Polarization effects on fluorescence emission of zebrafish neurons using light-sheet microscopy

As an advanced imaging approach, light-sheet fluorescence microscopy (LSFM) has been extensively used in developmental biology [1, 2], cell science [3], and neuroscience [4] in the past decade. Unlike epi- or trans-illumination fluorescence microscopy, LSFM illuminates the sample via a thin laminar laser beam, referred to as a light sheet, in the plane of focus and detects the emitted fluorescence signal in the orthogonal direction [5, 6]. This natural feature of optical sectioning brings many salient advantages to the LSFM, such as fast data acquisition, high signal-to-noise ratio, and low phototoxicity and photobleaching, making LSFM the very fit technique for imaging clear biological specimens in vivo [7, 8]. The zebrafish (Danio rerio), as a vertebrate model organism, shares a high similarity of approximately 70$\%$ with human disease genes [9], thus playing an important role in many research areas such as pharmacogenetics and neuropharmacology [10, 11]. The larval zebrafish, due to its transparent property, has become an ideal and popular organism to be studied through light-sheet fluorescence microscopy [12].

During a fluorescence imaging process, a fluorescent probe is usually introduced to the living tissue or cells to realize the visualization of fluorescent entities. Due to the photoselection effect induced by the linearly-polarized light, the fluorescence emission is spatially anisotropic [13, 14]. This is because the linear oscillating electric field modulates the randomly distributed molecules into a non-random distribution regarding the orientations of transition dipole moments[15]. The molecule with a transition dipole moment oriented in the same direction as the polarized light is excited preferentially and fluoresces primarily on an axis perpendicular to the transition dipole moment, leading to anisotropic fluorescence emission [13, 16]. Since the excitation probability strongly depends on the relative orientation between light polarization and emission dipoles, it allows one to experimentally maximize the fluorescence signal by selecting a specific polarization of excitation light in a microscope. However, if the molecules are light and can rotate without constraint in a medium of low viscosity, the photoselection effect is averaged out. The fluorescence emission will present a certain degree of anisotropy only if the rotation occurs on a time scale longer than the fluorescence lifetime. This happens when the fluorophore is attached to a heavier molecule such as proteins. Therefore, the anisotropic fluorescence emission may reveal the micro-architecture of the biological samples. In case there is a pre-existed anisotropic orientation of the dipole moments that the fluorophores are associated with, the polarization-dependent measurements when using linearly-polarized light can in principle be used to extract information about the spatially-ordered biological structures [17]. Therefore, the effects of excitation light polarization on fluorescence emission have attracted enormous interest and been well studied with scanning confocal microscopy [18, 19, 20].

The anisotropic fluorescence emission induced by the polarized excitation light is expected to be more pronounced in the LSFM geometry, where the orthogonal configuration of illumination and detection paths introduces an extra geometrical limitation on the fluorescence signal possible to be collected by the detector, than in the collinear geometry. But few works relevant to the anisotropic fluorescence emission have been carried out in LSFM so far. The polarization-dependent experiment using LSFM has first been conducted to study the contrast mechanism of second harmonic generating (SHG) nanoprobes for in vivo light-sheet imaging [21]. Recently, the importance of excitation polarization control has been reported via quantitatively demonstrating its influence on the detected fluorescence signal levels in a two-photon LSFM [22]. However, both works mentioned above involve nonlinear processes, i.e. SHG and two-photon excitation, where the polarization density responds non-linearly to the electric field of the light. This nonlinear effect would complicate the interpretation of images presenting the biophysical phenomena. Detecting anisotropic fluorescence emission using one-photon LSFM can directly map the relationship between light polarization and imaging properties but has not been done experimentally, thus deserves more research.

In this work, we have employed the self-built high-resolution LSFM to study the polarization effects on fluorescence emission using larval zebrafish as the biological specimen. The LSFM system has a lateral resolution of 0.47 $\mu$m, which can resolve the neuronal cells of the zebrafish larvae. We have performed polarization-dependent experiments on both fluorescent microspheres and GCaMP6f-expressed zebrafish larvae. When varying the polarization of the excitation light, the fluorescence intensities obtained from fluorescent microspheres are nearly constant, hardly depending on the polarization states. While in the case of zebrafish larvae, around a 40% difference in the fluorescence intensity has been observed under different polarization states. With the excitation light propagating along the illumination axis, the fluorescence intensity has a peak value or a valley value when the polarization is perpendicular to or parallel to the detection axis, respectively. To understand the cause of the difference in fluorescence intensity, we have registered the fluorescence images of zebrafish larvae taken under different polarizations on the accuracy of cellular resolution and subtracted them from a reference image. We found that most differential fluorescence signals occurred inside the nerve cells rather than in the extracellular space. The subtraction of fluorescence images of different polarizations eliminates the depolarization fluorescence signal and enables one to obtain a fluorescence image with increased contrast and resolution in a one-photon light-sheet microscope.

We first tested the effect of laser polarization on the fluorescence emission of microspheres. Fig. 2(a) shows a light-sheet fluorescence image of the microspheres. A single microsphere is of diameter 0.5 $\mu$m, occupying 3$\sim$4 pixels. Insets are the zoom-in images of the regions of interest (ROIs), enclosing an area containing either a single microsphere (Inset 1, 2) or a cluster of microspheres (Inset 3-7) or background noise (Inset 8), with sizes of 50 by 50 pixels. The relative intensities of all the ROIs are plotted as a function of HWP tuning angles in Fig. 2(b). The color of the scatter plot corresponds to the color of the outer frame of the inset in Fig. 2(a). The summation of values of each pixel in the ROI is defined as fluorescence intensity and the relative intensity is calculated as the intensity divided by the minimum intensity in the same plot, which is also referred to as the anisotropy value in this paper. We found that the background noise has the smallest anisotropy value of less than 1.032, which is attributed to the slight difference in the refractive index of agarose under different polarizations. The intensity modulation of the fluorescence emission from microspheres is similar to the one from the background, since the refractive index difference plays the same role in the fluorescence emission from the microspheres by changing the incident light intensity. The anisotropy values of a single microsphere (Inset 1, 2) or cluster of few microspheres (Inset 3, 4, 5) are between 1.032 and 1.045, while for a big cluster of microspheres (Inset 6, 7) are between 1.067 and 1.074. These values indicate the difference in fluorescence intensity due to different polarizations is about 4$\%$ or 7$\%$, meaning that the fluorescence emission hardly depends on the excitation polarization. In our experiment, the microspheres were prepared with agarose, which restricted their movement. The fluorophores embedded in the microspheres are constrained by their bindings and lack the degree of freedom. Therefore, we believe that nearly isotropic fluorescence emission comes from a population of randomly-oriented fluorophores. Despite weak, a single fluorescent microsphere still presents anisotropic fluorescence emission. With more and more carboxylate-modified microspheres bound by chemical bonds, we observed that the anisotropy value increases with the increase of cluster size, showing a moderate preference for light polarization and fluorescence emission direction.

We then examined the polarization effect on GCaMP6f-expressed zebrafish larvae utilizing the same self-built LSFM system. GCaMP6f is a genetically encoded fluorescent Ca[2+] indicator, consisting of a circularly permuted green fluorescent protein (GFP) and calmodulin, and expresses under a pan-neuronal promoter. Figure. 3(a), (c), and (e) display the fluorescence images of three different parts of zebrafish larvae. They were all recorded under p-polarization with maximized fluorescence signal. Similarly, we analyzed three ROIs with sizes of 100 by 100 pixels, two from the emission regions and one from the background. The intensity changes as a function of HWP angles within certain ROIs are plotted in Fig. 3(b), (d), and (f) in the same color as the square markers of ROIs in the fluorescence images to the left. The solid and hollow data points are the control results with the laboratory light turned on and off, respectively, during the record of fluorescence images. The anisotropic values are listed vertically in the same order as solid data, the values after the slash are for the hollow data of the same color. As shown in Fig. 3(b, d, f), the solid and hollow data points in blue color are overlapped, indicating that the intensity of background noise is unrelated to the laboratory lighting. The intensity gap between the light on and off from the other two ROIs (in color red and magenta) is mainly due to the reflection of the laboratory light from the zebrafish larva surface. Due to both the photoselection effect and the non-coaxial configuration of LSFM, the intensity of fluorescence emission taken at p-polarization is about 38-47$\%$ stronger than the one taken at s-polarization. This phenomenon has also been observed recently by another group [22]. The sharp sparks, e.g. the magenta solid data point at HWP=6^∘ in Fig. 3(b), signify the neuronal signal transmission that happened inside the magenta square during the 50 ms exposure time. These sparks affect the smoothness of the polarization correlation curve but without changing the trend of modulation. We have the same experiments repeated on two other zebrafish larvae and obtain comparable anisotropic emission properties. Considering the particularity of the biological samples, the anisotropy value can vary from 1.3 to 1.5.

According to the theoretical model, for single-photon excitation, the probability of excitation is proportional to cos${}^{2}\theta$, where $\theta$ is the angle between the transition dipole moment of the molecule and the polarization orientation of the excitation light [15]. We fit the experiment data from the emission regions with a cosine square function of HWP angles in solid and dash lines for light on and off, respectively, as shown in Fig. 3(b, d, f). In general, the data points are in good agreement with the theoretical model with the maximum fluorescence intensity $I_{\text{p}}$ observed at p-polarization and minimum intensity $I_{\text{s}}$ at s-polarization. However, the amount of fluorescence at s-polarization is more than what we expected. As sketched in Fig. 1(b), when the excitation light is s-polarized, the favored transition dipole moment oriented along the detection axis fluoresces predominantly in the xy-plane. The detection objective used in the experiment has a collecting angle of 37^∘ and can only gather a small portion of the fluorescence signal. We infer that the considerable amount of fluorescence at s-polarization is caused by the rotational diffusion of the fluorescent molecules, which is the dominant depolarization process that changes the direction of transition dipole moments. This intensity part is expressed as $I_{\text{depol}}$, also exists in p-polarization.

The ratio between p- and s-polarization reflects the fraction of the molecules attached by the fluorescence tag with a preferential direction. According to the electric dipole radiation, the intensity observed at angle $\alpha$ is proportional to $sin^{2}\alpha/r$, where $\alpha$ is the angle between the dipole orientation and the observation point, and $r$ is the distance from the dipole center to the observation point. In our experiment, the intensity signal is then the integral of $sin^{2}\alpha/r$ with the integration limits corresponding to the 37^∘ collection angle. Therefore, for the same amount of fluorophores, the intensity ratio of the two representative schemes in Fig. 1(b) is calculated to be $I_{\text{p}}/I_{\text{s}}\sim$2.25. If we consider a simple case that $(I_{\text{p}}+I_{\text{depol}})/(I_{\text{s}}+I_{\text{depol}})\approx 1.4$, then the number of molecules with dipole orientations along with p-polarization $N_{\text{p}}$, s-polarization $N_{\text{s}}$ and random orientations $N_{\text{depol}}$ would satisfy the relation as $2.25N_{\text{p}}-1.4N_{\text{s}}=0.4N_{\text{depol}}$. However, the actual situation of living animals is much more complicated. The fluorescence intensity detected under each linear-polarized light is the superposition of the fluorescence of molecules with different dipole moment orientations. In addition, the resonance effect would become an important consideration when the length of the dipole is comparable to the excitation wavelength.

As we have observed the difference in intensity due to the anisotropic fluorescence, which interests us next is whether this anisotropic fluorescence also exhibits a preference for certain regions within the biological sample. To explore this question, one needs to separate the fluorescence emission according to the excitation polarization. For this purpose, the fluorescence images taken at both s-polarization and p-polarization from the same recording sequence were registered at single-cell resolution to a reference image, which is an image also taken at s-polarization but at a different time. Then, the reference image was subtracted from the registered fluorescence images at two representative polarizations without normalizing the initial intensity. The resulting images for three parts of the larval zebrafish are shown in Fig. 4. Since the zebrafish experiments were conducted in vivo, subtraction of s-polarization images and the reference image was performed to provide a comparison to identify that the difference was induced by polarization rather than any occasional biological activity.

The upper and lower rows in Fig. 4 show the reference image subtracted from p-polarization and s-polarization, respectively. Compared to the original images recorded under p-polarization (see Fig. 3(a, c, e)), the subtracted images become flocculent. We note that the enhanced fluorescence due to anisotropic emission can be observed throughout the whole emission region and does not show a preference for specific biological sites. For Fig. 4(b, d, f), the subtractions of s-polarization images and the reference image, the difference is considered as noise signals, which are uniformly scattered over the whole fluorescence emission region, except for the hot spots/sparks arising from the neuronal transmission.

In order to further dissect the cause of anisotropic fluorescence, arbitrarily selected areas of 400 by 400 pixel size from fluorescence images taken under p-polarizations before and after subtraction are zoomed in Fig. 5, where Fig. 5(a, d, g) are from Fig. 3(a, c, e) and Fig. 5(b, e, h) are from Fig. 4(a, c, e). It can be seen that the biological structures become sharper in the subtracted images. The normalized intensity profiles along the blue and red lines in Fig. 5(a, b), (d, e), and (g, h) are plotted in Fig. 5(c), (f), and (i) in the same color, respectively. With the normalized peak height, in all three graphs, the fluorescence intensity after subtraction (blue lines) has a much low value at the valley than the fluorescence intensity before subtraction (red lines). Reflecting on the images, this suggests that the anisotropic fluorescence emission due to the polarization on the GCaMP6f-expressed zebrafish larvae mainly occurs within the nerve cells, rather than intercellular or extracellular substances. This observation is rational since the GCaMP is commonly used to measure increases in intracellular Ca²⁺ in neurons as a proxy for neuronal activity. While the fluorescence signal observed in the extracellular space may come from the scattered-fluorescence, auto-fluorescence, Poisson noise, etc, which are usually regarded as the noise signal and degrade the image quality [25]. Therefore, based on this phenomenon, a fluorescence image with better contrast and resolution can be obtained by subtracting images under different polarizations. Herein, we use the Michelson formula to quantify the contrast [26]. The contrast $C$ is calculated as $C=(I_{\text{max}}-I_{\text{min}})/(I_{\text{max}}+I_{\text{min}})$, where $I_{\text{max}}$ and $I_{\text{min}}$ are the intensity values of each peak and its corresponding valley, respectively. By averaging $C$ values of main peaks in the graph, we obtain the gross contrast with and without differential polarization subtracting method to be 0.30 and 0.13 (Fig. 5(c)), 0.61 and 0.35 (Fig. 5(f)), and 0.77 and 0.37 (Fig. 5(i)), respectively. Hence, an improvement of $\sim$2 times the image contrast is achieved. And this improvement is more significant in biological structure with many cells. Last but not least, in the first 20-pixel distance in Fig. 5(c), extra biological structures can be resolved into distinct objects. Elimination of depolarized fluorescence emission enables one to distinguish the boundaries of nerve cells more clearly, which is promising for future research with a superior requirement for structural information.

We have successfully examined the effects of excitation light polarization on fluorescence emission using self-built light-sheet microscopy. Zebrafish larvae (GCaMP6f-expressed) embedded in low-melting-point agarose have been used as the biological organism to explore the anisotropic fluorescence emission. As expected, the levels of the fluorescence signal follow the cosine square function with respect to the polarization orientation of the excitation light. An approximate 40$\%$ increased fluorescence emission is observed when the excitation light is p-polarized, which principally excites the molecules with transition dipole moments lined up in the direction perpendicular to the detection axis, therefore emitting the most fluorescence toward the direction of the detection objective. This provides a straightforward way to adjust the fluorescence intensity and obtain the maximum signal during the light-sheet imaging experiments. From the ratio of the fluorescence emission of p- and s-polarization, we discuss the orientation of the molecules the fluorescence tag attached to, which would be interesting for biologists to further investigate nerve cells based on the intrinsic properties of the biological structure. We have further analyzed the origin of anisotropic fluorescence emission via image processing. After the registration and subtraction of the fluorescence images of various polarization, we find the anisotropy is driven by the fluorescence signal arising in the nerve cells rather than the extracellular substance. The fluorescence emission from the extracellular substance is comparable under p- and s-polarization, and thus can to a large extent be eliminated through subtraction, leaving a fluorescence image with improved contrast and increased resolution of detached cell structure. We have demonstrated that using this differential polarization subtracting method, one can obtain a clearer structural fluorescence image of twice the contrast and resolve more details, which is propitious for future research relying on microscopy imaging techniques.

Funding This work is supported by the "Strategic Priority Research Program" of the Chinese Academy of Sciences (grant no. XDA16021304) and the "Doctor of entrepreneurship and innovation program" in Jiangsu Province (grant no. JSSCBS20211440).

Acknowledgments The authors thank Shan Zhao, affiliated with the Institute of Neuroscience, Chinese Academy of Sciences, for preparing the larval zebrafish samples.

Disclosures The authors declare no conflicts of interest.