Effects of excitation light polarization on fluorescence emission in two-photon light-sheet microscopy

Light-sheet microscopy (LSM) is a powerful imaging technique that uses a planar illumination oriented orthogonally to the detection axis. Two-photon (2P) LSM is a variant of LSM that exploits the 2P absorption effect for sample excitation. The light polarization state plays a significant, and often overlooked, role in 2P absorption processes. The scope of this work is to test whether using different polarization states for excitation light can affect the detected signal levels in 2P LSM imaging of biological samples with a spatially unordered dye population. We compared the fluorescence signals obtained using different polarization states with various fluorophores (fluorescein, EGFP and GCaMP6s) and different samples (liquid solution and fixed or living zebrafish larvae). In all conditions, linear polarization oriented parallel to the detection plane provided the largest signal levels, while perpendicularly-oriented polarization gave low fluorescence signal with the biological samples, but a large signal for the fluorescein solution. Finally, circular polarization generally provided lower signal levels. These results highlight the importance of controlling the light polarization state in 2P LSM of biological samples. Furthermore, this characterization represents a useful guide to choose the best light polarization state when maximization of signal levels is needed, e.g. in high-speed 2P LSM.


1
Introduction Light-sheet (LS) fluorescence microscopy is a powerful optical imaging technique 1 based on the principle of a planar illumination oriented orthogonally with respect to the detection axis 2 . It employs wide-field detectors that allow to parallelize the photon collection, thus offering a large increment in the acquisition speed. Moreover, it offers also a good optical section capability and reduced sample photodamage and photobleaching, compared to other optical imaging techniques 3 . Two-photon (2P) LS microscopy 4-8 is a technique developed from traditional 1-photon (1P) LS microscopy that exploits the 2P absorption effect for sample excitation 9 . The excitation wavelengths used in 2P absorption are usually in the infra-red region: a frequency range characterized by reduced scattering inside biological tissues compared to visible light 10 . This effect, combined with the quadratic dependence of the absorption rate on the excitation light intensity, offers several additional advantages: a larger penetration depth in the sample, a reduction of the sample-induced aberrations, a better uniformity of the illumination distribution and an improved image contrast 11 .
The polarization state of the excitation light plays a significant, and often overlooked, role both in 1P and 2P absorption processes exploited in microscopy, operating differently in the two cases 12 . In particular, in the 2P absorption process, the sum of angular momenta of the absorbed photons is required to be zero, since the total angular momentum change related to the electronic state transition in most fluorophores is null 13,14 . This is the reason why linearly-polarized light is associated with a higher 2P absorption with respect to circularly-polarized light, since in the former configuration there is a 50% probability for the fluorophore to interact with photons with opposite handedness that can reciprocally compensate their own angular momenta. On the other hand, the use of circularly-polarized light will lead to a spatially more homogeneous fluorophore excitation, whereas in case of linearly-polarized light the excitation probability depends on cos 2 (θ), for 1P excitation, or cos 4 (θ), for 2P excitation; where θ is the angle between the polarization axis of the exciting electric field and the dipole moment orientation of the dye molecule 15 . This means that when linearly-polarized excitation light is used, a photoselection of the dyes is performed based on their spatial orientation and this effect is much more pronounced with 2P excitation. We illustrated this situation in Fig. 1 that shows a dye emitting fluorescence only when excited with a linearly-polarized light with θ 0. ≃ It should be noted that this situation is different from the case in which the spatial anisotropy is present in the general dye population regardless of the excitation, i.e. when the dye population is spatially ordered due to the inherent biological properties of the sample. In the latter case polarization-resolved 1P or 2P fluorescence microscopy can be used to extract important information about the sample micro-architecture [16][17][18][19] ; however, in the present work we will focus on the more general case in which the native population of fluorophores is supposed to be randomly oriented.
The fact that, when using linearly-polarized excitation light, the population of excited dyes is spatially anisotropic induces a spatial anisotropy also in fluorescence emission. This is because the fluorescence emission happens preferentially on an axis perpendicular with respect to the emission transition moment 15 (the latter can be grossly approximated as parallel with the absorption transition moment for many dyes). The effects of this spatial anisotropy, as well as of the ellipticity of the excitation light, on biological imaging of randomly-oriented dyes were already experimentally characterized for 1P confocal microscopy and 2P microscopy 13 . Nevertheless, their characterization is still lacking for LS microscopy, where the different geometry of the excitation and detection optical axes makes the presence of this anisotropy even more significant. We illustrated this situation in Fig. 1. As it is shown, the different orientationally-defined dye populations photoselected by the polarization directions of the excitation light preferentially emit light toward different directions. This means that the polarization orientation of the excitation light could in principle be experimentally orientated as to maximize the light emitted toward the direction of the detection objective in a LS microscope.
The described situation assumes the time-scale of the dye rotational movement being much slower than the fluorescence life-time, meaning that the orientation population of the excited dyes could not randomize before the photon emission. This does not represent the general case, since in liquid solution the majority of fluorophores can rotate in a time-scale of 50 ÷ 100 ps, while the fluorescence life-time is usually in the time scales of 1 ÷ 10 ns 15 . On the other hand, rotational diffusion is limited in viscous media, such as biological tissues. In addition to medium viscosity, the rotational diffusion can be mitigated also by bonded molecules which damp the free rotational movement of the fluorophore. This can be relevant when the fluorophore is actually an internal moiety of a large bio-macromolecule, such as the Green Fluorescent Protein (GFP), or when it non-covalently interacts with larger biological molecules.
In general, there are several environment factors relevant in a biological sample context that can bring the rotational time-scale to be comparable to typical fluorescence life-times, thus affecting the anisotropy of the excited dye population and therefore of the spatial fluorescence emission.
Therefore, the scope of this work is to test whether the use of different polarization states for the excitation light (circular polarization or linear polarization, with two orthogonal polarization orientations) can affect the detected signal levels when performing 2P LSM of biological samples in which the dyes are randomly oriented.

Methods
We used two strains of transgenic zebrafish (Danio rerio) larvae: 3 Tg(elavl3:H2B-GCaMP6s) larvae 20,21 in homozygous albino background 22 and 6 Tg(actin:EGFP) larvae 23 . The former expresses, with nuclear localization, the fluorescent calcium sensor "GCaMP6s" under a panneuronal promoter, while the latter expresses enhanced GFP (EGFP) in all tissues owing to a ubiquitous promoter. Zebrafish strains were maintained according to standard procedures 24 . To avoid skin pigment formation, Tg(actin:EGFP) larvae were raised in 0.003% N-phenylthiourea (P7629, Sigma-Aldrich). All larvae were observed at 4 days post fertilization (dpf). Fish maintenance and handling were carried out in accordance with European and Italian law on Five of the larvae were subjected to live imaging. Immediately before the acquisition, each larva was anesthetized with a solution of tricaine (160 mg/L; A5040, Sigma-Aldrich), was included in 1.5% (w/v) low gelling temperature agarose (A9414, Sigma-Aldrich) in fish water (150 mg/L Instant Ocean, 6.9 mg/L NaH 2 PO 4 , 12.5 mg/L Na 2 HPO 4 , pH 7.2), and mounted on a custom-made glass support immersed in fish water thermostated at 28.5 °C. The other 4 larvae were fixed (2h in 4% paraformaldehyde in PBS at room temperature) before undergoing the same mounting procedure.
The imaging was performed with a custom-made 2P LS microscope. The setup scheme is shown in Fig. 2. Excitation light at 930 nm is generated by a pulsed Ti:Sa laser (Chameleon Ultra II, Coherent) and a pulse compressor is employed to pre-compensate for the group delay dispersion (PreComp, Coherent). The beam is attenuated using a half-wave plate and a Glan-Thompson polarizer and then it passes through an Electro-Optical Modulator used to rotate on command its linear polarization plane by 90°. Moreover, we use a combination of a half-wave plate and a quarter-wave plate to align the light polarization plane with the reference system of the microscope and to pre-compensate for the polarization distortions. The beam is then scanned by a fast resonant galvanometric mirror (CRS-8 kHz, Cambridge Technology), used to generate the digitally-scanned LS along larval rostro-caudal direction, while a closed-loop galvanometric mirror (6215H, Cambridge Technology) is used to scan the LS along larval dorso-ventral direction. The beam is finally relayed to an excitation dry objective (XLFLUOR4X/340/0,28, Olympus), placed at the lateral side of the larva, by a scan-lens (50 mm focal length), a tube-lens (75 mm focal length) and a pair of relay lenses (250 mm and 200 mm focal lengths) that underfill the objective pupil. When needed, we converted the light polarization state from linear to circular by placing a removable quarter-wave plate on the beam-path between the tube lens and the first relay lens.
The emitted green fluorescent light, coming either from GCaMP6s or EGFP, is collected by a water-immersion objective (XLUMPLFLN20XW, Olympus) placed dorsally above the larva. The objective is scanned along the axial dimension by an objective scanner (PIFOC P-725.4CD, Physik Instrumente) synchronously with the closed-loop galvanometric mirror movements. The optical image formed by the detection-objective tube lens (300 mm focal length) is then demagnified by exploiting a second pair of tube lens (200 mm focal length) and objective (UPLFLN10X2, Olympus), bringing the final magnification to 3×. Finally, the green fluorescence is spectrally filtered (FF01-510/84-25 nm BrightLine® single-band bandpass filter, Semrock) and relayed to a sCMOS camera (ORCA-Flash4.0 V3, Hamamatsu).
Imaging was performed with a pixel size of about 2×2 μm 2 , and a field of view of about 1×1 mm 2 . The acquisitions in fluorescein solution were performed on a single transversal plane with an exposure time of 100 ms. The larvae instead were imaged with volumetric acquisitions composed by 31 planes spaced by 5 μm and with an exposure time of 26 ms for each plane and a volumetric acquisition frequency of 1 Hz (~200 ms where reserved for objective flyback time). Each acquisition lasted 1 minute and then the 60 acquired volumetric stacks were averaged to obtain one final z-stack.
The laser power used for the acquisitions, measured at the excitation objective pupil, was 100 mW for the Tg(actin:EGFP) larvae both in living and fixed preparations, 200 mW for the live imaging of Tg(elavl3:H2B-GCaMP6s) larvae, 180 mW for the fixed Tg(elavl3:H2B-GCaMP6s) larvae and 162 mW for the fluorescein solution acquisition. Great care was taken to ensure that the excitation power remained constant when imaging with the three different polarizations. Moreover, we checked that this power range is far from the fluorescence saturation regime by measuring the average fluorescent signal generated by a fixed Tg(actin:EGFP) larva while varying the excitation power from 25 mW to 525 mW. The results, shown in Fig. S1 in the Supplemental Materials, clearly depict a quadratic dependence of the signal from the excitation power (coefficient of determination: 0.999), as expected in 2P microscopy, and therefore we can exclude the presence of a saturation effect.
Before each acquisition session, we monitored the residual polarization distortions by temporarily inserting on the beam-path a half-wave plate followed by a Glan-Thompson polarizer before the excitation objective pupil. We then manually rotated the retarder while measuring the power variation after the polarizer. For circularly polarized light the amplitude of the observed oscillations was less than 4% of the signal.
General linear mixed models were used to analyze the results for the Tg(actin:EGFP) larvae and the fixed Tg(elavl3:H2B-GCaMP6s) larvae. The models were implemented with the library "lmerTest" 25 for the R language for statistical computing. We used the fluorescent signal as dependent variable, the polarization state as fixed effect and the fish as random effect. A linear regression model implemented in R language was instead used to analyze the results for the living Tg(elavl3:H2B-GCaMP6s) larva. We used the fluorescent signal as dependent variable and the polarization state and the Region Of Interest (ROI) as independent variables. In both cases we used linear contrasts to compare the polarization groups and we used the Sidak method for the multiplicity correction. Fluorescein solution data were compared by computing 95% Confidence Intervals (C.I.) using the Student's t-distribution. In the following, all the fluorescence signal values are expressed in Arbitrary Units (A.U.).

3
Results and discussion In a medium where the fluorophores are able to rotate completely unrestrained, we would expect the fluorescence emission to be isotropic, because in this condition the thermally induced rotation movements happen on time scales much shorter than fluorescence lifetime, as discussed in Sec. 1.
To test this hypothesis, we excited fluorescence in a high-concentrated fluorescein solution employing linearly-and circularly-polarized light, and we show the results in Fig. 3. The polarization plane of the former was aligned either parallel or perpendicular to the plane where the optical axes of the detection and excitation objectives lay; in the following we shall refer to the parallel condition as "vertical polarization" and the perpendicular condition as "horizontal polarization" (corresponding to the z-axis and the y-axis in Fig. 1, respectively).
We indeed observed similar fluorescence levels when employing vertically-or horizontallypolarized light, nevertheless we revealed a small, albeit statistically significant, difference between the two polarization states: the signal in horizontal-polarization condition (544. ). This observation indicates that even in a medium that favors high-level of molecular mobility, as in solution, the fluorophores can still exhibit a residual degree of spatial anisotropy in their fluorescence emission. A much larger difference was instead observed for circularly-polarized light: for this condition, we observed a ~30% reduction in the fluorescence signal level (374.4 A.U.; 95% C.I.: [372.0, 376.8] A.U.) with respect to the two linear polarization conditions. This is consistent with the low 2P excitation efficiency that characterizes the circular polarization. Nevertheless, it is not a trivial result, since circular polarization can excite the dyes in a spatially-homogeneous fashion, while the linearly-polarized light can excite only the subset of dyes that are almost parallel to its polarization-plane, as discussed in Sec. 1. This result therefore indicates that the widening of the group of possible target dyes is not sufficient to compensate the decrease in excitation efficiency for the circular polarization with respect to the linear polarization. We then tested if this polarization-dependent effect is present also in tissue imaging. To do so, we observed zebrafish larvae expressing EGFP, both in fixed and in living conditions, and we show the results in Fig. 4. In this case, we selected an arbitrary ROI for each larva (as depicted in Figs. 4(a) and 4(b)) and we measured its mean fluorescence signal. In this case, we did not observe significant differences between the circular-polarization condition and the verticalpolarization condition in both the fixed (41.6 A.U., standard deviation: 15.3 A.U. and 45.0 A.U., standard deviation: 14.2 A.U., respectively) and the living (53.4 A.U., standard deviation: 18.8 A.U. and 48.9 A.U., standard deviation: 17.5 A.U., respectively) conditions. We observed instead a large and significant (p-value < 0.0001) signal increase in the horizontal-polarization condition with respect to the circular-and the vertical-polarization conditions, both in the fixed-condition (~67% and ~54%, respectively; horizontal-polarization value: 69.3 A.U., standard deviation: 20.4 A.U.) and in the living condition (~41% and ~54%, respectively; horizontal-polarization value: 75.2 A.U., standard deviation: 29.2 A.U.).
It should be noted how in the animal tissue the difference in the signal levels between the horizontal-and the vertical-polarization conditions is much more marked with respect to the fluorescein solution. We hypothesize that this effect could be ascribed to the different molecular rotational mobility in the two environments. Finally, we tested if this polarization-dependent effect can be observed also with a fluorescent calcium indicator, such as GCaMP6s. For the fixed-condition, we measured the average fluorescence signal emitted by arbitrarily selected ROIs, similarly to what we did for the EGFP experiments, and we show the results in Figs. 5(a) and 5(c).
Also in this case, we did not observe a significant difference between the circularpolarization condition (14.5 A.U., standard deviation: 1.1 A.U.) and the vertical-polarization (15.5 A.U., standard deviation: 1.3 A.U.) conditions. However, the measured fluorescence levels in the horizontal-polarization condition (26.5 A.U., standard deviation: 1.5 A.U.) showed a large and significant (p-value < 0.0001) increase with respect to the circular-polarization condition (~83%) and the vertical-polarization condition (~71%).
We tested if the GCaMP6s polarization-dependent effect can be observed also during liveimaging. These measures are different with respect to the previous ones, since the cellular calcium levels, and therefore the emitted fluorescence, variate during the time, reflecting the time-dependent neuronal activity. In particular, this means that, due to the fluctuations in basal neuronal activity, the fluorescence levels change in the time needed to switch the polarization state. For this reason, we decided to draw ROIs around individual neuronal cells (i.e. the individual sources of the time-dependent signal), and we show the results in Figs. 5(b) and 5(d).
In this case we did not observe a significant difference between the circular-polarization condition (3.6 A.U., standard deviation: 1.9 A.U.) and the vertical-(2.1 A.U., standard deviation: 1.4 A.U.) and the horizontal-polarization (4.8 A.U., standard deviation: 3.5 A.U.) conditions. However, we observed a large (~128.6%) and significant (p-value=0.0016) increase in the fluorescence signal level in the horizontal-polarization condition with respect to the verticalpolarization condition.
The slightly different trends observed for GCaMP6s between the fixed and the living conditions could be ascribed to several factors. The fixation procedure induces cross-linking between molecules that could alter the rotational mobility of the fluorophore. Moreover, the physico-chemical properties of the cytosol change between the living and the fixed states and this medium alteration could affect the motion of the dye. Finally, the fine spatio-temporal biological control of the calcium distribution is completely abolished in the fixed state and therefore also the distribution between the bound and the unbound states of the fluorescent sensor is altered in the two cases, affecting its fluorescence characteristics.  Conclusions In this work we compared the fluorescence signal levels obtained using different excitation light polarization states with various fluorophores and different samples in 2P LSM. In all the different conditions tested, horizontal polarization proved to provide the largest signal levels, while circular polarization generally provided low signal levels. Moreover, vertical polarization gave low fluorescence signal levels with all the biological samples, albeit it provided a large signal for the fluorescein solution. Taken together, these results highlight the importance of controlling the polarization state of the excitation light in 2P LSM of biological samples.
Furthermore, this characterization represents a useful guide to choose the properly-oriented linearly-polarized light when maximization of signal levels is needed. This is particularly important in high-speed 2P LSM, because in this situation (differently from 1P LSM) the acquisition frequency is usually limited by the signal-to-noise ratio and therefore increasing the signal levels is necessary to achieve a higher temporal resolution.

Disclosures
The authors declare that there are no conflicts of interest related to this paper.