Scanned LSFM with Confocal Detection
Improving Contrast and Resolution in Scattering Samples
- Fig. 1: Working principle of an sCMOS rolling shutter . Pixel exposure is initiated at the sensor´s top row and propagates along the sensor (red line). After expiry of the exposure time, the readout is carried out (yellow line). Between both lines is a slit-shaped region of simultaneously active pixels, which is termed "rolling shutter" and acts as a moving slit detector.
- Fig. 2: (A) Schematics of the specimen illumination optics. The beam exiting the fiber is focused into the sample and is parallel to the optical axis, which is accomplished by positioning the scanning mirror conjugate to the back focal plane (BFP) of the illumination objective. The relay optics widens the beam appropriately, and maps the scanning mirror into the BFP of the illumination objective, since the latter is located inside the objective housing. (B) Fluorescence light is collected by a detection objective orthogonal to the illumination path and imaged onto a sCMOS camera. (C) The image of the scanned beam has to coincide withe the rolling shutter. A rolling shutter width equal to the beam width produced the best results.
- Fig. 3: Deterioration of imaging quality in a scattering sample . (A) Illumination beam in a non-scattering dye solution. (B) Sample with fluorescent beads. The illumination beam is scattered and broadened on its way through the sample. Also scattered fluorescence sums up to a homogenous image background. (C) Improvement of contrast and signal-to-noise ratio by using rolling shutter detection. Illumination was from left. Scale bar, 10 µm.
- Fig. 4: C. tentans salivary gland cell nucleus. (A) Image taken in global shutter mode. During acquisition of the image the beam was swept once across the sample. (B) Image taken in rolling shutter mode. Shutter width was equal to the illumination beam diameter. (C) The increase in contrast and resolution is clearly confirmed by the intensity plots along the indicated lines. Image size, 70 x 70 µm2.
Light sheet fluorescence microscopy (LSFM) is a technique with a vivid development [1, 2]. In contrast to epi-illumination microscopy the sample is illuminated with a thin light sheet orthogonally to the detection path. Most often the sheet is formed by rapidly scanning a laser beam across the sample. This illumination yields intrinsic optical sectioning and reduction of photobleaching and phototoxicity since excitation is limited to fluorophores inside the observation plane. This is particularly beneficial for long term imaging in developmental biology and reduces image background [3-5].
However, in scattering samples the stray photons produce a nonspecific background reducing contrast, signal to noise ratio and the effective resolution. In order to diminish this problem, we developed a setup that exploits the rolling shutter of a scientific CMOS (sCMOS) camera as a confocal slit detector . This technique amalgamates the best of two strong techniques: the superior contrast and resolution capabilities of confocal imaging with the elegant illumination scheme, high detection efficiency and high frame rates of LSFM . It does not require any modifications of the optical detection path and no image post-processing.
Sheet Illumination versus Confocal Laser Scanning Microscopy
In LSFM images can be acquired with high frame rates and sensitivity by pa-rallel image acquisition devices such as CCD or sCMOS detectors. However, not all detected photons convey information, which is useful to create a sharp image, since they are also elastically scattered on their way through the specimen. Scattering occurs for both the excitation and the emission light. Though there is a high probability for forward scattering, the summation of many events leads to a random distribution of propagation directions. Scattered photons are likely to hit a random detector pixel and contribute to a nonspecific image background. A confocal detection scheme could certainly alleviate this problem.
Using a Scientific CMOS Camera for Confocal Slit Detection
This can be accomplished in a very elegant manner by taking advantage of a special readout mode featured by sCMOS cameras, the so-called "rolling shutter".
Each pixel of a sCMOS chip has its own readout and amplification unit. In the rolling shutter mode the exposure is initiated at the topmost pixel line and advances to the next row until it reaches the bottom of the chip (fig. 1). The time between activation of subsequent rows is referred to as the line activation time, and the exposure time is an integer multiple thereof. The shortest possible exposure time for a pixel is equal to the line activation time. Selecting it would result in only one single active row at a time moving from top to bottom of the chip. Increasing the exposure time means to increase the number of simultaneously exposed rows thus forming a slit-shaped region of active pixels, which moves along the detector and truly forms a "rolling shutter".
The readout of a classical CCD occurs in a principally different manner. The pixel exposure of a CCD camera is initiated and terminated for all pixels simultaneously. This image acquisition mode is designated as "global shutter". Scientific CMOS cameras also allow to utilize the global shutter mode.
By synchronizing sample illumination by a scanned Gaussian laser beam and the detection of the excited fluorescence by the rolling shutter of a sCMOS camera a confocal arrangement of spatially confined illumination and detection volumes can be realized perpendicular to the scanning direction. Reduction of the rolling shutter width allows to prevent a large fraction of scattered photons to hit the active detector lines and from contributing to image formation, while the ballistic photons directly hit the active pixels in the rolling shutter region.
Our setup was based on a conventional scanned light sheet microscope . A galvanometric scanning mirror was placed in the front focal plane of a relay lens system. It was imaged into the back focal plane of a 10X achromat such that its motion translated into a telecentric scanning of the object plane (fig. 2). The illumination beam radius was adjusted to a full width at half maximum (FWHM) of 5.2 μm yielding a Rayleigh length of zr = 138 μm along the illumination direction for an excitation wavelength of 633 nm. Optical sectioning could be enhanced by introducing a beam expander in front of the scanning mirror, thereby reducing the beam waist to a FWHM of 3.3 μm with zr = 38 μm. The setup was designed to augment an inverted microscope, which enabled the use of a variety of high NA detection objectives. We used sCMOS cameras (Orca Flash 2.8 and 4.0, Hamamatsu Photonics K.K., Japan) for image acquisition. The minimal time to acquire one full frame was 22 ms for the Flash 2.8 and could be decreased by vertical reduction of the active chip area. The Orca Flash 4.0, its successor, comprises two adjacent detector chips and was modified by the company in order to better fit the requirements of confocal slit detection. This included the ability to run the rolling shutter along both detectors without interruption and to control the speed of the rolling shutter. This allowed to utilize the complete detector area and to vary the width of the rolling shutter independently of the exposure time.
The movement of the scanned beam must perfectly be synchronized with the rolling shutter demanding a very precise control of scanning mirror and camera trigger. This was achieved by using the internal routing of a NI USB-6221 data acquisition and generation device (DAQ, National Instruments Corp., Austin, TX, USA). The mirror was driven by an analogue saw-tooth voltage curve synchronized with the camera trigger. Laser intensity was controlled by an acousto-optical tunable filter. The first images were acquired in the global shutter mode. Figure 3A shows an image of the illumination beam entering from the left into a sample containing a non-scattering fluorescence solution (Alexa Fluor 647 in buffer). Figure 3B reveals the deterioration of the beam and the fluorescence background in a scattering sample corresponding to immobilized red-fluorescent beads (∅ 200 nm) and additional Alexa Fluor 647 dye molecules.
Scattering of the excitation laser beam and the excited fluorescence created a substantial fluorescence background outside of the excitation volume. This grossly diminished contrast and sectioning quality. Next, we acquired images of the latter sample with a scanned beam, while the camera was operated in the synchronized rolling shutter mode, and compared the image contrast for various rolling shutter widths. Image contrast acquired with a rolling shutter with an extension equal to the beam waist diameter was almost doubled compared to the global shutter data (fig. 3C). The signal to noise ratio, measured as the ratio of the intensity amplitude divided by the background noise was more than tripled in this confocal detection mode.
Certainly, it is more relevant to exa-mine the usefulness of the confocal line scanning mode using biological specimen. To this end we prepared one of the biological specimen of our current inte-rest, namely salivary glands of the dipteran Chironomus (C.) tentans. This is an iconic biological system that has been used for decades to study multiple aspects of the mRNA life cycle. These cells contain the well-known polytene chromosomes comprising thousands of parallel aligned DNA molecules. The transcription sites of the polytene chromosomes were stained by green immunofluorescence. This created a characteristic chromosomal banding pattern, which is due to differently active transcription sites in the chromosomes. In these cells the fluorescence light has to travel a distance of more than 100 μm through scattering tissue, and they were previously used by us to show the advantageous effects of light sheet microscopy . Here, the rolling shutter was adjusted to the width of the illuminating scanned beam (fig. 2C). Like in the model system, the confocal detection mode yielded images that had a significantly higher contrast and an improved resolution compared to the global shutter data acquisition (fig. 4).
We built a scanned light sheet fluorescence microscope, in which illumination beam and detection slit as realized by the rolling shutter of a sCMOS were synchronized. This configuration lets the rolling shutter act as a confocal slit mask to reduce the detection of scattered light. A clear-cut increase in signal to noise ratio and contrast could be demonstrated in test samples and in intact tissue. In particular, this approach is valuable for augmenting the imaging quality of LSFMs deep inside large samples.
The authors thank W. Wendler and H.P. Königshoven for technical support, Dr. H. Spieker from LaVision BioTec and J.-H. Spille for constructive discussions, and Hamamatsu Photonics Deutschland GmbH for making the Orca Flash 2.8 and 4.0 available. TK is a fellow of the German National Academic Foundation.
 Siedentopf H. et al.: Ann. Phys. 10, 1-39 (1903)
 Voie A.H. et al.: J. Microsc. 170, 229-236 (1993)
 Reynaud E.G. et al.: HFSP J 2, 266-275 (2008)
 Huisken J. et al.: Science 305, 1007-1009 (2004)
 Keller P. J. and Stelzer E. H. K.: Curr. Opin. Neurobiol. 18, 624-632 (2008)
 Baumgart E. and Kubitscheck U.: Opt. Expr. 20, 21805-21814 (2012)
 Spiecker H. K. A.: PCT Patent 2011/120629 (2011)
 Ritter J.G. et al.: PLoS ONE 5, e11639 (2010)
Eugen Baumgart (corresponding author via email request)
Prof. Dr. Ulrich Kubitscheck
Rheinische Friedrich-Wilhelms Universität Bonn
Institut für Physikalische und Theoretische Chemie