Light-Sheet Confined Super-Resolution Using Two-Photon Photoactivation

Light-sheet microscopy is a useful tool for performing biological investigations of thick samples and it has recently been demonstrated that it can also act as a suitable architecture for super-resolution imaging of thick biological samples by means of individual molecule localization. However, imaging in depth is still limited since it suffers from a reduction in image quality caused by scattering effects. This paper sets out to investigate the advantages of non-linear photoactivation implemented in a selective plane illumination configuration when imaging scattering samples. In particular, two-photon excitation is proven to improve imaging capabilities in terms of imaging depth and is expected to reduce light-sample interactions and sample photo-damage. Here, two-photon photoactivation is coupled to individual molecule localization methods based on light-sheet illumination (IML-SPIM), allowing super-resolution imaging of nuclear pH2AX in NB4 cells.


Introduction
The growing interest in large organisms and tissue imaging led to a rapid development in two-photon excitation microscopy [1] and light-sheet based techniques [2] [3]. The basic idea behind light-sheet based techniques relies on the selective illumination of a planar region in the focal plane coupled with a CCD based detection system. These techniques provide optical sectioning capabilities and, recently, they have been widely used to image thick biological samples. The earlier idea of orthogonally oriented illumination and detection [4] has been successfully extended to Bessel beam illumination [5] [6], structured illumination [7] [8] [9], confocal detection [10] and multiview imaging [11] [12] [13] in order to improve the resolution and imaging capability in scattering samples [14]. Within this framework, two-photon excitation [1] based approaches have been successfully implemented in light-sheet based optical configurations [15] [16] [17], thus reducing scattering effects and light-matter interactions [18]. On the other hand, in the last few years several emerging fluorescence techniques based on single-molecule localization [19] [20] [21], allowed super-resolution imaging of biological structures at molecular level [22]. The concept behind these techniques is based on the repeated detection and localization of single fluorophores with nanometer precision. The molecule sparseness, necessary for the localization, can be obtained exploiting the photoactivation process the fast conversion of photoactivatable molecules to a fluorescent state by light irradiation [23]. The super-resolution image is then obtained mapping the position of all the localized molecules, thus allowing the reconstruction of subcellular structures circumventing the diffraction limit. This led to a rapid development in innovative solutions providing localizationbased techniques with multicolour [24] [25] and three-dimen-sional imaging capabilities [26] [27] [28]. Furthermore, in order to spatially confine the activation process, approaches based on temporal focusing [29] and selective plane illumination microscopy (IML-SPIM) [30] have been recently developed, allowing the 3D super-resolution imaging of whole cells and thick living cellular spheroids, respectively. All these new techniques extended the 3D imaging capabilities of localization-based methods [31] to thicker and thicker samples, thus taking a further step towards in vivo super-resolution imaging. Additionally, several algorithms [32], capable of improving the localization process within high-density samples [33] [34] featuring high background regimes [35], have been recently optimized. Despite this, one of the main limitations to super-resolution imaging of large samples is still represented by light-sample interactions and refractive index inhomogeneities since absorption and scattering effects may induce photon losses and spreading respectively. In the last decades, two photon excitation has been proven to be a golden standard to improve imaging depth capabilities and to reduce the scattering effects thanks to the use of higher wavelengths [36]. Furthermore, the two photon process has been demonstrated to be a suitable approach to perform photoactivation experiments using photoactivatable dyes or fluorescent proteins such as photoactivatable green fluorescent protein (PaGFP) [37]. In this case, after photoactivation, the excitation spectrum shifts maintaining the very same emission spectrum of the single-photon case for the native and photoactivated protein [38].
Within this framework, two-photon (2P) photoactivation may thus be a key tool for improving the performance of localizationbased super-resolution techniques, allowing higher imaging depths to be achieved and improving the confinement of the photoactivation process in thick samples. 2P process followed by widefield excitation was initially implemented in order to perform 2D super-resolution [39] [40] and 3D photoactivation localization microscopy (PALM) of ''whole'' cells [29], confining the photoactivation process and thus alleviating background noise. Within this framework, two-photon photoactivation can provide benefits to the localization performance of recently developed super-resolution imaging methods for thick samples such as IML-SPIM. In particular, when localization based super-resolution imaging is performed within selective plane illumination architecture, tissueinduced distortions and scattering effects may cause an enlargement of the light-sheet intensity distribution, thus generating an undesired photoactivation of molecules outside the light-sheet volume. Additionally, when super-resolution imaging of live biological samples is performed, two-photon photoactivation makes it possible to reduce the photo-damage compared with UV-induced activation. This paper demonstrates super-resolution obtained by two-photon photoactivation confined in a light-sheet optical illumination scheme.

Results
When imaging of scattering sample is performed using conventional SPIM, where illumination in the visible range is used, potential spreading of the exciting photons [18] and the consequent enlargement of the effective excitation volume can occur [41]. Within this scenario, in order to investigate the advantages of 2P photoactivation, we first assessed the effects induced by light-sample interactions on the effective excitation volume obtained using similar illumination conditions both for single-and two-photon regime. The light-sheet is created using a cylindrical lens (see fig. 1A) according to the conventional selective plane illumination configuration [2] shown in fig. 1B. The basic idea behind selective plane illumination microscopy (SPIM) relies on the confined illumination within a thin plane orthogonally oriented to the detection axis. The illumination and the detection paths are thus decoupled and only a planar region in the focal plane is excited. The light-sheet is generated using a cylindrical lens and a low magnification objective in order to obtain the suitable light sheet dimension and to reduce aberrations. The sample is held in a water chamber and is moved within the focus of the detection lens in order to perform optical sectioning.
First, we characterize the illumination intensity distribution and the light-sheet dimension imaging the illumination beam both in the visible and in the IR range. A direct measurement of the illumination light-sheet thickness can be made by reflection imaging of the incident light: a plane mirror is inserted in the sample holder and oriented at 45 degrees with respect to the illumination axis and the reflected light is acquired on the CCD camera (as shown in fig. 1B). Calibrated illumination conditions were set for single-photon (488nm) and for two-photon (790 nm) regime in order to obtain illumination light sheet distributions endowed with a comparable thickness. A fitting procedure ( fig. 1C) and the FWHM calculation provide the thickness of the illumination light-sheet, which proves to be (2. Thereafter, the excitation intensity distribution can be characterized by fluorescence SPIM imaging both in non scattering and highly scattering samples, in order to investigate effects due to incident scattered photons. The effective volume where 1P and 2P excitation occurs within the light-sheet geometry can be studied by imaging XY and XZ cross-sectional views of the fluorescence intensity distribution on an immobile fluorescent sample (materials and methods). A rotation of the cylindrical lens allows to switch easily between the XY and the XZ views. The XZ cross sectional view of the fluorescence intensity distribution is imaged on a CCD camera with the cylindrical lens rotated by 90 0 with respect to conventional imaging orientation both for single-photon excitation (see fig. 2A) and for two-photon excitation (see fig. 2B). The illumination source used for 2PE is a tunable Ti:Sapphire laser, Chameleon ULTRA II, Coherent Inc. (pulse durations of ,140 fs and repetition rate of 80 MHz across the tuning range (680-1060 nm). The non-linear dependence of the emitted fluorescent signal on excitation power was verified experimentally (data not shown) in order to demonstrate the 2PE regime despite the low photon density configuration provided by the light-sheet. In order to evaluate sample-induced distortions affecting the excitation volume, mainly due to scattering effects, we analysed the XZ profile of the fluorescence intensity distribution in uniform samples with different scattering coefficients. To this end, scattering phantom samples, capable of mimicking the optical properties of transparent biological samples (scattering coefficient m s = 5 mm 21 ) and mammalian cell tissue (m s = 50 mm 21 ) have been used [42]. Changes in fluorescence intensity distribution due to the increased illumination path length in different scattering phantom samples makes it possible to evaluate the distortions induced by lightsample interaction on the light-sheet shape. In fact, the more the light travels through the sample, the more evident the effects on the effective light-sheet excitation volume are. Imaging of the XZ cross-section of the excitation intensity distribution corresponding to different illumination depths (400, 600 and 800 mm within the sample) is performed (at a constant detection depth of 150 mm). The FWHM of the intensity profile along the axial path (see blue line in figs. 2A, 2B) allows us to evaluate the distortion of the effective excitation distribution damaging the optical sectioning capability. In fact, the experimental profile of the effective lightsheet distribution in a uniform scattering sample (m s = 50mm 21 ) shows a more pronounced distortion in the linear configuration (see fig. 2C) than in the non-linear configuration (see fig. 2D) as a function of the illumination depth increase. The FWHM values of the excitation intensity profiles, with the illumination light-sheet thickness being equal for single-and two-photon excitation, are obtained using a gaussian-lorentzian fitting procedure, both for single-photon excitation (488nm) and two-photon excitation (790nm), at different illumination depths (table 1). Average FWHM values are statistically obtained over 20 samples and the mean values of the adjusted R-square are shown in table 1 for highly scattering samples. When the light-sheet position is shifted from 400 mm to 800 mm deep within the scattering sample (m s = 50 mm 21 ) the thickness of the effective excitation distribution increases: a 73.5% increment in the linear configuration and a 23.7% increment in the non-linear configuration are reported. Considering the lower scattering phantom sample (m s = 5 mm 21 ), the FWHM of the light-sheet intensity profile shows a closer thickness increase (17% for 1PE and 18% for 2PE), while varying the illumination depth (from 400 mm to 800 mm). The corresponding values (corresponding to the mean values over 20 repeated measurements and the statistical errors) are reported in table 1 both for the low scattering and the high scattering regimes. The observed distortion of the effective fluorescence light sheet profile and the increased thickness of the excitation intensity distribution can be considered as a direct consequence of the photon spreading induced by the scattering properties of the sample. This effect is minimised when moving to two-photon excitation where the Gaussian shape is well preserved and the excitation process is confined better especially in highly scattering samples. The reduced axial confinement of the light-sheet excitation affects the optical sectioning capabilities of the SPIM system and induces a higher background when the illumination depth increases (fig. 3). This effect is less evident when 2P light sheet regime is used resulting in an image contrast improvement. This preliminary analysis on the effects provided by a 2PE approach within a selective plane illumination scheme suggests that single molecule localization can benefit from the confined excitation process demonstrated above. IML-SPIM can be promoted by confining the photoactivation process towards applications to imaging of large scattering samples in depth as a perspective. In this framework, fig. 4 represents the XZ crosssections of the light-sheet photoactivation volumes both for singlephoton (see fig. 4A) and two-photon induced activation (see fig. 4B). When photoactivation in scattering samples is performed using UV-VIS wavelengths, the thickness of the effective photoactivation volume increases and the confinement of the molecules involved in the image formation process is reduced, as shown in the schematic diagram in fig. 4A. On the other hand, when priming the photoactivation process in a confined volume using 2PE, molecules arising from a more homogeneous lightsheet distribution can be localized when imaging large scattering samples (see fig. 4B). The confinement of the photoactivation process and the improved contrast in scattering samples, direct consequences of the preserved light-sheet spatial distribution, suggests that two-photon process can be a useful solution for photoactivation in IML-SPIM. Here, we will focus on the demonstration of IML-SPIM imaging of subcellular structures exploiting 2P photoactivation. 2P photoactivation reduces scattering effects and preserves the activation volume shape, thus improving the confinement of the photoactivation process in scattering samples. In such a configuration, only the photoactivated molecules within the light-sheet volume are turned on (as shown in the schematic diagram in fig. 5A) and imaged by IML-SPIM. Following the experimental evidence and the previous considerations, we demonstrate, for the first time, individual molecule localization in a standard light-sheet regime taking advantage of 2P photoactivation at 760 nm. Individual molecule localization after 2P photoactivation is exploited to perform super-resolution imaging of the phospho-histonH2A.X (pH2AX) in NB4 cells. Immunostaining of NB4 cells with photoactivatable dyes (FLIP565, Abberior), previously used in literature for single-molecule detection and 2P photoactivation    43], was performed. The comparison between the conventional SPIM imaging ( fig. 5B) of the spatial distribution of the nuclear pH2AX and the corresponding super-resolution imaging with 2P photoactivation ( fig. 5C) shows the imaging improvement provided by 2P IML-SPIM. In the super-resolution experiment the readout was obtained exciting the photoactivated molecules at 561 nm. We used a CW readout laser (561 nm) for imaging and an IR pulsed laser at 760 nm to prime photoactivation, both of which ran continuously during the entire experiment (5000 frames). Prior to 2P photoactivation, an initial high-power illumination phase at 561nm is performed in order to bleach all the previously photoactivated molecules in the sample. The intensities used for 2Pn photoactivation and single photon readout were 50-200kW/cm 2 and 9KW/cm 2 respectively. The photactivation power is gradually increased during the entire experiment in order to control the number of active molecules/frame. The relatively low power used for photoactivation and the short pulse duration used allow to reduce thermal effects and sample instabilities. To analyse the IML-SPIM images, we used an established localization approach based on the procedure previ- ously indicated for super-resolution imaging through fluorescence photoactivation localization microscopy [21]. We collected photons from individual fluorescent molecules achieving a mean lateral localization precision of 40 nm. The localization precision achieved depends on the number of photons/molecule collected and is calculated according the previous work by Mortensen and colleagues [44]. Histograms show the distribution of the localization precision reached ( fig. 5D) and the number of photons/ molecule collected ( fig. 5E), thus demonstrating that 2P photoactivation IML-SPIM provides an efficient single molecule mapping.

Discussion
In this work, we demonstrated for the first time the utilization of 2P photoactivation in a light-sheet architecture towards IML-SPIM super-resolution imaging. The relevance of such an approach lies on coupling the benefits of 2PE imaging when dealing with highly scattering objects and super-resolution capabilities already demonstrated for 1PE IML-SPIM. First, the evaluation of the effects induced by light-sample interactions and a comparison between the performances of 1P and 2P regimes were performed. Particular attention has been addressed to the distortions induced on the effective excitation distribution when imaging non-transparent samples in depth. Experiments showed a distortion of the effective fluorescence light-sheet profile and an enlarged excitation intensity distribution due to the photon spreading induced by the scattering properties of the sample in the linear regime. The light sheet confinement and the improved contrast shown under 2P regime in thick scattering samples suggests this technique as a promising solution to improve the imaging capabilities of light-sheet based super-resolution techniques. Following the demonstration of the improved performances of the 2P based approach in terms of the homogeneity of the light sheet excitation, we showed IML-SPIM imaging based on 2P photoactivation of nuclear pH2AX in NB4 cells reaching a 40 nm localization precision. The 2P photoactivation IML-SPIM has great potential for the super-resolution imaging of complex biological specimens thanks to the reduction of scattering effects when realizing the sparse molecules photoactivation in depth. 2P-IML-SPIM can be used to browse through living or fixed samples with limited photodamage, thus targeting specific events for superresolution analysis with the unique penetration depth provided by infrared non-linear interactions.

Immobile Fluorescent Sample
The derived fluorescence emission intensity distribution was collected by imaging FITC Dextrans molecules (1 mg/10 ml, m.w. 500KDa) embedded in 1.5% agarose gel.

Phantom Scattering Sample
Fluorescent phantom samples with calibrated scattering coefficients were made by entrapping FITC dextrans (1 mg/10 ml, m.w. 500 KDa, Sigma Aldrich) in 1.5% low-melting-point agarose (Sigma Aldrich, low melting) diluted in 16 PBS (pH 7.4) and embedding different amounts of non-fluorescent polystyrene microspheres (d = 1.03 mm, Polysciences) in the agar solution in order to control the scattering properties of the sample. Scattering phantom samples capable of mimicking the mean optical properties of transparent biological samples (5 mm 21 ) and mammalian cell tissue (50 mm 21 ) were obtained using concentrations of 3.64, and 36.4 particles/ml (Mourant, et al., 1998) respectively. The sample was embedded in agarose gel and enclosed in a glass capillary (Socorex capillary tube, volume 60-100 ml, diameter 1mm).
Contrast measurements in a phantom scattering sample (1 mm sized scattering beads were embedded in 1.5% agarose gel resulting in a controlled scattering coefficient of 50mm 21 ) were performed by imaging sub-resolved fluorescent beads (2 mg/ml (W/V) Invitrogen FluoSphere, 173 nm diameter).

Two-photon IML-SPIM Optical Set Up
The optical system used ( fig. 1b) is based on the IML-SPIM [29] architecture adapted for two-photon photoactivation. The illumination unit is based on a conventional SPIM configuration [2] in which a cylindrical lens (Thorlabs LJ1653R-B, f = 200 mm) focuses the light in the back focal plane of an objective lens (Nikon Plan, 106, NA 0.3) to produce a thin light-sheet (Fig. 1a).
Cell Culture and Immunofluorescence of NB4 Cells NB4 cell line was established from a human Acute Promyelocytic Leukemia (APL) [45]. NB4 cells were grown in RPMI 1640 medium containing 10% FBS, 2 mM glutamine, 50 ng/ml Penicillin/Streptomycin (Lonza) at 37uC in 5% CO 2 and fixed in 2% formaldehyde for 10 minutes at 4uC. Fixed NB4 cells were washed in PBS and permeabilised for 10 minutes in a permeabilisation buffer containing 0.1% Triton X-100 (vol/vol) in PBS. The cells were then incubated for 30 minutes in a blocking solution, 5% BSA (wt/vol) in PBS, and then incubated for 1 hour at room temperature with primary antibodies anti phospho-histonH2A.X (pH2AX )Biolegend) diluited in the blocking solution. The cells were rinsed 3 times in PBS and incubated for 1 hour at room temperature with secondary antibody anti mouse FLIP565 (Abberior). The cells were rinsed 3 times in PBS once again.