, θ ⊥ denotes the phase, and is the solution to tan(θ ⊥/2)=(sin2 α-n 2 12)1/2/cosα(10). Therefore, in both TIRF and HILO, the commonly used eccentric-spot illumination breaks the radial symmetry and defines a propagation direction of the electromagnetic wave field along the positive x-axis.
Figure 2 shows the optical layout. A 405-nm laser beam (40 mW, Lasos, Jena, Germany) is cleaned up with a HQ410/40 band-pass, (EX1, AHF Analysentechnik, Tübingen, Germany) and directed on a pair of crossed acousto-optic devices (AOD 1, AA.Opto-Electronique, St. Rémy-en-Chévreuse, France). Laser and AODs are mounted on a tilt and pan table for injecting the (1,1)-order diffracted beam into the microscope. A personal computer (PC) and peripheral interface microcontroller (PIC) control the AODs and provide trigger signals to synchronize an AOD-based shutter (AOD 2, Brimrose, Baltimore, Maryland, U.S.A.) and detectors. A Galilean telescope (f 1=100 mm, f 2=50 mm) augments the full scan angle to ±2.1°. A Rodenstock Rodagon (FL, , Linos, f FL=135 mm) focuses the beam in the aperture plane of a NA-1.45 oil objective (see Table 1 for objectives used). Neutral density filters (ND) attenuate the beam to <10 µW impinging at the sample, corresponding to ~1W/cm2 intensity. A charge-coupled device camera (CCD 1, Foculus FO432SB, Elvitec, Pertuis, France, 4.65-µm pixel size) captures the 0.3% intensity transmitted by the Z405RDC diachroic reflector (DC, Sem-rock, Rochester, NY, U.S.A.) to provide a scaled (×0.3) image taken at a position equivalent to that of the objective BFP.
The generated fluorescence is filtered by a HQ575/150 band-pass and projected with a 160-mm doublet (Olympus) on a fast EMCCD camera (CCD 2, Cascade128+, Photometrics, Tuczon, AZ, USA). Latchable lenses match the 24-µm pixels to the calculated resolution. For inspecting and focusing at target cells, white light-emitting diode (HighLED, Linos, Göttingen, Germany, 4.0 mW at 440 nm) provides bright-field illumination through a f 6=200 mm lens and 0.8-NA condenser (WI-OBCD, Olympus). The addition of a band-pass filter (EX2, D425/60x, Chroma, Rockingham, VT, U.S.A.) permits wide-field fluorescence excitation. A computer-controlled 3-axes micromanipulator (HS 6/3 & MCL 3, Märzhäuser, Wetzlar, Germany) positions the sample relative to the objective. Focal stability is maintained by moving the objective with a piezoelectric focus drive (PIFOC, Physik Instrumente, Waldbronn, Germany) under feedback control.
2.3 Beam steering software and PIC microcontroller
Peripheral interface controllers provide an inexpensive stand-alone solution for in- and output control. We used a PIC16F877A chip (Microchip Technol., Chandler, AZ, USA) that has a Harvard architecture, separating memory from the central processing unit. The PIC runs autonomously on a low-level assembler code. For programming (12), we used the freeware integrated development environment package MPLAB, which includes the assembler, linker, software simulator and debugger (13). A DELL OptiPlex745 PC controlled the PIC through the serial port by a C++ program, written in Bloodshed C++ IDE 22.214.171.124 with a graphic user interface toolkit from Trolltech Qt 4.2.3 (14, 15). The adjustable 8-bit output from the PIC was set between 0 and 13.0 V and fed into the AOD driver, equivalent to an output of 100 to 160 MHz. We implemented different waveforms as scalable press-botton-driven shortcuts.
2.4 Sample preparation
Cover slips (Schott Desag D263, n=1.533 at 405 nm, Menzel, Braunschweig, Germany) were sonicated in absolute EtOH, rinsed with 6 M KOH and their autofluorescence bleached with a 254/365-nm UV lamp (UVGL-58, UVP, Upland, CA, U.S.A.). Following incubation at 37°C during 2 hours with 0.1 mg/ml biotinylated bovine serum albumin (pH 7.3) and thoroughly rinsing, they were incubated during 1 min with ~10 pM streptavidin-conjugated quantum dots (Qdots565-ITK, Invitrogen, CA, U.S.A.) in 50 mM borate buffer (pH 8.3). Unbound Qdots were flushed with buffer. We used Cargille (Cedar Grove, NJ, U.S.A.) FF low fluorescence immersion oil (n=1.489 at 405 nm) for optically coupling the cover slip and objective.
Cortical mouse astrocytes were prepared as described (16). We labeled astroglial lysosomes (16) with the styryl dyes FM1-43 (1 µM, 10 min, followed by 20 min wash, Invitrogen, Carlsbad). Alternatively, we transfected cells with 1.2 µg/ml vinculin-EGFP (17) (a kind gift of Dr Johannes Hirrlinger, Leipzig) using lipofectamine 2000 (Invitrogen). Cells were perfused at 0.5 ml/min with extracellular saline containing (in mM): 140 NaCl, 5.5 KCl, 1.8 CaCl2, 1 MgCl2, 10 HEPES and 20 glucose, pH 7.3 (KOH).
3. Results and discussion
3.1 Reading out relevant beam parameters from the BFP image
We built a microscope optimized for high-NA illumination, as used in TIRF or HILO microscopy. In addition to the usual fluorescence image, we simultaneously acquired a BFP image that contains all parameters relevant for characterizing the beam and azimuthal angle, as well as their angular spread, thus permitting true quantitative fluorescence analysis. Focused in the center of the BFP, an on-axis beam (α=0), produced a thin collimated beam that emerged vertically from the objective. To set the midpoint (x 0,y 0) on the BFP image, we positioned CCD 1 so that it showed a centered spot. We then increased r=(x 2+y 2)1/2 until we observed TIR, meaning that we reached the critical radius r c. The triplet r=0, r=r c and r NA=r max forms the basis of a three-point calibration for α(r). Corroborating earlier observations (4,5), the EW-excited fluorescence in a dilute fluorescein solution revealed an asymmetric intensity profile, tailing off toward the positive x-axis. Thus, even a low dye concentration is sufficient to produce observable forward scattering. We next used our AOD scanners to rotate the illumination spot at constant r. Azimuthal spinning restored the symmetry, averaged over directional effects and generated an even illumination across the entire field-of-view, Fig. 3(a). Spinning the spot also rotates the polarization vector of a linearly polarized input beam in the sample plane, and integration during fluorescence image acquisition averages over this effect. The precise measurement of the beam angle α allowed us to generate a look-up table α(r), for each objective lens used. Plotting the fluorescence generated in fluorescein solution vs. the angle of incidence displayed the expected steep drop of fluorescence due to the sudden confinement of excitation at α c=60±1°, Fig. 3(b).
3.2 Circular spinning permits the on-line quantification of fine morphological detail
The benefit for biological fluorescence detection of circular spinning the excitation beam is evident on fluorescence images of mouse cortical astrocytes stained with styryl pyridinium dyes excited at angles just below the critical angle (HILO). Astrocytes are a type of macroglia and are the dominant glial cells of the brain, where they play important roles in metabolism and signaling as partners of neurons. In cultured cortical astrocytes styryl dyes specifically label lysosomal/late endosomal compartments (16). These membrane-delimited and optically dense organelles appear as diffraction-limited spots on EW-excited fluorescence images. Compared to conventional eccentric-spot illumination (cardinal images for NWSE orienttation), fluorescence flare is absent from the central image acquired while continuously spinning the laser beam in the BFP (inset). A large and evenly lit field-of-view is an additional benefit, Fig. 4(a). Broadening the incident laser beam, which is easily achieved by a slight defocus in BFP, can also improve the image quality upon illumination from one side, but the resulting image inevitably suffers from power losses and effects of forward scattering. Zooming in on sub-cellular fluorescence and slightly increasing the beam angle beyond the critical angle provided a crisp TIRF image of near-membrane granular and tubular (probably mitochondrial) fluorescence. Importantly, circular spinning produced fluorescence images, from which we directly read off intensity profiles without the usually required local-background subtraction, filtering or pre-processing. Hence, restoring symmetry facilitates and speeds up on-line analysis of TIRF images, Fig. 4(b).
3.3 Changing magnification while maintaining a constant penetration depth
Conventional fluorescence microscopes accommodate objectives of different magnification by overfilling the objective back aperture. However, TIRF or HILO, changing objectives is then virtually impossible, because the diameter of the objective back pupil (Øpupil) varies reciprocally with magnification M, and so should the illuminating ring to maintain a constant penetration depth or inclined optical sheet. Re-adjusting the beam angle manually can be a solution, but has usually prohibited experimenters from combining different high-NA objectives during one experiment.
On our setup, we are able to swap objectives and to adapt instantly the illumination pattern under computer control, just by changing the α(r) look-up table. This feature permits to zoom in on characteristic features of the sample under study without. We demonstrate this feature by imaging focal adhesion contacts (FACs)— molecular structures by which cells make mechanical contact to the substrate on which they are grown.
We validated our approach by imaging cellular focal adhesion sites labeled with vinculin-EGFP. Vinculin is a three-domain adaptor protein that mediates focal adhesion by regulating integrin dynamics and clustering as well as transducing force to the actin cytoskeleton (17) and is believed to interact transiently with the lipid membrane. However, the nature and role of these interactions in situ is poorly understood, because high-resolution imaging of the strongly scattering protein-rich FACs has presented a major challenge.
To visualize adhesion sites, we transiently expressed in cortical astrocytes a vinculin-EGFP construct. On TIRF images, vinculin-EGFP appeared concentrated in bright spots, from which dimmer filaments emerged and radiated into deeper cytoplasmic regions, Fig. 5. Images at ×60, ×100 and ×150-magnification showed the same cell at increasing detail, whereas the axial section illuminated did not appreciably vary. Thus, our programmable light engine allows to instantly adapting the light path to objectives of different magnification. We expect this feature not only of use for zooming in and out, but also for variable-angle scans or multi-color EW-excitation, where changes in penetration depth w z resulting from wavelength changes can be compensated for by adjusting α. Also, this flexibility removes the current requirement for bulky and expensive multiple-port TIRF attachments when multiple lasers are coupled to one microscope. This feature will also be of particular interest for motorized microscopes,
3.4 kHz circular spins allow image acquisition with exposure times down to a few ms
Single-molecule detection (SMD) is an important application for TIRF. The sub-wavelength axial optical sectioning and near-field enhancement create the conditions for detecting single fluorophores in front of a reduced background signal. The same is true for the detection of single-pair fluorescence resonance energy transfer (spFRET) or the time-lapse observation and tracking of single fluorescently labeled organelles in biological TIRF microscopy. By building up the ensemble average from individual events, these techniques reveals heterogeneity otherwise masked by the average behavior but, in turn, require large a number of individual observations to discern a statistically relevant pattern. For such a compilation to be valid, care must be taken to maintain experimental conditions constant across observations.
We prepared a monolayer of single emitters by immobilizing streptavidine-linked semiconductor nanocrystals (‘quantum dots’) on a biotinylated microscope slide to verify if spinning the excitation spot improved the conditions for TIRF-based SMD. Circular scan rates depended on number of orientations to average over for a circular full scan and varied between 0.3 for 256 and 1.2 kHz for 64 directions. Thus, with 6-bit azimuthal angle resolution we could stream images frames at up to 500 Hz with only the number of signal photons limiting the instrument performance. Traces of 405-nm excited photoluminescence showed ON-OFF states associated with individual quantum dots, or the presence of two quantum dots in the diffraction-limited voxel, recognized as three intensity states, Fig. 6. As with earlier experiments, the spatially invariable background and signal-to-noise ratio allowed for the direct comparison of individual traces, thereby considerably speeding up image analysis.
4. Discussion and conclusion
Since its introduction in 1989 (1), objective-type TIRF is broadly being used in microscopy, spectroscopy and SMD. Unlike the alternative prism-based geometry (see refs. 18–22), it is easy to set up on an inverted microscope, permits the free access to the sample from above, and is readily combined with epifluorescence and bright-field illumination. The large-NA collection permits high lateral resolution (∝NA -1, in the limit NA≤n 1) and fluorescence collection efficiency (∝½[1-cosα NA]). Dedicated ‘TIRF illuminators’, high numerical aperture objectives (NA=1.44-1.65, ref.23) and compact, fiber-coupled diode lasers have contributed to its wide use. More recently, isosceles through-the-lens injection of two colliding high-NA beams has been used to generate the high-frequency illumination pattern for structured illumination in objective-base ‘super-resolution’ TIRF (24–26). The containment of the beam inside the microscope body facilitates compliance with laser safety regulations, which is a concern for multi-user facilities and imaging platforms.
Early objective-type TIRF achieved symmetric illumination by focusing a mercury arc into a ring with r≥r c, and blocking the central sub-critical rays (1). Olympus (Shinjuku-ku, Japan) developed but never commercialized a variant in which a highly aberrant lens images the arc to a crescent-shaped sector of the high-NA ring (Fig. 1(c), middle). However, a consequence of using an incoherent white-light source are intensity losses, because imaging the a mercury arc of dimension δ into a ring with width r 2-r 1 (r 2>r 1≥r c) decreases its luminous density by δ 2/[ζ·(r 2 2-r 2 1)]≪1, where ζ is the part of the full circle illuminated. In addition, the alignment of the peripheral rays is finicky when objectives with a high magnification or a NA close to n 2 are used. Hence, more recently, experimenters have preferred an eccentric focused laser spot rather than a full ring for providing supercritical illumination.
Our AOD-based programmable light engine restores the illumination symmetry through beam scanning and thereby combines the advantages of the original ring illumination and the laser-based eccentric-spot scheme. Unlike other devices, our programmable light engine offers a rapid, random-access computer-controlled beam steering in the objective back focal plane at moderate cost (<20k€). Flying spot scanning averages out illumination heterogeneities and permits the rapid shifting between TIRF, HILO and epifluorescence illumination by a simple change in the computer controlled scanning pattern. BFP imaging on an inexpensive second CCD detector permitted us direct visual control and on-line beam diagnosis. Our study shows that fluorescence image quality improves under a variety of conditions, and that even ms exposure times can benefit from azimuthal spinning, due to the high scan rate offered by the AODs. On-line image analysis of sub-cellular fluorescence without pre-processing is possible due to the lower background and even illumination, and should increase throughput while maintaining the high information content of TIRF images.
Our instrument provides an ideal platform for quantitative high-resolution TIRF and HILO imaging, for biological single-molecule detection, TIRF-FRAP; -photoactivation or FCS. It should equally be of interest for the growing community of academic and industrial researchers developing ‘screening by imaging’ assays for high content, high throughput microscopy.
We acknowledge the help of K. Herault, D. Li and A. V. Yakovlev with experiments. We thank R. Uhl and N. Ropert for discussion and S. Konzack (Olympus Europa, Hamburg, Germany) for the loan of equipment and K. Yamazaki for sharing unpublished technical data. Supported by the Agence National de la Recherche (GIP-ANR PNANO-2005-051-01) and the European Union (EU, FP6-2004-013880 ‘Single-motor FLIN’, FP6-2006-037897 ‘AUTOSCREEN’, to MO). MvH is the recipient of an EU Marie-Curie early-research career development fellowship (FP6-2005-019481 ‘From FLIM to FLIN’). VdS acknowledges support from the Centre National de la Recherche Scientifique (CNRS).
References and links
6. M. Oheim and F. Schapper, “Non-linear evanescent-field imaging,” J. Phys. D: Appl. Phys. 38, R185–R197 (2005). [CrossRef]
7. A. L. Mattheyses and D. Axelrod, “Direct measurement of the evanescent field profile produced by objective-based total internal reflection fluorescence,” J Biomed Opt. 11, 014006A (2006). [CrossRef]
8. A. L. Mattheyses and D. Axelrod, “Effective elimination of laser interference fringing in fluorescence microscopy by spinning azimuthal incidence angle,” Microsc. Res. Tech. 69, 642–647 (2006). [CrossRef] [PubMed]
9. R. Fiolka, Y. Belyaev, H. Ewers, and A. Stemmer, “Even illumination in total internal reflection fluorescence microscopy using laser light,” Microsc. Res. Tech. 71, 45–50 (2007). [CrossRef] [PubMed]
10. F. de Fornel, Evanescent waves: from Newtonian Optics to Atomic Optics, (Berlin, Heidelberg, New York, Springer, 2000) ISBN 3540658459.
11. N. K. Lee, A. N. Kapanidis, and H. R. Koh, et al, “Three-color alternating-laser excitation of single molecules: monitoring multiple interactions and distances,” Biophys. J. 92, 303–312 (2007). [CrossRef]
12. M. Predko, Progamming and Customizing PICmicro® Microcontrollers (McGraw-Hill, New York, 2000)
13. Microchip PIC16F87XA data sheet and MPLAB Integrated Development Environment, www.microchip.com
14. gcc C++ compiler for Windows, http://www.mingw.org/
15. Trolltech, Qt graphical library, www.trolltech.com
16. D. Li, N. Ropert, A. Koulakoff, C. Giaume, and M. Oheim, “Lysosomes are the major vesicular compartment undergoing Ca2+-regulated exocytosis from cortical astrocytes,” J. Neurosci. 28, 7648–7658 (2008). [CrossRef] [PubMed]
17. J. D. Humphries, P. Wang, C. Streuli, B. Geiger, M. J. Humphries, and C. Ballestrem, “Vinculin controls focal adhesion formation by direct interactions with talin and actin,” J. Cell Biol. 179, 1043–1057 (2007). [CrossRef] [PubMed]
19. W. P. Ambrose, P. M. Goodwin, and J. P. Nolan “Single-molecule detection with total internal reflection
20. excitation: comparing signal-to-background and total signals in different geometries,” Cytometry 36224–31 (1999).
21. P. B. Conibear and C. R. Bagshaw, “A comparison of optical geometries for combined flash photolysis and total internal reflection fluorescence microscopy,” J. Microsc. 200, 218–29 (2000). [CrossRef] [PubMed]
22. B. P. Ölveczky, N. Periasamy, and A. S. Verkman, “Mapping fluorophore distributions in three dimensions by quantitative multiple angle-total internal reflection fluorescence microscopy,” Biophys J. 73, 2836–47 (1997). [CrossRef] [PubMed]
23. Y. Kawano, C. Abe, T. Kaneda, Y. Aono, K. Abe, and K. Tamura, “High-numerical aperture objective lenses and optical system improved objective-type total internal reflection fluorescence microscopy,” Proc. SPIE 4098, 142–53 (2000). [CrossRef]
24. O. Gilko, G. D. Reddy, B. Anvari, W. E. Brownell, and P. Saggau, “Standing wave total internal reflection fluorescence microscopy to measure the size of nanostructures in living cells,” J. Biomed. Opt. 11, 064013 (2006). [CrossRef]
25. E. Chung, D. K. Kim, and P. T. C. So, “Extended resolution wide-field optical imaging: objective-launched standing wave total internal reflection fluorescence microscopy,” Opt. Lett. 31, 945–947 (2006). [CrossRef] [PubMed]
26. M. R. Beversluis, G. W. Bryant, and S. J. Stranick, “Effects of inhomogeneous fields in superresolving structured illumination microscopy,” 25, 1371–1377 (2008).