The modern world of biological light microscopy is a demanding one, increasingly sophisticated research applications require instrumentation capable of keeping pace.  Biological imaging has been moving consistently towards experimentation utilizing increasingly physiologically relevant systems for many years, and new technologies are making such experimental approaches increasingly feasible.  Such systems, including whole model organisms, tissue explants, and three-dimensional (3D) cell cultures, must furthermore be imaged in a manner minimizing perturbations in order to maintain their physiological integrity as experimental models.  Light-induced photodamage and phototoxicity have long been problematic in the field of biological imaging, and can have quite a dramatic effect on health and function at all biological levels of organization.  Imaging such large and sensitive specimens thus necessitates an efficient approach to 3D imaging that minimizes exposure of the sample to light. 

Light Sheet Fluorescence Microscopy (LSFM) is a general name for a constantly growing family of planar illumination techniques that have revolutionized how optical imaging of biological specimens can be performed.  Fundamentally, LSFM techniques are made possible by decoupling the illumination and detection optical pathways, allowing for novel illumination strategies that optimize the photon collecting efficiency of the instrument, this general concept is illustrated by Figure 1 above.  A light sheet is formed by laser light either shaped into a hyperbolic ‘sheet’ of illumination, or approximated by scanning a weakly focused low numerical aperture beam in the lateral directions, analogous to traditional laser scanning microscopies.  Importantly, detection is performed along a different axis than illumination, most often in an orthogonal direction in order to maximize detection efficiency by minimizing fluorescence from out-of-focus features. 

With standard episcopic approaches (i.e. confocal and widefield), the cone of illumination and detection is stretched in the axial (z) directions, unnecessarily exciting significant out-of-focus fluorescence and degrading the quality of the in-focus signal, as illustrated by Figure 2 below.  With LSFM, lower illumination intensities can be used in conjunction with planar (or otherwise structured) illumination to provide improved signal to noise (SNR) with minimal sample exposure, enabling imaging with high frame rates and over long periods of time.  Scanning the light sheet or the specimen in the axial directions allows for rapid and minimally invasive 3D imaging.

Axial confinement of illumination results in dramatically reduced photobleaching and phototoxicity, allowing for long-term imaging over many days, or even weeks.  However, LSFM can be notoriously difficult to implement, usually requiring two or more objectives, non-standard sample preparation protocols, difficult alignment procedures, and a well planned workflow for dealing with the massive amounts of data generated by such techniques (often in the terabyte regime).  Fields that benefit substantially from LSFM include developmental biology/embryology, neurobiology, drug discovery, plant biology, and more.  This text will review the fundamentals of LSFM (including a brief history), explore its various implementations and applications, and delve into basic aspects of sample preparation.

Historical Perspective

LSFM is rooted in a technique termed ‘ultramicroscopy’ pioneered in 1902 by Richard Adolf Zsigmondy, an organic chemist and experimental physicist, and Henry Siedentopf, an optical physicist.  The original ultramicroscope focused sunlight for side-on illumination of colloidal gold solutions, with the resultant scattering detected orthogonally to the plane of illumination.  This approach significantly increased the SNR compared to existing techniques, making possible the observation of single gold molecules.  The French scientist Jean Perrin would further Zsigmondy’s work by applying ultramicroscopy to plot particle movement, providing invaluable evidence for Brownian motion.  Zsigmondy was later awarded the Nobel Prize in Chemistry for his work with ultramicroscopy and colloidal solutions.  An illustration of Zsigmondy’s original ultramicroscope is provided in Figure 3.  For many years this method was largely forgotten, with planar illumination seeing some minor use in the field of photomacrography in the 1960s, as popularized by the ‘Dynaphot’ microscope.

The first combination of planar illumination with fluorescence microscopy came in 1993 when Arne Voie and David Burns of Francis Spelman’s laboratory at the University of Washington pioneered Orthogonal-Plane Fluorescence Optical Sectioning (OPFOS) microscopy, an LSFM they applied towards imaging and mapping the guinea pig inner ear cochlea.  One of the challenges Voie faced was the opacity of the calcium-rich bone.  Calcium atoms strongly scatter light, making optical imaging through bone nearly impossible, but can be removed using a 10% (w/v) EDTA solution in water.  Next their group applied an optical clearing solution, a 100-year-old formulation known as Spalteholz solution, consisting of a mix of oils for approximating the refractive index of protein. 

The images obtained by Voie and colleagues were comparable to those obtained using high-sensitivity tomographic methods, such as X-ray micro computed tomography (µCT), but with a superior ability to resolve soft tissue structures.  This is important as Voie and colleagues were attempting to correlate hair cell morphology with hearing loss, and so needed to be able to image tissue with high fidelity.  Correlative µCT and OPFOS would later be realized.  Note that this and later work by the Spelman group would treat OPFOS as a tomographic technique, not necessarily one for resolving sub-cellular details, as they demonstrated 10 μm lateral and 26 μm axial resolution with a very wide field of view (FOV) of 1.5 millimeters (mm) x 1.5 mm.  However, despite the success of OPFOS, LSFM would remain a relatively obscure technique for several years.

In 1994 Ernst H. K. Stelzer and Steffen Lindek of the European Molecular Biology Laboratory (EMBL) published a paper describing a new technique: confocal theta microscopy.  This technique uses orthogonal illumination and detection optics, similar to modern LSFM systems, but as part of a point-scanning confocal microscope.  The illumination and detection point-spread functions (PSF) only overlap at their common focus, resulting in 3-4x better axial resolution compared to conventional confocal imaging methods.

In 2002 LSFM gained renewed interest when Eran Fuchs and colleagues applied ‘Thin Light Sheet Microscopy’ (TLSM) towards imaging unperturbed aquatic microenvironments, specifically seawater samples.  Though ‘thin’ is used to describe the light sheet, the dimensions are similar to those originally used for OPFOS, a 23 μm beam waist thickness and 1 mm x 1 mm FOV.  One of the major reasons their group chose TSLM is because it allowed them to counterstain bacteria with SYBR Green I without subsequently filtering out the bacteria from the seawater sample to remove the free dye.  TLSM proved sensitive enough to image labeled bacteria with minimal background from the dye in solution.  Though confocal microscopy could have provided optical sectioning and reduced background, TLSM does so with superior acquisition speed and significantly higher SNR, important for capturing the dynamics of fast-moving microbes.

Finally, in 2004, the Stelzer group introduced a variation of the confocal theta technique using planar light sheets and widefield (rather than confocal) detection, publishing their landmark paper introducing Selective Plane Illumination Microscopy (SPIM).  Though SPIM featured relatively minor technical differences compared to its predecessors, the real break-through was its application to in vivo imaging of transgenic GFP-expressing model organisms, including GFP-labeled muscle in the naturally-transparent Medaka fish Oryzias latipes, and embryogenesis of the common fruit fly Drosophila melanogaster using GFP-moesin (a plasma membrane marker).  To this day one of the most popular applications of LSFM is towards fluorescence imaging of embryogenesis.  The introduction of SPIM signaled a very important step in the maturation of LSFM as the years following its release would see an explosion of new techniques and improvements, which we will explore in detail.

Light Sheets

Light Sheet Properties

Perhaps the most important component of any LSFM is the light sheet itself, which can take the form of either a static planar sheet, as used by the original SPIM technique, or as a scanned beam approximating a light sheet over time.  Each method has its respective advantages and disadvantages, but first it is important to understand some of the fundamental properties of light sheets.  Figure 4 illustrates different types of light sheets and important parameters discussed herein.

First, it is impossible to create a perfectly planar light sheet, as one can only be approximated, and only over a given range.  Axial resolution is ultimately limited by the thickness of the light sheet and the NA of the detection objective.  Both static planar and scanned light sheets have a hyperbolic profile in the xz plane (Figure 4), with the beam waist (w₀) specifying the thickness of the most tightly focused mid-point of the illumination profile, and calculated using the following equation:

Where w₀ is the beam waist thickness, f is the focal length of the illumination optics, λ is the wavelength of the illumination light, and D_lens is the diaphragm of the illumination optics.  Of particular concern is the confocal parameter (b) of a light sheet, which defines the distance through which the sheet can be said to provide near-homogenous planar illumination, with thickness  at the beam waist and expanding to  at either edge of the confocal parameter.  The distance from the center to where the beam waist has thickened by a factor of $\sqrt{2}$ in one direction is the Rayleigh range (X_R), which is by definition half the confocal parameter.  The confocal parameter thus defines the maximum usable lateral extent of the light sheet in the direction of propagation.  Increasing the confocal parameter by creating a larger light sheet directly results in a thicker beam waist (limiting axial resolution), as described by the equation below relating the confocal parameter and beam waist thickness: 

Where b is the confocal parameter, X_R is the Raleigh range, w₀ is the beam waist thickness, and λ is the illumination wavelength.  For example, if a light sheet thickness of ~2 micrometers is chosen and imaging is performed with 561 nm green light, the confocal parameter will be approximately 22.4 micrometers (μm), which is sufficient for single cell imaging, but not for imaging larger specimens.  For comparison, a light sheet with a beam waist w₀ = 10 μm has a confocal parameter b with a value of over 1 millimeter (mm).  It should be noted that these equations are for characterizing light sheets propagating in air (refractive index n = 1.0), in a medium of refractive index n, the confocal parameter becomes:

Where n is the refractive index of the medium and b_n denotes the confocal parameter in medium with refractive index n. Thus it is possible to extend light sheets by allowing them to propagate in higher refractive index materials, such as water (n = 1.33).  Research attention is being given to specialized approaches for extending the confocal parameter of light sheets without thickening the beam waist using specialized techniques that will be discussed in more detail.  Bear in mind that the provided equations can be used to characterize not just static planar light sheets, but ‘virtual’ light sheets created by a scanned beam.

Planar Light Sheets

Older LSFM techniques, such as OPFOS, included a single cylindrical lens to craft and project the light sheet into the sample. Like spherical lenses, cylindrical lenses are used to expand or converge light, but only along a single axis.  More sophisticated iterations of the technique use a cylindrical lens in combination with an illumination objective to refine the optical properties of the light sheet.

Scanned Light Sheets

One solution to decoupling the interdependence of the confocal parameter and beam waist thickness involves scanning a thin light sheet in the lateral directions and subsequently stitching image sections together to create a larger composite image than dictated by the confocal parameter alone.  This approach has been utilized by High Resolution OPFOS (HR-OPFOS) and scanning Thin Sheet Laser Imaging Microscopy (sTSLIM), and applied towards imaging the mouse cerebellum and inner ear cochlea.  The downside of this approach is that it requires lateral as well as axial scanning of planar light sheets, resulting in longer acquisition times and greater light exposure - partially negating one of the greatest advantages of LSFM, the selective illumination of only those features to be imaged.  Additionally, image registration and stitching can be problematic, though registration algorithms have improved in recent years and allow for the use of fiducial markers for registering the locations of sample features.

Typically the specimen is scanned axially through the light sheet to generate a z-series, though objectives can be translated relative to a stationary sample.  More recently many research groups have opted for using standard Gaussian laser beams rather than a light sheet, scanning the beam laterally to imitate a planar light sheet.  This is followed by axial translation of the beam, using an f-theta lens and appropriate scanning mirrors, allowing for the acquisition of a complete z-series without moving the objective or sample.  This approach is referred to as Digital Scanned Laser Light Sheet Microscopy (DSLM) and was developed by Philipp Keller and colleagues in 2008. 

Multiphoton Light Sheets

Both planar and digitally scanned light sheets can be used in conjunction with multiphoton excitation.  Two-photon excitation (2PE) occurs when a fluorophore absorbs two photons in a single quantum event where the combined energies are similar to that needed to excite the fluorophore to an excited state using a single photon.  2PE is an extremely rare process in the natural world, only occurring at extremely high photon densities, as realized at the focal point of an extremely high-power femtosecond-pulsed laser.    The excitation wavelength for 2PE should be approximately twice that used in the typical one-photon case, though this is much more of a guideline than a rule and published two-photon absorption spectra should be consulted.

The imaging field is somewhat reduced in size with multiphoton excitation compared to conventional single photon approaches due to the reduced focal area, however the long-wavelength infrared light used for 2PE is superior for sample penetration in scattering media.  Indeed, the optical window of biology exists between approximately 650 nanometers (nm) and 1200 nm, where autofluorescence is minimal and absorption by hemoglobin, water, and protein is low.  This is especially helpful for LSFM experiments since specimens are often several millimeters thick and more highly scattering and/or absorbing compared to thin samples.  In summary, the primary advantages offered by multiphoton excitation are decreased scattering/absorption and consequently deeper sample penetration; products of the longer-wavelength infrared light used by this technique.

Scanned Bessel Beam and Optical Lattice Light Sheets

The lab of Alexander Rohrbach first demonstrated the use of self-reconstructing Bessel beams to create a digitally scanned light sheet.  Bessel beams are generated via projection of an annular pattern about the periphery of the rear pupil of the illumination objective, or by shaping the illumination light with an axicon.  In the ideal case these beams can be thought of as non-diffractive: if obstructed a Bessel beam will reform downstream of the obstructing object, resulting in the commonly used ‘self-healing’ and ‘self-reconstructing’ descriptors. 

A Bessel beam core can be made much thinner than a standard Gaussian beam, allowing for thinner scanned light sheets.  The most significant problem associated with Bessel beam illumination is that a large proportion of the beam energy resides in side-lobes, similar to, but more pronounced, than seen in a typical Airy diffraction pattern.  These sidelobes are actually vital to the self-healing properties of the beam, but effectively thicken the beam waist.  One solution to this problem is the combination of Bessel beam illumination and two-photon excitation (2PE), which has been demonstrated for shaping very thin light sheets (less than half a micrometer).  The illumination intensity is only high enough in the core of the Bessel beam to achieve 2PE, effectively eliminating out-of-focus excitation by the side-lobes.  Using this approach the laboratory of Eric Betzig has demonstrated 300 nm isotropic resolution in live cells at speeds of approximately 200 image planes per second.  Bessel beams have also been used in combination with incoherent structured illumination microscopy (SIM) methodologies, allowing one to computationally discriminate and reject the low spatial frequency information arising from out-of-focus fluorescing structures.  The benefits of SIM and 2PE combined with Bessel beam illumination are similar.

Expanding upon this idea of applying Bessel beams to LSFM, the Betzig lab introduced a powerful new LSFM:  lattice light sheet microscopy.  Instead of using scanned Bessel beams, this method uses similar techniques to project novel optical lattice patterns into the sample plane.  An optical lattice is a periodic interference pattern created by the superposition of two or more plane waves and having 2D or 3D structure.  However, unlike Bessel beams, which are created by confining illumination to a very thin region in a complete ring about the rear pupil plane of the illumination objective, an optical lattice is created by illumination of discrete points about the annulus.  Optical lattice illumination allows the user to select a pattern optimized for axial resolution or illumination confinement. 

Lattice light sheet microscopy has been demonstrated in several different ‘modes’.  The ‘standard’ mode involves quickly dithering the optical lattice pattern using a galvanometer to provide temporally uniform illumination of each lateral section, resulting in distinct time-averaged illumination patterns.  The structured illumination (SI) mode provides approximately 1.3-1.5x greater spatial resolution than the dithered, but with about 7.5x worse temporal resolution.

The utility of lattice light sheet imaging has been further demonstrated for single molecule tracking and super-resolution imaging experiments.  Small and tightly focused light sheets generally have a beam waist of about 4-5 μm, several times thicker than the typical depth of field (DOF) of high NA objectives.  In contrast, dithered optical lattices have an effective beam waist closer to 1 μm, more closely matching the DOF of high NA objectives and superior for single molecule super-resolution imaging.  Confined excitation of single emitters within the objective focus allows a greater proportion of single molecule emission events to be detected. 

Scanned Airy Beams

Similar to Bessel beams, Airy beams are propagation-invariant and ‘self-healing’. Airy beams are formed by modulating a typical Gaussian beam at the back aperture of the objective using a spatial light modulator, resulting in the Airy beam’s distinctive asymmetric transverse structure (Figure 4). Interestingly, Airy beams seem to extend even further than Bessel beams through thick and scattering media.  An Airy light sheet is formed similar to other DSLMs with the beam dithered along the y-axis to create a virtual sheet of illumination.

LSFM Imaging Modalities

Multiview and Multidirectional Imaging

One of the most common problems with LSFMs is the presence of striping artifacts.  These artifacts arise as a result of single-direction illumination; sample features that scatter, absorb, or otherwise perturb the incident beam upstream result in weaker downstream illumination, visible as dark “stripes”.  These striping artifacts are illustrated by Figure 5(b,d), where a hypothetical sample containing optically opaque structural elements (orange circles) is subjected to one-sided illumination, with the resulting shadows/stripes clearly visible in the figure.

An early solution to the striping problem was the application of multiple-view (multiview) imaging of the sample, realized by serial xy rotation and imaging of the specimen at each z plane, followed by fusion of the images from each view to create a high quality composite. Usually, several z-stacks are acquired sequentially, each stack from a single view. For example, the first z-series is acquired, the sample is rotated by a pre-defined value, a second z-series is acquired, and so forth until a satisfactory number of views are obtained, usually 4-6. Once a z-stack is acquired at each view, composite optical sections are created using specialized fusion and deconvolution algorithms to combine the images, producing a high-resolution image with low deviations in image quality throughout the sample volume. The benefits of multiview imaging are explored further in Figure 5.

Multiview imaging allows one to simulate homogenous illumination of each z plane. Though all sections of the specimen are not entirely uniformly illuminated, especially with low view numbers and approaching the center of thick samples, image quality is greatly enhanced compared to single view methods. Alternatively, pivoting light sheets can be employed in a technique referred to as multidirectional SPIM (mSPIM). Light sheets propagating from the illumination objective(s) are pivoted in the detection focal plane over an angle of approximately 10^o using a resonant mirror operated at high frequency. Illumination from multiple angles is realized near-simultaneously from each illumination objective during the camera exposure, reducing illumination artifacts without compromising acquisition speed as with multiview imaging. However, this approach is not as effective as multiview SPIM in reducing artifacts since the range of illuminated angles is still relatively limited. Multiview and multidirectional imaging can be combined for increased sensitivity.

There are concerns regarding the multiview method, primarily because it increases acquisition time several fold. Rotating the specimen can be problematic due to rotation times typically in the tens of milliseconds. Long rotation times are necessary to prevent mechanical strain and movement of the specimen, which could compromise the experiment or result in ‘jerk’ artifacts where the sample appears to move drastically between image frames. Furthermore, each plane is exposed to illumination light several times longer than with comparable single-view LSFM experiments. This is due to the need to image features in each plane 4-6 times. Such increased acquisition times often preclude multiview imaging of living specimens. Thus the technique is more conducive to high-quality structural imaging of fixed samples instead.

A popular alternative and/or addition to multiview imaging is dual-sided light sheet illumination: a pair of opposed illumination objectives provide alternating (or simultaneous) light sheet illumination, penetrating from opposite sides of the specimen, with the detection objective(s) still placed orthogonally to both illumination objectives, and providing more homogenous illumination of sample features. This is a fast and elegant approach for reducing striping artifacts resulting from single-sided illumination. Also, if properly aligned, dual-sided illumination can be used to effectively double the confocal parameter, allowing the use of thin light sheets for imaging larger specimens at higher resolution but without the lateral scanning and stitching processes utilized by sTSLIM, HROPFOS, and similar LSFM tiling techniques.

As stated, DSLM and similar techniques simulate light sheets by line-scanning a typical diffraction-limited Gaussian-distributed collimated laser beam both laterally and axially through the sample. One of the primary advantages of this technique is that it provides a more homogenous light dosage than static planar light sheets, preferred for more quantitative imaging applications. Additionally, DSLM doesn’t require beam-shaping apertures, reducing optical aberrations and resulting in an illumination efficiency of approximately 90-95%, compared to about 5% for planar light sheets. This allows one to employ much cheaper and lower power laser light sources. Finally, DSLM and related approaches allow for novel illumination patterns via modulation of the illumination intensity with an Acousto-Optical Tunable Filter (AOTF).

Many recent standardized approaches for high-sensitivity LSFM involve the use of four total objectives: two for illumination and two for detection. A great benefit of this technique is that it doubles the amount of light collected in each plane. The standardized SiMView and MuVi-SPIM imaging platforms popularize this modality. More recently, the IsoView variation was introduced, whereby light sheet excitation and fluorescence detection is performed by four custom-made objectives sequentially, and each with their own sCMOS camera for detection, allowing for the collection of four views without physically rotating the specimen. An early technique, termed multiple imaging axis microscopy (MIAM) also used 4 objectives, but placed in a tetrahedral arrangement, and providing isotropic resolution, also without rotating the sample. However MIAM is difficult to align and implement.

Guides for assembling high-quality LSFM instruments, including SiMView, MuVi-SPIM, IsoView, and the open-source OpenSPIM project serve to help standardize instrumentation. Additionally, SPIM software and algorithms are made available from these sources. In time LSFM will most likely reach the same degree of commercial availability and standardization enjoyed by other microscopies.

iSPIM, diSPIM, and triple view SPIM

The introduction of inverted SPIM (iSPIM) techniques by the laboratory of Hari Shroff represented a powerful new approach in adapting LSFM for widespread use. iSPIM consists of a ‘head’ on which both the illumination and detection objectives are attached. The ‘overhead’ orientation of the illumination and detection objectives allows for the use of more conventional sample preparation techniques, including adherent cells grown on coverslips. This is in contrast to most LSFMs where the sample must be embedded in a gel and suspended between the objectives in a non-traditional manner. The head also contains galvanometer scanners for translation of the light sheet through the sample, and objective z piezos for translating the detection objective so as to retain the light sheet in the plane of focus. The head provides an incredibly flexible solution that can be mounted on virtually any inverted microscope stand, allowing for relatively simple integration on existing platforms.

The more advanced dual inverted SPIM (diSPIM) variant is a clever adaptation of the original iSPIM design, but illumination and detection alternates between both objectives, with subsequent image registration and deconvolution applied to fuse the complimentary views and providing 330 x 330 x 330 nm isotropic resolution. As the sample is kept stationary, acquisition speeds of 200 images/second have been reported, allowing researchers to acquire a full z-stack in less than a second, perfect for capturing 3D dynamics in living systems. Objective selection for diSPIM (and most LSFMs) is limited by steric hindrance – the close proximity of the illumination and detection objectives limits the application of bulky, high NA objectives with correspondingly short working distances. This is problematic for diSPIM as two identical objectives are used, in contrast to using lower magnification and NA illumination objectives with longer working distances combined with larger, higher NA detection objectives, as with other LSFMs. Currently, diSPIM and related techniques are most popularly performed using a pair of Nikon Apo 40x 0.8 NA water dipping objectives (working distance = 3.5 mm), as illustrated in Figure 6.

Most recently, the diSPIM technique was adapted to utilize an objective mounted in the conventional nosepiece position of the inverted microscope stand on which the diSPIM head is mounted, directly beneath the coverslip. Previously, low magnification and NA objectives in this position were utilized to assist with sample placement, but not for data collection. However, the latest iteration of the iSPIM approach – triple view SPIM – utilizes a high NA objective to provide a third view, capturing even more of the solid angle of fluorescence emission. Triple view SPIM has been demonstrated with a 1.2 NA water immersion objective in the nosepiece position, improving the optical resolution of this setup to 235 x 235 x 324 nm, and with higher signal levels. However, as the objective in the nosepiece position is at a 45^o angle to the light sheet (from either objective on the diSPIM head), it must be translated along the z-axis to capture the in-focus fluorescence generated by the light sheet along the entire angle. This is optimized by synchronizing the z scanning with the rolling shutter of a sCMOS camera in the nosepiece objective imaging path.

Single-Objective LSFMs

In addition to iSPIM and related methods, there are several modular single-objective LSFM designs intended to make the technique more accessible to non-specialists, and/or for integration on existing microscope platforms. One design, termed objective coupled planar illumination (OCPI) is a modular unit where the illumination and detection objectives are rigidly coupled, mounted together on a single arm and permanently aligned.

Many LSFM approaches utilize alternative illumination strategies in order to facilitate the use of high NA oil-immersion objectives. Several use oblique illumination, similar to that realized when operating a typical total internal reflection fluorescence (TIRF) microscope with sub-critical illumination. These techniques, variably referred to as highly inclined and laminated optical sheet microscopy (HILO) and oblique plane microscopy (OPM), provide many of the same benefits as more standard light sheet approaches, but with smaller fields of view. Oblique illumination provides a small, stationary light sheet through which the sample can be scanned using a typical piezo z-stage for volumetric imaging. However, this requires that at some point along the optical train the angled image field be corrected for (due to the oblique illumination). This is accomplished by projecting the image for detection by a second objective, which is placed at an angle correcting for the oblique geometry, as illustrated by Figure 7 below. This can be difficult to align properly; the objective arrangement must be near perfect in order to simulate a flat field.

The OCPI approach has been demonstrated using a 60x NA=1.35 objective. However, one of the drawbacks of this approach is limited field size, precluding its use for large samples (e.g. model embryos). Also, the full NA of the imaging objective cannot be utilized as lower NA objectives are used when reimaging the sample plane, limiting this technique to NA < 1.0 in practice. These strategies are premised upon streamlining LSFM instrumentation more than resolution improvement, helping to ease the design requirements for researchers seeking to construct their own instruments for light sheet analysis of smaller biological samples, such as single cells.

A second strategy involves using an angled mirror or specialized prism coupled to the objective for projecting the light sheet directly into the sample medium.  One approach utilizes a disposable mirror on an atomic force microscopy-type cantilever, positioned immediately adjacent to the detection objective to reflect the light sheet into the sample.   Z imaging is achieved via vertical translation of the sample, with illumination remaining stationary so as not to disturb the system alignment.  As with oblique illumination strategies, this method has limited field size compared to other LSFMs, typically about a 1 um thick beam waist and a 10 um confocal parameter.  Also, there exist concerns with the difficulty of implementation, as well as with the overall optical stability of such a setup.  A more recent variation of this technique sought to address these issues by replacing the mirror with a Pellin-Broca prism, on which the sample is directly placed.  Light is directed through the angled prism, and emerges as a light sheet in the sample.  Though the technique has only been demonstrated on the apical membrane of adherent cells, the approach is promising, offering high enough signal-to-noise and resolution to perform single particle tracking experiments. 

Super-resolution with LSFM

There are a variety of alternative imaging modalities for realizing super-resolution imaging beyond the diffraction limit with LSFM.  Single molecule tracking and localization is becoming a fairly widespread technique for use in conjunction with LSFM.  One approach, termed individual molecule localization SPIM (IML-SPIM) combines light sheet illumination with single molecule localization microscopy (SMLM).  A high-powered light sheet is used to induce photoswitching of suitable probes, individual emission events are well separated spatiotemporally, allowing for the identification and sub-diffraction limited localization of individual fluorophores.  Given suitable acquisition times, a super-resolution reconstruction can be created by mapping all the localizations in a single plane. 

By combining LSFM with SMLM, super-resolution optical sections can be obtained from samples with thickness orders of magnitude larger than typically realized using non-LSFM instruments (e.g. TIRF).  Specifically, IML-SPIM has been demonstrated with nuclear-labeled cellular spheroids in excess of 100 um thick, localizing histone proteins labeled with the photoactivatable fluorescent protein PAmCherry1.  Sub-diffraction axial localization of individual fluorophores within the light sheet is achieved by introducing astigmatism into the emission point-spread function via a weak cylindrical lens.  Localizations appear stretched in the x or y direction, dependent upon the fine axial location.  Localization precisions of 35 nm in the lateral directions and 65 nm in the axial have been achieved.  Two-photon activation of SMLM probes in an IML-SPIM configuration further confines fluorescence in the axial directions, making single molecule identification easier and mitigating the deleterious effects of scattering media.

A more commonly employed resolution-enhancing approach involves combining LSFM with incoherent structured illumination microscopy (SIM), referred to as SPIM-SI.  SIM relies upon symmetric, periodic illumination patterns projected into the image space in order to discern in-focus from out-of-focus fluorescence.  Out-of-focus features are not affected by the spatial modulation of the pattern, allowing their contribution to be computationally identified and rejected.  In order to create a SIM image, a sinusoidally varying stripe pattern of alternating maxima and minima is created in the specimen, generally by modulating the illumination intensity with an AOTF to create virtually structured illumination.  The stripe pattern is imaged at three different phases of the same orientation.  This approach should not be confused with super-resolution structured illumination, which uses beam interference for generation of a diffraction-limited pattern.

A related technique, known as HiLo microscopy (not to be confused with HILO: highly inclined and laminated optical sheet microscopy) can also be used to effectively discriminate out-of-focus information.  HiLo is similar to SIM, but only requires two images:  one structured illumination image and one conventional widefield image.  The structured illumination image is used to identify low frequency out-of-focus information, while the conventional widefield image contains high spatial frequency details that may have been lost via application of the illumination pattern. 

It is also possible to combine LSFM with super-resolution stimulated emission depletion (STED) microscopy.  STED is a confocal-type point-scanning technique reliant upon a high power ‘depletion laser’ to drive excited fluorophores to a ground state without fluorescence emission by exploiting the process of stimulated emission.  The depletion laser is ‘wrapped around’ the conventional excitation beam, stimulating emission from excited fluorophores residing away from the center of the excitation beam in a non-linear fashion.  Super-resolution is thus achieved by spatially confining the area where probes are allowed to fluoresce to a sub-diffraction limited area, with the size of this area inversely proportional to the power of the depletion laser.  As applied to LSFM, the depletion laser is shaped into a double sheet, sandwiching the excitation light sheet, and effectively shrinking the beam waist.  The result is a reported ~30% increase in both axial and lateral resolution compared to standard LSFM imaging modalities.  Note that the related RESOLFT technique has similarly been applied to LSFM to improve axial resolution.

Other LSFM Implementations

Some advanced additions to LSFM include the implementation of adaptive optics (AO).  AO reduces wavefront distortions resulting from normal sample-induced aberrations in optical microscopy by using wavefront sensors combined with components that compensate for distortion, most often deformable mirrors (DMs).  This allows for wavefront correction in response to sample aberrations, increasing resolution and accounting for inhomogeneities in the imaging medium, a common problem for LSFM, especially with deeper fields of view that are more prone to scattering and absorption artifacts.  AO can be implemented using several approaches, but is especially intriguing for LSFM, as it can technically be applied to both the illumination and detection pathways independently.  AO has been used in LSFM to allow for imaging through a glass capillary, compensating for distortion resulting from the changes in refractive index between the glass and aqueous sample medium.  A similar approach uses differential interference microscopy (DIC) to estimate the changes in refractive index through the samples, allowing for space-variant deconvolution.

Several LSFM implementations have developed that use a multimode optical fiber to deliver the light sheet in place of a typical illumination objective lens.  Such approaches have an incredibly small footprint, ideal for labs with limited space or who are interested in the potential application of LSFM with an endoscope.

LSFM Hardware

Constructing an LSFM can be difficult, in large part because it differs so greatly from traditional microscope designs. Figure 8 provides an illustration of the optical train of a typical open-source dual illumination SPIM-type instrument.

Perhaps the most vital component of any imaging system is the camera. The use of charge-coupled devices (CCDs) is considered the industry standard for most biological imaging applications. Additionally, the use of high-sensitivity (and expensive) electron-multiplying CCDs (EMCCDs) is popular for many advanced imaging applications requiring single molecule level sensitivity. However, the cameras of choice for most modern LSFMs are scientific complementary metal-oxide-superconductors (sCMOS). The primary advantages of sCMOS cameras are high imaging speeds (as fast as 100 Hz for a full field of view), moderate cost, and much larger chip sizes compared to CCDs and EMCCDs, great for wide field of view imaging. Fast image acquisition rates are vital for realizing the full potential of LSFM, especially for in vivo volumetric imaging where important events occur in three dimensions very quickly. For very large specimens, the chip size can make the difference between being able to image the entirety of the sample and having to resort to an alternative strategy, such as image stitching. An increasingly common approach uses the rolling shutter of a sCMOS camera to act as a confocal slit aperture for a line-scanning DSLM, providing axial resolution improvement similar to existing line-scanning confocal systems. Note that it is also possible to simply include a confocal slit aperture to reject out-of-focus light, rather than using the rolling shutter.

Objectives

When choosing an objective lens for LSFM a number of different factors must be taken into consideration. Foremost among these are the characteristics of the sample and the type of light sheet microscope. The sample volume determines the optimal field of view, working distance, and depth of field. Additionally, the optical design of an LSFM places many inherent spatial constraints on objective selection that the user must be cognizant of when designing an experiment.

In most cases, short working distance objectives are prone to steric hindrance resulting from having to be placed very close to the second, perpendicular objective lens. Most high NA immersion objectives (NA > ~1.1) fall into this category. Additionally, oil immersion usually isn’t of great utility for LSFM as most specimens are large and/or mounted in an aqueous medium, such as an agarose gel or the culture medium, and thus optimally imaged using either a water objective. Though refractive index mismatch between the objective and the sample medium usually isn’t problematic with very thin samples (e.g. adherent cells on a coverslip), with samples displaying non-negligible 3D structure spherical aberration quickly becomes a major issue.

Generally, detection objectives for LSFM are low-to-mid NA (~0.2-1.1) dry or water immersion/dipping objectives with long working distances and steep approach angles (objectives for electrophysiology are usually a good place to start). The ideal specifications depend on the specimen of interest. High-resolution confocal and related techniques often use oil-immersion objectives with high NA (~1.3 - 1.49), enabling inherently greater lateral resolution. However, up to NA = 0.8, LSFM provides superior optical resolution in the axial directions compared to a confocal microscope using the same NA objective. Ultimately, LSFMs are designed to prioritize acquisition speed and minimize sample exposure rather than maximize optical resolution, and thus cannot be judged based off objective NA alone.

The illumination optics generally include a low NA and low magnification illumination objective, generally 2-10x magnification with an NA = 0.1-0.3. Other beam shaping optics for shaping the light sheet to the proper dimensions are often included, varying with system design. Such optics include a cylindrical lens, slit apertures, and beam-expanding optics, as illustrated by Figure 8. Popular Nikon objectives for LSFM include the CFI Apo 40XW NIR (NA = 0.8), CFI75 LWD 16XW (NA = 0.8), CFI75 Apo LWD 25XW (NA = 1.1), and the CFI Plan Fluor 10XW (NA = 0.3).

Light Sources

The light source for LSFM is almost always a laser, either shaped into the planar light sheet or a scanned Gaussian, Bessel, or Airy beam. With planar light sheets, only about 5% of the laser intensity is conserved due to beam shaping optics, compared to about 95% intensity in a line-scanning configuration. Special considerations for advanced techniques, such as combining LSFM and with two-photon excitation (2PE), require more complex instrumentation. As with confocal microscopy, a high quality laser light source is a must.

Though wide-spectrum mercury and halide arc lamps can, in principal, be shaped in to planar light sheets, the resulting intensity is generally insufficient. Light emitting diodes (LEDs) provide similarly insufficient illumination intensity, though it is common for an LSFM to include a red LED for transmitted light illumination, providing a simple and relatively non-invasive method for positioning the specimen prior to imaging. We expect this to change though as increasingly powerful and cost-effective LEDs continue to become available.

Data Analysis and Storage

A vitally important aspect to consider is data analysis. LSFM is capable of producing massive amounts of data, often many terabytes from a single experiment. To this end, it is necessary to establish a data management pipeline before bringing an LSFM online. A pipeline to an analysis computer or cluster and subsequent archiving workflow is paramount, and practices such as image compression are a must. Many institutions enjoy access to super-computing resources, speeding up analysis. The continuing advancements in cloud computing services may also prove beneficial for data storage. Computationally efficient and well validated open source image registration and analysis algorithms are available for real-time image analysis/processing.

Sample Preparation for LSFM

As stated, the primary advantage of LSFM is for imaging biological specimens larger than typically considered feasible using more traditional fluorescence techniques, such as confocal microscopy. LSFM is also advantageous for fast 3D imaging of adherent cell cultures with drastically reduced light dosage. Confocal point-scanning techniques cannot penetrate more than approximately 700 micrometers into the sample (along the optical axis). This is a generous limit; usually image degradation is readily observable at depths of less than 100 micrometers. LSFM is ultimately capable of imaging at depths in the tens of millimeters, rivaled only by multiphoton microscopy in this regard. Though multiphoton is still preferable for intravital imaging of very large specimens (e.g. mature rodents), LSFM is superior for visualizing embryogenesis and early development of most model organisms. Popular model organisms for LSFM include fruit flies, zebrafish, nematodes, and more.

Some of the most intriguing model systems for LSFM evaluation are cellular spheroids/cysts. These ~100 micrometer diameter hollow spheres mimic the structure and function of actual tissue to a much greater degree than typical two-dimensional cell cultures. Additionally, there exist significant deviations in the proteome, respiration modalities, and communication networks of 3D cultured spheroids compared to 2D flat cultures of the same cell lines. Many of the genes up-regulated in spheroids are also up-regulated in in vivo tumor growths. Spheroid cultures are especially well suited towards the study of epithelial development, with Madin-Darby canine kidney (MDCK) cells and mammary epithelial cells (MEC) serving as prime models. Unlike 2D cultured epithelia, 3D spheroids are polarized, similar to in vivo epithelial tissue. 3D cultures also tend to be homeostatic, regulating cell migration, apoptosis, proliferation, and differentiation. It is also possible to create heterotypic spheroids, containing several cell types as found in actual tissue, e.g. containing both an epithelial and mesenchymal/endothelial layer. Cell spheroids thus play an important role in cancer research and drug discovery, providing a superior model system for treatment.

The final type of specimen that has repeatedly been demonstrated as suitable for LSFM is excised boney organ structures, as described for the OPFOS technique. Specifically, much research has been performed on characterizing the inner ear components using LSFM, and sometimes in concert with more traditional tomographic techniques such as µCT. This approach has also been used to map neurons in fully-grown mouse and zebrafish brains.

Sample Preparation

Sample preparation for LSFM is often difficult. On one hand sample preparation can be relatively simple for techniques such as iSPIM and Lattice Light Sheet, where the samples are often adherent cells grown on coverslips. Due to its ubiquity, preparation of 2D adherent cell cultures will not be discussed. On the other hand, SPIM-type imaging of larger and more complex model systems requires non-standard sample preparation techniques. Such specimens include cultured three-dimensional cellular spheroids/cysts, smaller model organisms (e.g. Drosophila, zebrafish, medaka, and Xenopus embryos), and explanted tissues/organs. There are two primary approaches to sample preparation for LSFM. The first uses specialized optical clearing reagents and calcium chelators to render normally opaque structures optically penetrable, allowing one to image boney structures such as the inner ear cochlea unimpeded. The second approach uses un-cleared specimens, usually model embryos or spheroids no more than about a cubic millimeter in volume, mounted in a low-melting point agarose cylinder for insertion into a specialized imaging chamber.

When Arne Voie and colleagues first introduced OPFOS in 1993, they utilized Spalteholz solution for clearing, which is composed of 5 parts methyl salicylate to 3 parts benzyl benzoate and having a refractive index of approximately 1.47, similar to that of most proteins. Thus the entire imaging volume is refractive index matched, making it optically transparent by dramatically reducing scattering artifacts. However, Spalteholz clearing does not account for bone or other calcified tissues. Remember that calcium strongly scatters light. Thus preliminary application of a calcium chelator such as 10% (w/v) dehydrate ethylenediaminetetraacetic acid (EDTA) in dH₂O is required before submersion in the Spalteholz solution. Keep in mind that such harsh sample preparation protocols are for fixed sample imaging. In the case of Voie and colleagues, the excised, de-calcified, and optically cleared cochlea were imaged tomographically and subjected to finite element analysis for delineating relatively large features. Several alternatives now exist to Spalteholz solution that can be explored, such as CLARITY, DISCO, and related methods. Furthermore, the use of cleared specimens provides a unique opportunity for using higher NA objectives, as many microscope manufacturers produce objectives featuring a refractive index adjustment collar specifically for matching the refractive index of the objective with that of the cleared specimen. Specific clearing protocols are well reviewed elsewhere.

As stated, a common method of LSFM sample preparation is sample embedding in an agarose cylinder, which acts as mechanically stabilized water. Alternatively, samples can be similarly embedded in gelrite, galactan, gelatin, agar, alginate, or carrageenan. Most commonly, a 0.8-1.0% (w/v) low melting point agarose in water or PBS is chosen. Importantly, this approach is compatible with living specimens, with in vivo imaging of many naturally transparent model organisms now commonplace. First, the gelling agent should be prepared and left to cool until just above the gelling temperature. This can be performed on a hot plate or heated water bath, in general it is best to make sure the agent is close to 37^oC immediately prior to use (physiological temperature for mammalian systems). At this point the sample is ready to be mounted, which should be performed quickly so as not to introduce bubbles into the gel and to avoid pre-mature gel polymerization. Once mounted, samples should be imaged immediately.

Embedding is most commonly achieved by shaping agarose into a cylindrical chamber from which the sample can be suspended. Most simply, the molten agarose is pulled into a syringe where the top has been removed. The sample can then be positioned within the gel, which is subsequently allowed to polymerize so that the cylinder can be pushed out using the plunger. Figure 9 illustrates this method, which can be adapted for smaller samples using a glass capillary (with plunger) instead of a syringe. In general, the sample width should be between 1/3-2/3 of the agarose cylinder diameter and positioned as close as possible to the surface facing the DO so as to minimize aberration. Also, embedded samples should not be imaged still inside of the capillary/syringe (if possible) due to the change in refractive index.

Two primary methods exist for agarose embedding.  The most straightforward is to mix the sample with the agarose directly prior to pumping into the syringe/capillary to polymerize.  This method works well for smaller samples, such as fluorescent beads, cellular spheroids/cysts, yeast, and even small embryos.  However, for larger samples such as developing model organisms, the embedding mold is first filled with agarose, then the sample carefully positioned within prior to polymerization.  This allows for finer control of sample alignment compared to the first method.  Similarly, for especially difficult samples, the agarose cylinder can be formed alone, a hole or notch cut into it, the sample placed into it, and fresh molten agarose poured in to cover the specimen.  Lastly, for samples that shouldn’t be embedded, a chamber can be constructed within the cylinder for holding PBS, growth medium etc. and subsequently sealed off using more agarose.  This internal chamber can be molded by attaching a solid cylinder to the tip of the plunger, allowing the agarose to polymerize around it.  Note that for this approach a higher concentration of agarose (~1.5%) should be used to ensure the mechanical stability of the thin chamber walls.

Alternative sample mounting methods are very common.  Samples can be hooked on a curved glass capillary, which can be rotated and/or translated as necessary for imaging.  Another common practice is to glue the sample to a rod/capillary or to leave it partially extruded from a syringe.  Specialized imaging chambers for LSFM are garnering more attention; one promising approach uses three-dimensional printers to create custom-designed holders out of choice materials.

LSFM Applications

As previously stated, LSFM is conducive to imaging large samples, up to tens of millimeters in length. Common samples include anatomical structures and explanted tissues, such as mice brains, mouse/guinea pig cochlea and middle/inner ear components, and more. Model organisms for LSFM include the common fruit fly Drosophila melanogaster, the zebrafish Danio rerio, medaka fish Oryzias latipes, the nematode Caenorhabditis elegans, the midge Chironomus tentans, the African clawed tadpole Xenopus laevis, and the small flowering plant Arabidopsis thaliana. These organisms are generally imaged during the embryonic stage, but whole adult organism imaging has been performed for many of the listed species.

Early implementations of LSFM, specifically OPFOS, focused on hearing research by imaging the structure of middle and inner ear components. Early studies utilized thick light sheets approximately 20-30 micrometers thick, allowing very large fields of view several millimeters in length. Excised, optically cleared, guinea pig cochlea and scala tympani were among the first specimens to be imaged using OPFOS. These images allowed for revolutionary 3D reconstruction and mapping of these structures. Later, guinea pig tympanic bulla would be imaged and modeled using a similar methodology, with a resolution of about 16 micrometers. The use of laterally scanned thin light sheets (HR-OPFOS) allowed for much greater optical resolution (~2.0 micrometers) over fields 30 mm in diameter or greater, yielding even better images of the cochlea and other ear structures.

Perhaps the most popular application of LSFM is for tracking embryogenesis. The first such example was visualizing Drosophila embryogenesis labeled using EGFP fused to moesin for visualizing plasma membranes. Since then embryogenesis has become a prime target for visualization via LSFM, with most researchers opting to use nuclear markers, both genetically-expressed like EGFP-histones and/or exogenously induced as with the nuclear stain Hoescht 33342. This has allowed for comprehensive cell lineage tracing and tracking in developing embryos. Figure 10 contains DSLM images of mouse embryogenesis, tracking different sections of the developing embryo over time and in 3D. Later, two photon DSLM was implemented to perform similar experiments using Drosophila instead. More focused studies have been performed, for example lineaging only cells involved in the lateral line development in zebrafish. Using modern computational methods, fully automated cell tracking has been demonstrated.

The speed and large field of view provided by LSFM has proved useful for calcium imaging in living specimens. The calcium-sensitive FRET probe Cameleon (YC3.6), which consists of a cyan and yellow fluorescent protein tethered by a calcium-sensitive calmodulin domain, was used to probe calcium dynamics in Arabidopsis roots in response to external stimuli, such as the addition of ATP (Figure 11). In one recent study the time-variant activity of about 80% of a zebrafish neural network was visualized using the calcium indicator GCaMP5G, discovering then-unknown correlations in the activity of disparate parts of the brain. OCPI was demonstrated by visualizing calcium dynamics in mouse vomeronasal neurons exposed to a variety of stimuli. The very first 2PE LSFM was demonstrated for calcium imaging of C. elegans using Cameleon.

One of the earliest extensions of LSFM was towards fluorescence correlation spectroscopy (FCS), a technique that correlates fluctuations in fluorescence intensity to changes in fluorophore concentration, and other physical parameters.  FCS is commonly used to determine concentration, singlet-triplet dynamics, diffusion coefficients, and more.  Though most often applied in conjunction with confocal or multiphoton microscopy, the combination of SPIM and FCS (termed SPIM-FCS) has been validated for recording over 4000 spectra in under a minute, and with minimal sample perturbation.  Confocal systems typically require about a minute to collect a single spectrum.

Another exciting application of LSFM is towards optogenetic control of biological functions.  Optogenetics is an emerging field that relies upon light-sensitive proteins to control cellular functions and thus organism behavior.  By combining channelrhodopsin-2, halorhodopsin, and LSFM, Arrenberg and colleagues were able to exercise very precise temporal control over ion channels in zebrafish cardiac pacemaker cells (cardiomyocytes).  A digital mirror device (DMD) was used to create arbitrary illumination structures in three dimensions through the detection optics, independent of light sheet illumination, allowing for single-cell activation.  Many emerging methods rely upon novel illumination strategies through the detection optics independent of light sheet generation.  This optogenetic approach allowed for simulated cardiac arrest, tachycardia, and more.

Another extension of LSFM is towards laser-based microsurgery.  Laser ablation microsurgery is a relatively common technique employed in biological research, with applications such as cytoskeletal surgery (e.g. cutting microtubule filaments), cell membrane disruption, and other morphogenetic studies.  As with the optogenetic approach discussed, the light sheet illumination is uninterrupted, with the surgical laser being projected using the detection optics.  Cell membrane disruption of parts of an MDCK cyst, zebrafish fin cutting, microtubule cutting, and hemocyte migration were all demonstrated using this technique.

The non-invasive nature of LSFM imaging has also proven conducive to single-particle tracking experiments in difficult media and at depths problematic for most microscopies.  Single molecule tracking has been demonstrated several hundred micrometers deep within a sample, where optical scattering and absorption generally preclude such precision approaches, especially using slower methods such as point scanning confocal.

Conclusions

Light sheet fluorescence microscopy (LSFM) is undergoing a veritable renaissance, with many exciting approaches having been invented and optimized in the past several years.  Though an old concept, LSFM encompasses a revolutionary set of techniques premised upon breaking the co-dependence of the illumination and detection pathways, allowing for drastic reductions in volumetric imaging times, phototoxicity/photobleaching, and an overall improvement in image quality of large fluorescently-labeled samples.  Imaging is performed efficiently; planar illumination ensures that structures in the focal plane are selectively excited and detected, rather than needlessly irradiating the entire sample volume in order to detect fluorescence from a small sub-volume.  This is especially important as experimental demands shift towards quantitative in toto imaging of large biologically relevant samples (e.g. entire model organisms, tissues, and three dimensional cultures) with non-negligible 3D structure.  Also, thinner samples such as adherent cells benefit greatly from drastically reduced light exposure.

LSFM has opened a number of doors in terms of imaging fast and efficiently.  Large (100+ µm diameter) cell cysts and spheroids are great three-dimensional homeostatic model systems with applications in cancer treatment and drug discovery, as their functions imitate those of tumors.  LSFM provides a method for long-term and high performance imaging of such culture systems.  This is even more pronounced when imaging model organisms such as zebrafish and fruit flies, allowing advanced applications such as comprehensive cell lineage reconstruction and FRET to be performed in whole developing embryos.  LSFM is also applicable to structural and tomographic imaging of very large organ structures, most famously the inner ear cochlea, and has been used in correlative imaging with established tomographic techniques such as µCT.  LSFM methods represent a paradigm shift, an answer to the increasing demands for better 3D imaging techniques.  Many challenges still exist, especially in the realm of data management, but the increasing popularity and usefulness of LSFM bodes well for the future of this technique.