2.1.

Sample Preparation

Retinal samples were prepared in a dark room with dim red light illumination. After 4 h of dark adaptation, eyes were enucleated from euthanized frogs or mice. The eyeball was hemisected along the equator with fine scissors and the anterior structures were removed from the retina. The retina was separated from the retinal pigment epithelium (RPE). The isolated retina was then cut into retinal slices with $150\mu \mathrm{m}$ thickness. Without RPE, retinal slices provided a simple preparation to enable easy observation of individual photoreceptors and quantitative calculation of photoreceptor movements. The entire sample preparation procedure was performed in Ringer’s solution containing $110.0\mathrm{mM}/\mathrm{L}$ NaCl, $2.5\mathrm{mM}/\mathrm{L}$ KCl, $1.6\mathrm{mM}/\mathrm{L}$ ${\mathrm{MgCl}}_2$, $1.0\mathrm{mM}/\mathrm{L}$ ${\mathrm{CaCl}}_2$, $22.0\mathrm{mM}/\mathrm{L}$ ${\mathrm{NaHCO}}_3$, and $10.0\mathrm{mM}/\mathrm{L}$ D-glucose.^14–16 Retinal slices were immersed in a chamber filled with Ringer’s solution and placed under an NIR light microscope for optical imaging. Retinal slices were illuminated from the side, which comprised a cross-section of retinal photoreceptors and other functional layers. All experiments were performed following the protocols approved by the Animal Care Committee at the University of Illinois at Chicago.

2.2.

Experimental Setup

An NIR light microscope (BX531WI, Olympus, Japan) with a $40\times$ NA 0.8 water immersion objective lens was used for this study. Figure 2 shows a schematic diagram of the experimental setup that has been used for intrinsic optical signal (IOS)^17,18 and TRP¹⁰ studies in freshly isolated retinas. A high-speed camera (Neo 5.5, Andor Technology Ltd., Belfast, Ireland) was used for dynamic NIR microscopy imaging. The NIR light for optical imaging was produced by a halogen lamp with a band-pass filter (wavelength band: 775 to 1000 nm). The stimulus light was provided by a fiber-coupled light-emitting diode with a central wavelength of 550 nm (wavelength range: 450 to 650 nm). The retinal slices were stimulated by a rectangular shape stimulus light pattern. The width of the stimulus light pattern was controlled by an adjustable slit in the stimulus path. Adjustable light intensity was controlled with neural density filters. For illustrated measurements in this article, we selected the following stimulus intensities: $\sim 1.0\times {10}^7$, $1.0\times {10}^8$, $1.0\times {10}^9$, and $1.0\times {10}^{10}\text{photon}/{\mu \mathrm{m}}^2·\mathrm{s}$, with 1 s stimulus duration. All illustrated images in this article were captured at a speed of $100\text{frames}/\mathrm{s}$, with a frame resolution of $512\times 512\text{pixels}$. The entire experiment for each retinal preparation was completed within 2 h after animal euthanasia.

Fig. 2

Schematic diagram of experimental setup. The NIR light was used for retinal imaging. The visible light stimulator was used for retinal stimulation. A slit was used to limit the visible stimulus light into a rectangle shape. In front of the high-speed camera, the NIR filter was used to block visible stimulus light.

2.3.

Data Processing

Optical flow is a well-established method for calculating target movements between two images. It identifies movements in sequential images, which makes the method suitable for tracking movements of retinal cells.¹¹ To verify the photoreceptor displacement, we adopted optical flow MATLAB^® software developed by Sun et al.¹⁹ Stimulus-evoked photoreceptor movement was quantified in magnitude and direction maps by comparing the images before and after the stimulation. In order to automatically segment active areas and inactive areas in a displacement magnitude map, Otsu’s thresholding method²⁰ was used. The active area is the location with detectable photoreceptor displacement, while the inactive area is the location without detectable displacement. Segmented active areas were color coded in the direction map.

Centroid detection algorithm was used to study the stimulus-evoked OS changes’ kinetics of individual rod photoreceptors. A similar algorithm has been well established for single molecule localization to achieve super-resolution imaging, such as photoactivated localization microscopy (PALM) and stochastic optical reconstruction microscopy, at nanometer resolution.^21–24 In this study, a small rectangular window was drawn on the tip of the rod OS and the intensity centroid of the window was tracked in each frame of the video. If the background of the image was completely black (i.e., zero intensity), the centroid movement in the vertical direction should be exactly half of the physical length change of the photoreceptor tip; i.e., if the photoreceptor was shrunk by 2 pixels in the vertical direction, the centroid should move 1 pixel in the vertical direction. However, the background of the image recorded with microscopy was not totally black. In fact, the contrast between the photoreceptor and background was poor due to the transparency of retinal tissues. Because the centroid detection method is sensitive to the background level, it requires background correction for reliable measurement. The performance of the centroid detection method can be improved by using a simple background filtering approach or by using a more robust background-corrected centroid estimation method, such as super-resolution analysis tools (QuickPALM).²¹ We resolved this problem by calibrating each photoreceptor in the first frame of each image sequence. First, a small rectangular window was drawn which covers approximately half background and half photoreceptor tip as shown in Fig. 3(a). The window was shifted up to 4 pixels above and 4 pixels below the original position, with one pixel interval in the vertical direction. At each position of the selected window, the centroid was calculated. The centroid movement in the vertical direction was plotted against the magnitude of the window shift in the vertical direction. The relation between these two variables was almost linear. A best fit line was plotted and the slope of the line was calculated. The slope of the line provided the relation between the photoreceptor movement and centroid movement when the background was nonzero.

Fig. 3

Effect of the background on the centroid calculation. The first row shows a simulated photoreceptor at three different contrasts compared to its background: (a) high contrast, (b) intermediate contrast, and (c) low contrast. The calibration factor was calculated by shifting a small rectangular area [i.e., a region of interest (ROI)] one pixel at a time in the vertical direction. The second row shows shifting of x-coordinates of the centroid when ROI shifts in the vertical direction. The calibration factor for the high-contrast image (a) was 0.50, the calibration factor for the intermediate-contrast image (b) was 0.14, and the calibration factor for the low-contrast image (c) was 0.05.

We tested this technique on a simulated rectangular photoreceptor created with intensities $128+(\text{random values}\times 10)$, as shown in Fig. 3(a). The centroids were calculated by placing the photoreceptor in three different (high, intermediate, and low) background conditions. First, the simulated photoreceptor was placed on a completely black background [Fig. 3(a)]; i.e., all the background pixels had zero intensity. The calibration factor was 0.50, which indicated that the photoreceptor movement was two times that of the centroid movement. Second, the simulated photoreceptor was placed on backgrounds with intensities that were nearly half of the photoreceptor intensity [Fig. 3(b)]. The background was formed by creating pixel values with intensities $60+10\times \text{random variables}$ to create slightly different intensities in the background region. The calibration factor for intermediate contrast conditions was 0.14, which indicated that photoreceptor movements were nearly seven times that of the centroid movement. Third, the simulated photoreceptor was placed on the background with a slightly lower intensity ($100+10\times \text{random variables}$) than that of the photoreceptors [Fig. 3(c)]. The calibration factor for low-contrast conditions was 0.05, indicating that the photoreceptor movement was 20 times of the centroid movement. The aforementioned centroid displacement method was used to calculate the stimulus-evoked OS change kinetics of individual rod photoreceptors of retinal slices. To verify the reliability of the observed rod OS changes, we repeated the measurements with nine retinal slices under identical experimental condition.

3.1.

Video Microscopy of Stimulus-Evoked Rod Outer Segment Changes

Figures 4 and 5 illustrate representative measurement with mouse and frog retinal slices, respectively. Individual mouse photoreceptors can be ambiguously observed in Fig. 4(a), and stimulus-evoked OS change of a single mouse rod photoreceptor (yellow arrowhead) is shown in Fig. 4(b). Compared to a mouse photoreceptor, individual photoreceptors can be easily observed in Fig. 5(a1). Movie clips associated with Figs. 4(a) and 5(a1) revealed rapid rod OS length and orientation changes in mouse and frog retinas. Each movie represents a 5.0 s time-lapse microscopy video clip that consists of a 0.5 s prestimulus, 1.0 s stimulus, and 3.5 s post-stimulus recordings. The magnitude and direction of rod OS change varied with the photoreceptor position within the stimulation area. In the center of the stimulation region, the length of the OS shrunk, while in the peripheral region, the OS swung toward the center of the stimulation area in the plane perpendicular to the incident stimulus light. The diameters of mouse photoreceptors (both rods and cones $\sim 1$ to $2\mu \mathrm{m}$) were relatively smaller than those of frog photoreceptors (rod: $\sim 5$ to $8\mu \mathrm{m}$, cone: $\sim 1$ to $3\mu \mathrm{m}$).^25,26 For easy calculation of individual photoreceptor movements, we used frog retinal slices for quantitative analysis of rod OS changes at the center and periphery of the stimulus pattern.

Fig. 4

(a) Stimulus-evoked mouse rod OS movement (Video 1, MOV, 56 KB) [URL: http://dx.doi.org/10.1117/1.JBO.21.6.065006.1]. The yellow window indicates the stimulation area. (b) OS changes of a mouse rod photoreceptor (yellow arrowhead) at the center of the stimulus window.

Fig. 5

(a1) Stimulus-evoked frog rod OS movement (Video 2, MOV, 64 KB) [URL: http://dx.doi.org/10.1117/1.JBO.21.6.065006.2]. The yellow window indicates the stimulation area. Corresponding optic flow visualization results are shown in (a2) and (a3). (a2) Rod OS magnitude map derived from the image and (a3) rod OS direction map. Inactive area mapped by black background. Active/inactive areas are separated according to the displacement magnitude using Otsu’s thresholding method. The active area is the location with detectable photoreceptor displacement, while the inactive area is the location without detectable displacement. The distribution around 0 and 180 deg represents moving toward right and left, respectively, while that around 90 deg represents shrinkage. Stimulus-evoked OS changes in frog rod photoreceptors at the left edge (rod 1), center (rod 2), and right edge (rod 3) of the window are shown in (b1), (b2), and (b3), respectively. At the center of the stimulus pattern, the rod OS tip movement was predominated by shrinkage, i.e., OS length reduction. At the left and right edges of the stimulus pattern, the rod OS tip movement was predominated by a shift toward the center of the stimulus.

3.2.

Quantitative Comparison of Transient Rod Outer Segment Changes

Figures 5(a2) and 5(a3) show the magnitude and direction map derived from the two microscopy images taken before and after 100 ms of the stimulation using the optic flow method described in Sec. 2. The magnitude map is color coded [Fig. 5(a2)]. The direction map revealed that photoreceptors located at the left edge of the stimulation center moved toward the 0 deg direction (right direction), whereas photoreceptors located at the right edge of the stimulation center moved toward 180 deg (left direction). While the optical flow map revealed robust rod OS shifts at the right and left edges of the stimulus pattern, it was not sensitive enough to disclose the small rod OS shrinkage at the center of the stimulus pattern. Figure 5(b) illustrates stimulus-evoked OS changes’ kinetics of individual rod photoreceptors. At the left edge (rod 1) of the stimulus pattern, the rod OS tip movement was predominated by a $0.77\mu \mathrm{m}$ shift to the right [Fig. 5(b1)]. At the center (rod 2) of the stimulus pattern, the rod OS tip movement was predominated by shrinkage of $0.14\mu \mathrm{m}$, i.e., OS length reduction [Fig. 5(b2)]. At the right edge (rod 3) of the stimulus pattern, the rod OS tip movement was predominated by a $1.11\mu \mathrm{m}$ shift to the left [Fig. 5(b3)].

We repeated the measurements with nine retinal slices under identical experimental conditions. The average transverse movement (tip shift) and shrinkage (length reduction) of nine left-edge, nine center, and nine right-edge rods are shown in Fig. 6. Our results revealed that the peripheral rod OS tip shifted toward the center of the stimulus pattern, while the center rod OS length reduced unambiguously. As shown in Fig. 6, the average shrinkage and shift of the photoreceptors located at the center of the stimulation area were $0.16\pm 0.06$ and $0.06\pm 0.02\mu \mathrm{m}$, respectively. Similarly, the average shrinkage and shift at the left peripheral region of the stimulation area were $0.21\pm 0.10$ and $0.48\pm 0.25\mu \mathrm{m}$, respectively, and the average shrinkage and shift at the right peripheral region of the stimulation area were $0.16\pm 0.05$ and $0.38\pm 0.16\mu \mathrm{m}$, respectively. The average shift of the photoreceptors located at the center of the stimulation area was smaller than the average shrinkage. In contrary, the average shrinkage of the photoreceptors located in the peripheral regions was smaller than the average shift. The left and right photoreceptor shifts were significantly different ($p<0.01$) from that of the central photoreceptor, whereas the remainder of the comparisons were not statistically different.

Fig. 6

Averaged stimulus-evoked OS changes from nine rod photoreceptors at the left edge, center, and right edge of the stimulus pattern. The OS tip movement was predominated by a shrinkage response at the center of the stimulus pattern and by a shift response at the peripheral region of the stimulus pattern.

3.3.

Effect of Stimulus Strength on Rod Outer Segment Changes

To further understand the stimulus-evoked photoreceptor kinetics, we measured the effect of stimulus light intensity on the rod OS shift response. Individual photoreceptors at the left/right edges of the stimulus pattern were selected for measurement. Experimental data were collected from 12 retinal slices with four different stimulus intensities. For each stimulus intensity, three retinal slices were measured. We measured the latency time (the time delay before detectable response), peak time (the time to peak magnitude), and peak amplitude. Figure 7 shows the averaged rod OS responses (latency, peak time, and peak amplitude) corresponding to four stimulus intensities. It was observed that the peak amplitude increased with enhanced stimulus intensity [Fig. 7(b)]. However, the latency decreased with enhanced stimulus intensity [Fig. 7(c)]. Similarly, the peak time, i.e., time to peak, decreased with enhanced stimulus intensity [Fig. 7(d)].

Fig. 7

Effect of stimulus strength on rod OS changes. (a) Rod OS changes elicited by different stimulus intensities, (b) averaged peak amplitude ($n=3$ rods) versus stimulus strength, (c) averaged latency ($n=3$ rods) versus stimulus strength, and (d) averaged peak time ($n=3$ rods) versus stimulus strength.