Standard-unit measurement of cellular viability using dynamic light scattering optical coherence microscopy

: Dynamic light scattering optical coherence microscopy (DLS-OCM) integrates DLS, which measures diffusion or flow of particles by analyzing fluctuations in light scattered by the particles, and OCM, which achieves single-cell resolution by combining coherence and confocal gating, integratively enabling cellular-resolution 3D mapping of the diffusion coefficient, and flow velocity. The diffusion coefficient mapping has a potential for the non-destructive measurement of cellular viability in the standard unit but has not been validated yet. Here, we present DLS-OCM imaging of intra-cellular motility (ICM) as a surrogate of cellular viability. For this purpose, we have simultaneously obtained and compared ICM-contrast DLS-OCM images and calcium fluorescence-contrast images of retinal ganglion cells, and then characterized the responses of the measured ICM to a change in cellular viability induced by environmental conditions such as temperature and pH. The diffusion-coefficient-represented ICM exhibits consistent changes with the manipulated cellular viability.

Label-free measurement of ICM has a potential to enable non-destructive assessment of cellular viability without any chemical injection or genetic manipulation and thereby facilitating longitudinal studies of drug effects or environmental cytotoxicity. The potential, however, has not been fully realized; previous label-free methods mostly measured ICM in an arbitrary unit such as the decorrelation of OCT signals [23,24], which is less suitable for quantitative comparison across different samples or systems than a standard unit-represented measurement, and/or have not sufficiently shown that the measured ICM indeed represents the cellular viability by testing if it varies with a change in viability [19,20,[25][26][27]. Previously, we used dynamic light scattering optical coherence tomography (DLS-OCT) [19] and found that neurons exhibit high diffusion coefficients [20], but have not tested if they vary with viability changes. Here, we present label-free, cellular-resolution measurement of ICM in the standard unit of diffusion coefficient and demonstrate that it correlates with and thus represents the cellular viability. For this purpose, we used DLS-OCM that enables us to produce a micrometer-resolution, three-dimensional map of the diffusion coefficient, optimized the method to match its dynamic range with the typical range of ICM, tested if the diffusion coefficient map can visualize cells based on the contrast of ICM by comparing it with simultaneously-imaged intracellular calcium fluorescence, and then investigated how the measured ICM varies with a change in the cellular viability induced by altering conditions with well-known effects such as temperature and pH. We expect that the DLS-OCM assessment of cellular viability in the standard unit will be widely utilized in various studies from drug and toxin tests using human tissue spheroids to longitudinal disease development researches using animal models.

Simultaneous OCM and fluorescence imaging
A commercial spectral-domain (SD) OCT system (Thorlabs, Newton, NJ, USA) was used and modified for this study. The system uses a large-bandwidth near-infrared light source with a center wavelength of 1310 nm and a wavelength bandwidth of 170 nm, which leads to high axial resolution (3.5 μm). 40x objective lens (1-U2M587, Olympus America, Inc) was used for cellular imaging with the lateral resolution of 0.78 μm. The system uses a high-speed 2048-pixel line-scan camera to achieve 147,000 A-scan/s with a relatively large imaging depth (2.5 mm maximum). The field of view (FOV) has 256 (X) × 256 (Y) × 1024 (Z) voxels with the transverse and axial sampling rates of 0.5 μm and 3.46 μm, respectively, leading to 128 μm × 128 μm × 3.54 mm imaging volume. Due to the high numerical aperture, the effective depth of field was ~15 μm (twice of the Rayleigh length). To compare label-free DLS-OCM images to one of the current standards for cellular imaging, GCaMP3 fluorescence imaging that needs no exogenous calcium dye [28], we built a wide-field fluorescence microscope and combined it to the OCM system, as shown in Fig. 1. Blue light with nominal wavelength of 490 nm from the light emitting diode (LED) and driver (M490L4 and LEDD1B, Thorlabs) travels through a diffuser (ACL2520UDG6, Aspheric Condenser Lens, Thorlabs), an excitation filter (MF469-35, GFP Excitation Filter, Thorlabs), and a dichroic beamsplitter (69-899, Dichroic Longpass Filter, Edmund Optics) to a sample. Green fluorescence light (typically 500-540 nm in wavelength [28]) emitted from the sample travels through the beamsplitter and emission filter (FELH0500, Premium Longpass Filter, Thorlabs) to a camera (OCTG-1300, Thorlabs). Images were acquired and displayed in real time on a computer using custom-built LabVIEW software. The specific excitation and emission wavelength ranges were chosen to enable simultaneous OCT and fluorescence imaging while minimizing crosstalk. We aligne 10, Fluoresbr suspended in imaged with t data and the lumps of fixed that more flu fluorescence m uses additiona intensity proje images could plane and ima size) compari

DLS-OC
The dynamic function of a A-scan rate) and the maximum time lag in the autocorrelation function. This study uses a faster OCT system (147,000 A-scan/s) than the one used in our previous publication (47,000 A-scan/s) and specifically focuses on diffusion-coefficient mapping of cellular viability, whereas the previous study focused on distinctive mapping of the blood flow and intracellular diffusion [20]. For this reason, we optimized our DLS-OCM measurement of the diffusion coefficient through numerical simulation and phantom experiment, to match its dynamic range for the reported values of ICM (0.1-10 μm 2 /s [19,20,25]) prior to its application to biological specimen imaging.
The numerical simulation tested combinations of three cases of the data acquisition time (0.68, 1.4, and 2.7 ms, corresponding to the numbers of A-scans repeated at each X-Y position, 100, 200, and 400 A-scans, respectively) and two cases of the maximum time lag (1/2 and 1/4 of the acquisition time), to find which combination results in the most accurate estimation of the diffusion coefficient over the range of 0.3-30 μm 2 /s. Five diffusion coefficients in the range were chosen with even intervals in a log scale. A longer acquisition time may be more advantageous to measure lower diffusion coefficients, so the range of data acquisition time was chosen based on our initial guess that it is appropriate for the diffusion coefficient range. The detailed simulation procedure was similar to the one we previously published [19]. In short, for each given set of parameters (including the diffusion coefficient, number density of diffusing particles, velocity of flowing particles (as a potential noise), and the number density of static particles within the OCM resolution volume), we generated timeseries position data of the particles, determined true diffusion coefficient and flow velocity from the position data, generated a time-series OCM signal from the position data, obtained its autocorrelation function, fitted it to our model to estimate the diffusion coefficient and flow velocity, and then compared them with the true values. We repeated this simulation for 100 random initial positions of the particles, repeated it for 1,050 parameter sets (5 diffusion coefficients, 5 flow velocities, 7 flow angles, and 6 number densities), and repeated the analysis again for six cases of the acquisition time and maximum time lag. As a result, the 0.68-ms acquisition time and the 0.17-ms maximum time lag produced the minimum rootmean-square error in the diffusion coefficient measurement (1.7%) when more than 60% of the particles in the OCM resolution volume exhibit diffusive movements and the flow velocity as noise is lower than 10 mm/s ( Fig. 3(a)). Based on this optimization, in the following DLS-OCM imaging, 128 A-scans were repeated at a fixed position, it was then moved to the next scanning position, and their autocorrelation functions with the 1/4 maximum time lag (0.22 ms) were obtained and processed, producing a 3D map of the diffusion coefficient of a sample.
The simulation was followed by a phantom experiment to confirm the accuracy and dynamic range of DLS-OCM measurements. As a standard sample, microspheres with various sizes were used because the diffusion coefficient and diameter of microspheres have an inversely linear relationship as given by Stokes-Einstein equation [29], D = k B T/8πηr 3 where k B = 1.38 × 10 −23 J/K, T = 293 K, η = 1.00 cP. Monodisperse polystyrene microspheres in 2.5% solids (w/v) aqueous suspension (Polysciences, Inc.) with diameters of 0.05, 0.088, 0.15, and 0.22 μm were used. Following the simulation results, repeated A-scans were obtained to acquire a 3D volume data of the sample. From the raw OCT-spectrum data, 4D complex-valued SD-OCT reflectivity signals ( ( , ) R t r ) were computed, from which a field autocorrelation function was generated as following; Here, E[ ] denotes averaging over initial positions, * denotes the complex conjugate, and denotes the time averaging. The field autocorrelation function was then convolved with a 3 × 3 × 3 (voxels) spatial kernel to average over initial positions of particles and fitted to the previously developed model shown in Eq. (2) [19] to derive the dynamic parameters of the sample (see [19] for details).
Here, the scatterers are assumed to be static or flowing in an OCT resolution volume, and the flowing particles should exhibit translational motion with a velocity or diffusive motion with a diffusion coefficient. The dynamic parameters obtained are the fraction of static particles, five dynamic parameter maps were generated which were then convolved with a 5 × 5 × 5 (voxels) Gaussian kernel. As shown in Fig. 3(b), our DLS-OCM makes relatively accurate measurements (80.4% in accuracy) while it becomes inaccurate for higher diffusion coefficients than ~8 μm 2 /s. This upper limit of the dynamic range is attributed to the limited temporal sampling rate of our OCT system (0.68 μs), so one may need a faster system to increase the dynamic range, although the current range is suitable for our purpose because the reported typical ICM ranges from 0.1 to 10 μm 2 /s [19,20,25]. Finally, we determined the lower limit of the dynamic range using a static sample (WS-1, Oceans Optics) to 0.05 μm 2 /s.

DLS-OCM imaging of mouse retina with a genetically encoded calcium indicator
As a biological specimen to test DLS-OCM measurements of cellular viability, intact retinal tissues dissected from mice were used. The retinal explant was chosen for this study because it has well-known laminar cellular distribution, it is relatively easy to be encoded with genetically-induced calcium fluorescence indicators [28], which is advantageous for the simultaneous fluorescence and DLS-OCM imaging, and it is closer to in vivo than cell culture as it has other tissue constituents as well (e.g., extracellular matrix). Thus, we established a colony of GCaMP3 mice (B6.Cg-Tg(Thy1-GCaMP3)6Gfng/J) and euthanized a mouse following our IACUC-approved protocol before dissecting retinas. For the retinal dissection, a mouse was anesthetized with 3-5% isoflurane with a mixture of 100% oxygen in the air. Immediately after the initiation of anesthesia, pharmaceutical grade Dexamethasone (0.2 mg/kg) was administered to decrease inflammation, and buprenorphine (0.05-0.1 mg/kg) was administered to ensure pain relief. Whole retina was harvested directly from the euthanized animal by puncturing the eye, cut out the cornea, and cut around the eye at the limbus to release the retina. Then, the retina was freed from the eyecup, detached from the optic nerve, cut in a clover shape to put on a filter (Millicell Cell Culture Insert, EMD Millipore Corp.), and then sucked from the other side to be fixed at the surface (Fig. 4(c)). This fixing method is widely used in retinal physiology experiments to keep a retinal explant in place even when the medium is slowly perfused. While we repeated the experiment, we visually checked the position of the retina sample between OCT data acquisitions and found that the fixing method works well in our setup as well. The sample and filter were placed in our imaging chamber (Figs. 4(a-b)), where the explant was supplied with Ames medium (Sigma-Aldrich Corp.) bubbled by O2 (5% CO2), through a syringe pump (AL-1000, WPI Inc.) for precise perfusion rate. The controlled flow went through a heater and controller (SH-27B, TC-324C, Warner instrument) to maintain the physiological temperature (31°C). All animal-based experimental procedures were reviewed and approved by the Institutional Animal Care and Use Committee of Brown University, according to the guidelines and policies of the office of laboratory animal welfare and public health service, National Institutes of Health.
To investigate how the measured ICM consistently varies with change in cellular viability, we manipulated the viability by altering conditions with well-known effects on it: the temperature and pH. DLS-OCM data were acquired before and after the condition changes (30 min later exposure to the change). The routine metabolic rate decreases at low temperature [30]; thus, we lowered the temperature to 15°C by packing the chamber surroundings with ice packs, to decrease the metabolic rate of retinal cells. On the other hand, cellular oxygen consumption decreases at low pH [31], so we decreased the pH to 3.5 by adding hydrogen chloride to the continuously perfused Ames medium.

DLS-OC coefficient
Simultaneous intact retina fluorescence 5(a)), the diff higher diffusi the cells are n difference in t the red arrow of soma (10coefficient ma ith the contr GCaMP3-encod aging system. with intrinsic c ows that the ce (Fig. 5(b)

CM-measured
in reaction to Fig. 6. In a r CM at low tem arent in the hi different betwe ween the initial 6. Finally, Fig. 8 summarizes statistical percentage changes in the measured ICM from the physiological condition to the changed condition. Both pH change and lowered temperature have produced statistically significant negative changes in ICM (−30% ~-40%; p = 0.016 and p = 0.034, respectively). The reheated case has shown a slight increase in ICM in comparison to that of the initial normal temperature, but it is statistically insignificant (p = 0.28). These results support that DLS-OCM-measured ICM can serve as an effective estimate of the cellular viability.

Discussions and conclusion
This paper proposes DLS-OCM measurement of ICM as a means of label-free, standard unitrepresented estimation of cellular viability. DLS-OCM has been optimized for the dynamic range matching with ICM, its diffusion coefficient map has visualized ICM of individual cells in the ex vivo retinal tissue as confirmed by comparison with simultaneously acquired coregistered fluorescence imaging, and finally, the measured ICM has consistently varied with the manipulated cellular viability.
When a cell is viable, active metabolism gives rise to active motions of intracellular organelles. Active and directional motions are primarily driven through cytoskeleton, which fluctuates and transports small particles. On the other hand, intracellular motions that are similar to diffusive motions, are also active motions consuming ATP that are different from passive diffusions [12,32]. Although small organelles exhibit directional motions, it can appear to be diffusive due to the size of the particle, the nature of the viscoelastic surroundings, and the force applied [33]. Some of the previous studies investigating the intracellular organelle movements used the standard unit of diffusion coefficient to quantify them. The previously reported diffusion coefficients in viable cells range from 0.1 to 10 μm 2 /s [19,20,25], and our previous study using DLS-OCT reported the diffusion coefficient of the neural cytoplasm in the range of 0.5-1 μm 2 /s. In this paper, ICM exhibits 1-4 μm 2 /s (see the histograms in Figs. 6 and 7), which is consistent, in the order of magnitude, with the previous reports even though they used different systems and biological specimen.
When the temperature of a tissue reaches a lower value than the physiological temperature, the active metabolism rate decreases [30]. Extremely low or high temperature 15°C reheated % change 0 ---are known to limit the oxygen delivery to mitochondria and cause changes to the anaerobic metabolism that has a lower efficiency of ATP reproduction compared to the aerobic metabolism [34]. In turn, with lower ATP, cellular energy levels also drop and the metabolic rate decreases [35]. The lowered metabolic rate will also lower the associated active motions of intracellular organelles (i.e., ICM). The previous study tracked polystyrene beads using in vitro motility assay and found that the active motion of microtubule-based transport decreased by 56% at a low temperature of 15 °C [36], which agrees with our measurements (Fig. 8). pH condition that is lower than a physiological pH surrounding a tissue results in a metabolism rate decrease, due to the oxygen consumption rate decrease and glucose utilization rate [31,37]. pH also affects cell membrane permeability by changing the net electric charge of the amino acids. A study showed that both oxygen and glucose consumption rate of brain slices decreased by ~20% and ~60%, respectively, after 20 minutes of exposure to pH 6.2 compared to pH 7.4 [37]. In accordance with the metabolism decrease in the lower temperature above, a decrease in the metabolism by lowering pH will decrease ICM. Our observation of the decreased diffusion coefficient under pH 3.5 is in harmony with the previous studies. With the acidic condition we have achieved with a low pH of 3.5, cells may have also undergone necrosis, which may also contribute to low ICM.
According to the Mie scattering theory, relatively large organelles with the size ranging from 0.1 to 10 μm (e.g., mitochondria and cytoskeleton) may dominantly contribute to our OCT signals (with the center wavelength of 1.3 μm) and thereby the measured diffusion coefficient [20]. Although Actin and microtubules are the two main drivers in the ATPconsuming intracellular motility, they are too small to substantially contribute to the OCT signals. It is currently unclear which organelles exactly are responsible for DLS-OCMmeasured ICM. A mechanism study that employs various drugs that affect a specific process targeting a specific organelle is required to answer the question. In addition, although this study shows the possibility of longitudinal monitoring of cellular viability changes in the same sample, which is difficult with destructive traditional cell assay methods, the present report does not demonstrate an additional interesting potential of tracking the viability cell by cell. When realized, the capability will enable unprecedented drug and toxin effect studies that track the effects over individual cells in a longitudinal manner (e.g., how an environmental toxin invades a human tissue spheroid and thus how the cells in its necrotic core respond differently to the toxin in time). In this study, the conditional changes were made through the continuously perfused medium, which had the sample slowly move and thereby made it difficult to track the exactly same cells between the imaging sessions 30-min apart. Therefore, another study using either a real-time visualization and adjustment of FOV (in both transverse and axial directions) or a treatment causing minimal movements is required to validate the capability. Finally, the current scanning and data processing time for our DLS-OCM may need further improvement for practical applications. For the FOV of 128 μm × 128 μm, our current system with 147,000 A-scan/s takes about 1 minute to repeat 128 A-scans at every X-Y position, which is faster than the previous DLS-OCT but still needs to be improved for high-throughput viability assessment on multiple microwells that is needed for drug and toxin effect studies. The current data processing time for the fitting of 5 coefficients takes 4 hours for the above FOV data when the algorithm we previously published [20] runs on our computer (two 2.30-GHz CPUs), which precludes the possibility of real-time analysis and visualization. A faster OCT, as well as a computationally-optimized algorithm for fitting, should be developed for the high-throughput application.
In conclusion, we have demonstrated the potential of DLS-OCM for non-destructive, standard-unit assessment of cellular viability. The technique can measure ICM of individual cells in biological tissue in the standard unit of diffusion coefficient, and the measured ICM has been shown to correlate with cellular viability. Such label-free, cellular-resolution measurement of viability will be useful for a wide range of drug and toxin effect studies.