Initiation of the optokinetic response (OKR) in mice

Abstract

To study the initial part of the mouse optokinetic response, OKR (∼500 ms from the onset of visual stimulus motion), we recorded the ocular response to a vertical sinusoidal grating moving at a constant velocity. We found that the magnitude of the response monotonically increased as the stimulus contrast increased. The response showed a narrow band-pass property for the spatiotemporal frequency, with the largest sustained response observed at 0.125 cycle/deg and 1.5 Hz. We also found that temporal frequencies higher than 1.5 Hz elicited transient increase in the eye velocity, but weak or no sustained eye movements. Thus the initial OKR in mice is characterized by the spatiotemporal frequency of the visual stimuli. Our results suggest that the initial OKR contains two components: a transient that diminishes within approximately 200 ms, and a tonic that is maintained for more than 400 ms, and that the initial part of the OKR in mice is an appropriate measurement parameter for studies of the visual and motor systems, like ocular following response (OFR) in primates.

Introduction

The optokinetic response (OKR) is an eye movement driven by certain visual stimuli. The input to the brain is a neural signal encoding information about visual motion on the retina, which is transformed to a motor command mainly via a subcortical neural pathway, i.e., the accessory optic system (AOS) and the nucleus of the optic tract (NOT; see review by Wallman, 1993). Therefore, the OKR plays an essential role in stabilizing visual images on the retinas of a number of species, including those in which the visual cortex is relatively less developed. Studies of the OKR have been performed in various animals, regardless of whether they are foveate or afoveate (see reviews by Büttner & Kremmyda, 2007; Collewijn, 1991; Distler & Hoffmann, 2003; Leigh & Zee, 2006; Stahl, 2004, 2008). 

In primates, the initial part of the OKR is considered to be closely related to the ocular following response (OFR; Miles, 1998). The OFR is a smooth eye movement characterized by an ultra-short latency: ∼50 ms in monkeys (Miles, Kawano, & Optican, 1986) and ∼70 ms in humans (Gellman, Carl, & Miles, 1990). Studies of the OFR neural circuitry have revealed that areas in the cerebral cortex, such as the medial superior temporal (MST) area, play significant roles in generating the eye movements (Kawano, Shidara, Watanabe, & Yamane, 1994; Takemura, Murata, Kawano, & Miles, 2007). Because the OFR is stereotypically smooth, time-locked eye movements with a short latency, the association of the visual input with the neural activity and oculomotor behavior is clear. In addition, the open-loop phase of the eye movement or of the neural activity reflects the pure dynamics of the response to the visual input. Therefore, the OFR has been used as a measure of how our central nervous system computes visual information and transforms it into smooth movements (see review by Chen, Sheliga, Fitzgibbon, & Miles, 2005; Kawano, 1999; Kawato, 1999; Masson, 2004; Miles, 1998; Takemura & Kawano, 2002). No study, however, has focused on the initial part of the OKR in rodents, in which the cerebral cortex is less developed than in primates. Therefore, it is still unclear whether the initial part of the mouse OKR can be used to study signal processing in the mouse brain. 

Recently, many studies have begun to use the OKR as a phenotype output of the mouse applied genetic engineering techniques that affect brain functions. For example, the effect of retinal abnormalities on the mouse OKR has been studied (Cahill & Nathans, 2008; Sato et al., 2008; Yoshida et al., 2001). However, none of these reports specifically examined the initial part of the OKR. Because eye movement affects visual motion on the retina, the relationship between these two parameters during the closed-loop phase is not clear in comparison with observations during the open-loop phase. Studying initial part of the OKR in mice may provide an opportunity to link results obtained via genetic engineering with visuomotor processing. This will facilitate our understanding of the cellular and molecular substrates of neural computations, not only in the retina, but also in other brain regions involved in the generation of the OKR. Furthermore, understanding the initial part of the mouse OKR will allow comparisons of visuomotor processing in mice, monkeys, and humans. Here, we have characterized initial stages of the mouse OKR as a behavioral measure. 

Classic experiments examining the OKR employed vertical black and white stripes or a similar high-contrast pattern on a cylinder (drum) that rotated around the animal. In the present study, we used computer monitors to present the visual stimuli, which allowed us to manipulate the stimulus parameters systematically. In addition, most of the classic experiments included observation of the OKR for 20 or 30 s. Long-lasting visual motion induced saccade-like resetting eye movements (the quick phase of the OKR) subsequent to the smooth compensatory eye movement. To eliminate the quick phase and focus on the initial component of the smooth eye movement, we observed the OKR for a maximum of 500 ms. 

Methods

Animals

The data were collected from 14 adult C57BL/6J mice, weighing 15–35 g. The mice were obtained from Oriental Bio Service (Kyoto, Japan). A stainless steel platform (Yamamoto Seisaku-syo, Kyoto, Japan) was glued to the cranial bone with dental cement under ketamine-induced anesthesia. In each experiment, a mouse was mounted on an experimental board by fixing the platform, and its body was supported with a rubber sheet to prevent excessive movement (Iwashita, Kanai, Funabiki, Matsuda, & Hirano, 2001; Sakatani & Isa, 2004). All experiments were based on protocols approved by the Animal Care and Use Committee of Kyoto University. 

Eye movement recording system

We recorded eye movements from the right eye of each animal. To monitor the eye movements, the right eye was illuminated using an infrared LED and the image reflected with a hot mirror (43957-J, Edmund) was monitored using an infrared-sensitive CCD camera (lens: VS-MC0510, VS Technology, Tokyo, Japan; controller: HR50, Sony, Tokyo, Japan; Figure 1A). We set a hot mirror with an azimuth of 60° in front of the animal. The hot mirror reflects infrared light but transmits visible light, such that the mouse can see the visual stimulus behind the mirror. Meanwhile, the hot mirror also transmits the visual stimulus in front of the mice. As a result of that, the camera does not only record the reflected infrared image but also transmitted visible light. To cut the transmitted visible light, we set an infrared filter (RT-830, Edmund) very close to the lens of the camera. By this, the camera could record only the reflected infrared image. The angles of incidence and reflection were separated by 90°. Thus, the camera recorded images of the eye at an azimuth of 60°. The camera, infrared filter, and hot mirror were fixed on the experimental board. 

Figure 1

 

(A) Experimental equipment used to record the mouse eye movements. Infrared images of the right eye were collected using a CCD camera. The position of the pupil was calculated using LabVIEW-based software (Get-Eye). (B) The system used to present the visual stimuli. To cover the entire visual field of the mouse, three LCD monitors were set around the test subject. Although we collected the data from the right eye, the mice looked at the visual stimuli with their both eyes. (C) Scheme of the experiment. A stationary visual pattern was presented, and then moved counterclockwise (temporal-nasal motion for the right eye) or clockwise (nasal-temporal motion for the right eye) at a constant speed. After a defined period, the pattern was removed.

The images of the mouse eyes were processed using software (Get-eye, Matsuura-Denko-sha, Kanazawa, Japan) on a PC (Epson; Figure 1B). This eye movement recording system was the same as that used in the study from Yoshida, Funabiki, and Hirano (2007). Briefly, the image sampling rate was 200 Hz, the black image, i.e., the pupil, was detected, and the center of mass of the image was calculated. Thus, the system measured the position of the pupil every 5 ms. 

Using the software, the eye position was recorded as the center pixel of the black image. To associate the pixel of the image and the angle of the eye, we conducted calibration. Since mice do not have foveae, we could not execute the calibration by using the visually guided saccade like in monkeys and in humans. We instead used a metal model of the mouse eyeball. The diameter of the model was 3.4 mm (Remtulla & Hallett, 1985). We recorded the eye position by rotating the eye model (2° for each rotation), and computed how the change in the pixel was related to the eye angle. 

Visual stimulus presentation system

In many previous studies, a rotating drum—a cylindrical screen with vertical black and white stripes—has been used to record the OKR. In the present study, we developed a visual stimulus presentation system using computer displays, which allowed us to change the properties of the visual stimulus systematically. To cover all of the angles in the visual scene, we used three displays (19-inch LCD; refresh rate, 75 Hz; Mitsubishi, Tokyo, Japan) and distributed them on the left, right, and in front of the mouse ( Figure 1B). The horizontal and vertical angles of the visual stimuli were 270° and 67.5°, respectively. Note that, in all experiments, the mice looked at the visual stimuli with their both eyes, because the hot mirror transmits the visible light. 

The presentation of the visual stimuli was controlled with Psych-toolbox (MATLAB; Mathworks), and a local area network (UDP protocol) was used to communicate with the eye movement recording system (Lab View, National Instruments). 

Visual stimulus

The visual stimulus was a drifting vertical sinusoidal grating with a sinusoidally modulated luminance. We adjusted the spatial frequency (SF) of the grating on the computer displays to mimic a rotating drum ( Figure 1B). Therefore, the SF was constant for the animal, regardless of the horizontal location of the monitor. In other words, in order for the visual angle of each stripe seen from the mouse to be the same, the widths of the stripes on the display were adjusted; narrower in the center and wider toward the edges (see Figure 1B). 

We manipulated the parameters of the visual stimulus based on the purpose of the experiment. In the experiment in which we manipulated the contrast, we calculated the Michelson contrast, which is defined as [(Lmax − Lmin) / (Lmax + Lmin)] × 100%, where L is the luminance. 

We used contrasts of 1, 2, 4, 8, 16, 32, 64, and 96%. We set the SF, temporal frequency (TF), and the mean luminance to 0.125 cycle/deg, 1.5 Hz, and 100 cd/m ², respectively. To compare the contrast dependency with another mean luminance, we performed the same experiment while setting the mean luminance to 25 cd/m ². We defined 100 cd/m ² and 25 cd/m ² as the high and low mean luminances, respectively. In these experiments, we also observed the eye movements elicited by a visual stimulus with a luminance modulated based on a square-wave function instead of a sinusoidal function. In all of the experiments described below, we used a sinusoidal grating pattern with a contrast and mean luminance of 64% and 100 cd/m ², respectively. 

To study the dependence of the eye movements on the spatiotemporal frequency of the visual stimuli, we systematically changed the SF of the sine wave and the speed of the visual stimulus in each trial. The SFs of the visual stimulus were 0.0313, 0.0625, 0.125, 0.25, and 0.5 cycle/deg. The motion of the visual stimulus was defined by the TF, which was selected from 0.1875, 0.375, 0.75, 1.5, 3, 6, 12, or 24 Hz. The speed of the visual stimulus was calculated using TF/SF; e.g., the speed of a visual stimulus with SF and TF of 0.125 cycle/deg and 1.5 Hz, respectively, is 12 deg/s. 

Trial scheme

Diagrams of each trial are depicted in Figure 1C. First, a stationary visual pattern was presented on the monitors. We defined the period between the appearance of the visual pattern and the onset of motion as the motion onset delay (MOD). In most cases, we set the MOD to 333 ms, although we varied the MOD in one experiment. After the MOD, the visual pattern started to move either counterclockwise (from temporal to nasal positions for the right eye) or clockwise (from nasal to temporal positions for the right eye). The speed of the visual stimulus was constant during each trial. A gray background, whose luminance was consistent with the mean luminance of the stimulus, was presented on the monitors for 2 s after the end of the stimulus presentation. Then, the next trial started automatically. 

Data analysis

We processed the eye position data using MATLAB. Before we calculated the eye velocity, we checked the initial eye position and rejected outliers. We detected trials in which the initial eye position was 10° away from the center position. The eye velocity data were obtained using backward difference, and smoothed with a digital Butterworth filter (3 poles; 3 dB at 15 Hz and 20 dB at 30 Hz). To focus on the smooth eye movements, we also detected any saccade-like eye movements (acceleration > 200 deg/s ²) and rejected trials that contained them. All eye movement data were aligned based on the onset of the visual stimulus motion. 

Latency

The sampling rate of the eye movement recording system used in the present study was 200 Hz. Therefore, the reliability of the system for determining the latency of the eye movements is inferior to that of a search-coil system with a high sampling rate (500 Hz). van Alphen, Stahl, and De Zeeuw (2001) used a search-coil method to determine that the latency of the OKR in mice is 70 ms. We calculated latency of the OKR in our stimulus condition to compare with those measured in the previous study. 

We computed the latency using receiver operating characteristic (ROC) analysis. This type of analysis made it possible to determine when the eye velocities elicited by counterclockwise visual motion and those elicited by clockwise visual motion differed. We constructed the ROC curve for each time point. The “Hit rate” and the “False Alarm rate” corresponded to the responses to counterclockwise visual motion and those to clockwise visual motion, respectively. We computed these rates using step increases of 0.1 deg/s, and then calculated the areas under the ROC curves. Thus, the areas under the ROC curves were calculated at each time point and plotted as a function of time. We fit the data with the Weibull distribution function: y = p − ( p − q) e ^{(− t/ r) s}. The parameters p, q, r, and s were computed based on a least-squares criterion. We defined the time at which the Weibull distribution curve crossed 0.75 as the latency of the eye movements. 

The dependency of velocity tuning on spatial frequency

We quantified the dependence of velocity tuning on spatial frequency by using a variant of a two-dimensional Gaussian on logarithmic axes:  

$\begin{array}{c}R(sf,tf)=A*\exp \left(\frac{-\left({\log }_2\right(sf)-{\log }_2(s{f}_0){)}^2}{{\sigma }_{sf}^2}\right)\\ \\ *\exp \left(\frac{-\left({\log }_2\right(tf)-{\log }_2(t{f}_p(sf)){)}^2}{{\sigma }_{tf}^2}\right),\end{array}$

(1)

where  

$t{f}_p(sf)=2^{(Q+1)*\left({\log }_2\right(sf)-{\log }_2(s{f}_0\left)\right)+{\log }_2\left(t{f}_0\right)}.$

(2)

 

The sf and tf indicate spatial and temporal frequencies, respectively. The sf ₀ and tf ₀ show the center of the Gaussian. The σ _sf and σ _tf show the standard deviation of each axes. The Q value affects the slope of the Gaussian and shows the dependence of velocity tuning on spatial frequency. We used the Q value as the velocity tuning index (see also Priebe, Cassanello, & Lisberger, 2003). 

We applied this method to analyze the spatiotemporal frequency map of the eye movement shown in Figure 6. The response distribution was fitted by these equations, and the Q value was used to evaluate whether the response showed the velocity tuning on the spatial frequency. When Q is 0, the preferred temporal frequency is proportional to spatial frequency, and thus preferred speed is constant across spatial frequency (“velocity tuning”). In this case, the response distribution on the spatiotemporal frequency map would show the velocity tuning, i.e., the slope of Gaussian would tilt parallel to the line connecting data point of the same velocity (=TF/SF). When Q is −1, the preferred temporal frequency is constant independent of spatial frequency, and thus the preferred velocity changes with spatial frequency. 

Results

Visually driven smooth eye movements in mice

We calculated the eye velocity based on the measured eye position. Sample velocity profiles of the eye movements when the visual stimulus moved counterclockwise (temporal to nasal motion for the right eye) at 12 deg/s are shown in Figure 2A (the SF and TF of the visual stimuli were 0.125 cycle/deg and 1.5 Hz, respectively). Smooth eye movements were elicited soon after the visual stimulus started to move, and the eye velocities were maintained for at least 400 ms after the onset of visual motion. In this paper, we use open-loop and closed-loop responses to represent the mean eye velocities at 100-ms intervals starting 50 ms and 300 ms after the onset of motion, respectively. 

Figure 2

 

Sample response profiles obtained from mouse A. (A) The eye movements when the visual stimuli (SF, 0.125 cycle/deg; TF, 0.75 Hz) moved in a counterclockwise direction. The dotted lines show the standard deviation of the individual eye velocity (44 trials). The solid line shows the mean of the individual eye velocity. (B) The latency of the eye movement was calculated from the eye movements elicited by counterclockwise and clockwise visual motions and is shown using arrows. Upward and downward deflections represent eye velocity toward counterclockwise (CCW) and clockwise (CW).

To calculate the latency, we used the eye velocities elicited by counterclockwise and clockwise motions ( Figure 2B, the black and the gray lines denote responses to counterclockwise and clockwise motions, respectively). The latency for these sample eye movements calculated from ROC analysis was 90 ms (arrow). The mean latency for all of the animals was 101.4 ± 10.0 ms (mean ± SD; n = 11; distributed from 85 ms to 115 ms). Thus, we found that sudden movement of a visual stimulus causes smooth eye movements with short latencies in mice. 

Dependence of the eye movements on contrast

We measured smooth eye movements elicited by visual stimuli with various contrasts (sinusoidal grating; mean luminance, 100 cd/m ²). In the experiment, the visual stimulus with SF = 0.125 cycle/deg and TF = 1.5 Hz was used. The eye velocity profiles of mouse A are shown in Figure 3A. Larger ocular responses were elicited by visual stimuli with higher contrast values. We calculated the mean eye velocity for two temporal periods (open-loop phase: 50–150 ms; closed-loop phase: 300–400 ms) and plotted the results as a function of the contrast in Figure 4A (blue lines). In both phases, the eye velocity gradually increased as the contrast increased. 

Figure 3

 

(A) Eye movement velocity profiles induced by visual stimuli with various contrast levels in mouse A. The SF and TF of the stimulus were 0.125 cycle/deg and 1.5 Hz, respectively (stimulus velocity = 12 deg/s). The different colors represent the different contrast levels of the visual stimuli. (B) Eye movement velocity profiles for different types of visual stimuli (contrast, 64%) in mouse A. High lum and low lum shows the mean luminance of the visual stimuli: 100 and 25 cd/m ², respectively. Sin and sq denotes the sinusoidal and square-wave visual patterns, respectively. Upward deflections represent eye velocity toward counterclockwise (CCW).

Figure 4

 

(A) Dependence of the initial OKR on the contrast of the stimulus in mouse A. The left and right columns show the results from the open-loop and closed-loop phases, respectively. High lum and low lum denotes the mean luminance of the visual stimuli: 100 and 25 cd/m ², respectively. Sin and sq signifies the sinusoidal and square-wave visual patterns, respectively. Upward and downward deflections represent eye velocity toward counterclockwise (CCW) and clockwise (CW). (B) Contrast dependence fit using the Naka–Rushton equation for mouse A. Black circles and gray squares denote the data obtained when the luminance was high and low, respectively. The solid lines show the best-fit curve. Upward and downward deflections represent eye velocity toward counterclockwise (CCW) and clockwise (CW). The parameters for the best-fit Naka–Rushton equation ( c ₅₀ and n) are also shown.

To examine the effect of the mean luminance, we measured the eye movements elicited by a dim visual stimulus (sinusoidal grating; mean luminance, 25 cd/m ²). The eye movement profiles of mouse A are shown in Figure 3B. For stimuli with a low luminance, the eye velocity in the sustained phase was smaller (orange dashed line), even though the eye velocity initially increased as observed with the high luminance. The quantitative results are shown in Figure 4. In most cases, the eye movements elicited by the low-luminance stimulus (green lines) were smaller than those obtained with the high-luminance stimulus (blue lines). Of the results in three mice (mice A, D, E), the differences were statistically significant in 8 of 48 open-loop phase cases and in 34 of 48 closed-loop phase cases ( P < 0.05; two-tailed t-test). This indicates that a higher luminance is required for larger eye movements, but the luminance effect is relatively small during the open-loop phase. 

To compare the ocular response elicited by the sinusoidal grating with those elicited by a visual stimulus consisting of black and white stripes, we utilized the visual stimulus whose luminance modulated in a square-wave manner (square-wave grating; mean luminance, 100 cd/m ²), instead of the sinusoidal grating. The eye movement elicited by the square-wave pattern is shown in Figure 3B (black solid line). The data show an eye velocity profile that is similar to the eye movement elicited by the sinusoidal visual stimulus (orange solid line, Figure 3B). Quantitative measurements demonstrated that the differences between the eye velocities obtained with the sinusoidal and square-wave patterns were statistically significant in 4 of 48 open-loop phase cases and in 10 of 48 closed-loop phase cases ( P < 0.05; two-tailed t-test). This indicates that, in most cases, the eye movements elicited by the square-wave visual pattern were not significantly different from those obtained with the sinusoidal visual pattern. 

To quantitatively evaluate the dependence on contrast and compare it with the primate OFR, we fit the results with the Naka–Rushton equation (Naka & Rushton, 1966): 

$y=R_{\max }\frac{c^n}{c^n+c_{50}^n},$

(3)

where y is the eye velocity, R_max is the maximum response, c is the contrast, c₅₀ is the semisaturation contrast, which elicits a response at half the maximum value, and n is the exponent that sets the steepness of the curves. The results from mouse A are shown in Figure 4B. The black circles and gray squares show the data for the high-luminance and low-luminance sinusoidal visual stimuli, respectively. The values of c₅₀ and n are also shown in Figure 4B. For the open-loop phase responses to the high-luminance stimuli (Figure 4B, left panel), the values of c₅₀ were 17.1 and 21.8 for the eye movements elicited by counterclockwise and clockwise visual motions, respectively. In other words, the open-loop response of the OKR was relatively weak if the contrast was less than 17.1% and 21.8%, respectively. Similar results were obtained when the luminance was low (c₅₀ = 25.4 and 19.0 for the eye movements elicited by counterclockwise and clockwise visual motions, respectively). The values were much higher than those associated with the OFR in humans and monkeys (see the Comparison with the OFR section). This tendency was similar for the closed-loop response (Figure 4B, right panel). In two cases, the c₅₀ value was more than 100, indicating that the dependence on the contrast did not become saturated. The results for all of the mice are shown in Table 1 (high and low luminances). The results in mice B and C were consistent with mouse A. 

Table 1

 

Summary of parameters from the Naka–Rushton equation. We calculated mean values of c₅₀ for the values less than 100. CCW and CW correspond to the eye movements elicited by counterclockwise and clockwise visual motions.

Direction of visual stimulus		High luminance				Low luminance
		Open-loop phase		Closed-loop phase		Open-loop phase		Closed-loop phase
		CCW	CW	CCW	CW	CCW	CW	CCW	CW
Mouse A	c 50	17.1	21.8	9.9	>100	25.4	19.0	25.3	>100
	n	2.33	1.27	1.57	0.59	1.91	1.64	1.81	0.72
Mouse B	c 50	11.5	>100	4.6	18.9	15.6	96.6	19.4	65.8
	n	6.75	0.43	1.73	1.02	1.54	1.12	1.65	1.07
Mouse C	c 50	6.7	>100	5.9	17.2	17.1	8.45	22.0	>100
	n	6.17	0.68	1.38	0.85	3.00	3.66	1.54	0.63
Mean	c 50	14.28		11.30		30.36		33.15
	n	2.94		1.19		2.15		1.24

Sensitivity of the visually driven smooth eye movements to the stimulus spatiotemporal frequency

We measured the smooth eye movements elicited by visual stimuli with various spatiotemporal frequencies (contrast, 64%; mean luminance, 100 cd/m ²). The eye velocity profiles elicited by the counterclockwise motion of the visual stimulus presented to mouse A are shown in Figures 5A– 5E. The SFs of the visual stimuli were 0.0313, 0.0625, 0.125, 0.25, and 0.5 cycle/deg for the results shown in Figures 5A– 5E, respectively. The different colors in each figure show eye velocity profiles elicited by the different TFs of the visual motion. We found that, for all responses except for SF = 0.5, the peak eye velocities were significantly larger than zero (one-tailed t-test, p < 0.05). However, the results also demonstrate that the spatiotemporal frequency range of the visual stimuli that elicit a large OKR is limited. For example, the visual stimuli with the lowest and highest SFs each caused weak and almost no responses, respectively ( Figures 5A and 5E). In addition, even for the middle SF ( Figure 5C), the visual stimuli with lower and higher TFs elicited weaker responses. We found that a visual stimulus with SF and TF values of 0.125 cycle/deg and 1.5 Hz, respectively, elicited the largest sustained eye movement ( Figure 5C, blue line). 

Figure 5

 

Eye velocity profiles from mouse A for visual stimuli with various spatiotemporal frequencies. (A–E) The SFs were 0.0313, 0.0625, 0.125, 0.25, and 0.5 cycle/deg. The different colors in each panel denote the different TFs. Upward deflections represent eye velocity toward counterclockwise (CCW).

We also observed a transient response elicited by stimuli with higher TF values. For example, when the SF was set at 0.125 cycle/deg, stimuli with TF values of 6 or 12 Hz elicited transient increases in the eye velocity, whereas the sustained eye velocity was close to zero. Interestingly, the build-up of the eye velocity occurred more rapidly than that of the eye movement showing the highest sustained eye velocity (TF, 1.5 Hz). The transient responses for higher TFs (TF = 6 or 12 Hz) also were observed with other SFs (SF = 0.0313, 0.0625, and 0.25 cycle/deg). The velocity profiles except for SF = 0.5 showed that the time to peak was earlier for high TF stimulus ( p < 0.01, Spearman's rank correlation). The results suggest that, first, the range of spatiotemporal frequencies of the visual stimulus that can elicit large sustained smooth eye movements is limited. Second, visual stimuli with higher TFs induce transient smooth eye movements, but the eye velocity during the sustained phase is near-zero. Third, the eye velocity reaches to its peak more rapidly for higher TF stimulus. 

To quantitatively evaluate the dependence of the eye movement on the spatiotemporal frequency of the visual stimulus, we calculated the mean eye velocity for each condition, which is shown as response fields ( Figure 6A). In these graphs, the amplitude of the ocular response is indicated by the diameter of the symbol. The left and right panels correspond to the open-loop (100-ms interval starting at 50 ms) and closed-loop (100-ms interval starting at 300 ms) phases of the eye movements, respectively. In the closed-loop phase ( Figure 6A, right panel), we observed a sharply tuned distribution of the responses. The maximum response was located at an SF value of 0.125 cycle/deg and a TF value of 1.5 Hz (indicated by the arrow). In contrast, in the open-loop phase ( Figure 6A, left panel), the responses were distributed at relatively higher TF region, and the maximum response was located at a higher TF (SF = 0.125 cycle/deg, TF = 3 Hz, indicated by the arrow). Concordant results were obtained from other animals for the eye movement elicited by counterclockwise visual motion ( Figures 6B and 6C). In both of mice D and E, the maximum response field in the closed-loop phase was located at an SF value of 0.125 cycle/deg and a TF value of 1.5 Hz (arrow). The results were consistent with mouse A. In the open-loop phase, SF and TF of the peak were 0.25 cycle/deg and 3 Hz in mouse D (arrow), respectively. In mouse E, the maximum response was located at 0.125 Hz and 1.5 Hz (arrow), however, the response amplitude at an SF value of 0.125 cycle/deg and a TF value of 3 Hz was almost identical to that at the peak (the difference was less than 2%). These results suggest that the spatiotemporal frequency tuning in the open-loop phase is a bit varied among mice, however, the preferred temporal frequency was relatively higher than that of the closed-loop phase. Similar results were obtained for the eye movement elicited by the clockwise motion, although the amplitude was slightly smaller than the opposite direction in some mice. These results indicate that the preferred stimulus changes with time; the visual stimulus with SF and TF values of around 0.125 cycle/deg and 1.5 Hz, respectively, is optimal for the sustained response, however, relatively higher TF is optimal for the open-loop response. 

Figure 6

 

Performance of the initial OKR elicited by counterclockwise visual motion on spatiotemporal property of the visual stimulus is plotted in the coordinate system of spatial and temporal frequencies. The horizontal and vertical axes show SFs and TFs of the visual stimuli, respectively. Left and right columns correspond to eye velocities in open-loop and closed-loop phases, respectively. (A–C) Response field of mice A, D, and E. The amplitude of the response is indicated by the diameter of the symbol. The black arrow shows the location of the peak. (D) Contour plot of animal's mean (mice A, D, and E) fitted by a two-dimensional Gaussian on logarithmic axes shown in Equations 1 and 2. The responses were normalized by adjusting the maximum response as 1.

To quantitatively clarify the tuning property, we fitted the averaged response over three mice with a two-dimensional Gaussian on logarithmic axes shown in Equations 1 and 2. The responses were normalized by adjusting the maximum response as 1. The best-fitted functions are exhibited in Figure 6D. The estimated peak for the open-loop response was located at an SF value of 0.117 cycle/deg and a TF value of 2.99 Hz ( R ² = 0.865), and it moved to SF = 0.152 cycle/deg and TF = 1.43 Hz ( R ² = 0.876) for the closed-loop response. We also showed the movie of the temporal development of the tuning function, starting from the open-loop phase to the closed-loop phase ( Movie 1). The movie clearly shows that the best-fit Gaussian gradually moves and reaches to the position of the closed-loop phase. These results support the idea that the optimal spatiotemporal frequency changes through the development of the eye movement. 

 

Movie 1

 

This animation shows dynamic change of best-fit two-dimensional Gaussian starting from the open-loop ( Figure 6D, left) to the closed-loop ( Figure 6D, right) phases.

Then, we examined whether the responses showed the velocity tuning. We fitted the response of individual mouse by the two-dimensional Gaussian and compared the velocity tuning index Q of Equation 2. When Q is −1, the preferred stimulus velocity (=TF/SF) changes among SFs. When Q is 0, the preferred velocity does not change among SFs, i.e., velocity tuning (see Methods section for details). For the eye movements elicited by counterclockwise visual motion, the Q values of the open loop and the closed loop were −0.414 ± 0.07 and −0.012 ± 0.127, respectively (with standard deviation). The difference was statistically significant (two-tailed t-test, p < 0.01). This indicates that the ocular responses in the closed-loop phase are tuned for the velocity (tuned for about 12 deg/s of stimulus velocity), and the responses in the open-loop phase are weakly tuned for the velocity in comparison with that in the closed loop. For the eye movements elicited by clockwise visual motion, we could not obtain the significant results due to small open-loop response in some mice. 

For closed-loop responses, we also calculated the gain (eye velocity/target velocity) of the eye movement. The mean values are shown in Figure 7. For both counterclockwise ( Figure 7A) and clockwise ( Figure 7B) motions, the gain was larger when the stimulus velocity was smaller. This result is consistent with those from a classic OKR study (Collewijn, 1991). 

Figure 7

 

Gain of the eye movement calculated from the closed-loop eye velocity (eye velocity/target velocity) in three mice, for which the dependence on the spatiotemporal frequency is shown in Figure 6. (A) Eye movement elicited by counterclockwise visual motion. (B) Eye movement elicited by clockwise visual motion. Different colors denote different target velocities.

Effect of the duration of the visual stimulus presentation

To examine the dynamic properties of the development of the eye velocity, we observed the velocity profiles of the smooth eye movements while systematically changing the duration of the visual stimulus presentation. The eye movements induced by a counterclockwise visual stimulus with an SF of 0.125 cycle/deg and a TF of 1.5 Hz (contrast, 64%; mean luminance, 100 cd/m ²) are shown on the left in Figures 8A (mouse A) and 8B (mouse E). For this visual stimulus (SF, 0.125 cycle/deg; TF, 1.5 Hz), elongation of the stimulus presentation period caused a gradual increase in the eye velocity, with a 160-ms presentation resulting in the sustained level. In contrast, for a visual stimulus characterized by SF and TF values of 0.0625 cycle/deg and 6 Hz, respectively (contrast, 64%; mean luminance, 100 cd/m ²), elongation of the stimulus presentation period did not cause an increase in the eye velocity. Similar results were obtained for the clockwise visual motion. The results clearly showed that visual stimuli with higher TFs did not elicit a sustained OKR, instead inducing only a transient response. In addition, a 40-ms stimulus presentation is sufficient to cause a transient response. 

Figure 8

 

Eye velocity profiles during the initial OKR for various stimulus presentation durations. The results for two different visual stimuli are shown in the left (SF, 0.125 cycle/deg; TF, 1.5 Hz) and right (SF, 0.0625 cycle/deg; TF, 6 Hz) columns. The eye movements of mouse A and mouse G are shown. Upward deflections represent eye velocity toward counterclockwise (CCW).

The effect of MOD

We then investigated the effect of the motion onset delay (MOD) on the OKR. We changed the length of the period between the presentation of the visual stimulus and the onset of visual motion. We calculated the amplitude of the eye velocity during the open-loop phase for each mouse. The quantitative data was obtained by averaging the amplitude of the counterclockwise and the clockwise eye velocities. The data for the visual stimuli with an SF of 0.125 cycle/deg and a TF of 1.5 Hz and an SF of 0.0625 cycle/deg and a TF of 6 Hz (contrast, 64%; mean luminance, 100 cd/m ²) are shown in Figures 9A and 9B, respectively. For both visual stimuli, the open-loop responses in the conditions of MOD ≥ 67 ms was larger than the response when MOD was 0 ms. The statistical analysis showed that the difference was significantly different, but the responses among the conditions of MOD ≥ 67 ms was not significantly different ( P < 0.01, two-way ANOVA with a post-hoc test; factor 1, animal; factor 2, MOD; test effect of the MOD). In other words, the initial eye movements were delayed if the MOD was 0 ms, whereas no significant differences were observed for MOD values ≥67 ms. 

Figure 9

 

(A) The mean eye velocity during the open-loop phase as a function of the MOD for the visual stimulus with an SF of 0.125 cycle/deg and a TF of 1.5 Hz is shown ( n = 7; mice A, B, C, I, J, K, N). (B) The mean eye velocity during the open-loop phase as a function of the MOD for the visual stimuli with an SF of 0.0625 cycle/deg and a TF of 6 Hz is shown ( n = 4; mice I, J, K, N). Asterisks denote statistical significance ( P < 0.01).

Discussion

We found that smooth eye movements were elicited in mice by the motion of visual stimuli with a range of spatiotemporal frequencies. A visual stimulus with an SF of 0.125 cycle/deg and a TF of 1.5 Hz elicited the strongest sustained eye movements. We also found that stimuli with higher TF values elicited transient increases in the eye velocity. The eye velocity soon decayed toward zero, even though the visual motion continued. These results suggest that the initial stage of the OKR in mice has at least two components: a transient component that increases the eye velocity before decaying toward zero after approximately 200 ms, as well as a sustained component that shows a sustained velocity for more than 400 ms. In other words, the mouse OKR system cannot be modeled using a simple spatiotemporal filtering of the velocity of the visual motion on the retina. The mouse OKR system uses at least two channels to process visual motion; one processes the velocity of the visual motion on the retina and prefers visual stimuli characterized by an SF of 0.125 cycle/deg and a TF of 1.5 Hz, whereas the other processes acceleration and/or the onset of the visual motion and prefers higher TF values. 

The OKR as a behavioral measure to study mouse visual system

The present study shows that visual stimuli with an SF near 0.125 cycle/deg caused the largest smooth eye movements. The result is consistent with the recent report showing that a stimulus with SF = 0.17 cycle/deg evoked optimal sustained OKR (van Alphen, Winkelman, & Frens, 2009). If the dependence on the SF is a property of the mouse visual system, eye movements can be used as an excellent measure to investigate the mouse visual system. Prusky and Douglas (2004) studied the contrast sensitivity of mice to visual stimuli with five SFs (0.059, 0.119, 0.209, 0.297, and 0.445 cycle/deg) and reported that 0.209 cycle/deg and 0.119 cycle/deg were the most and second most sensitive SFs, respectively. This is consistent with the SF-dependent nature of the OKR reported in the present study. 

In addition, Sinclair, Jacobs, and Nirenberg (2004) studied the electrophysiologic activity of retinal ganglion cells in mice and reported that, in control mice, the neural activity of this population peaked with a stimulus characterized by an SF of 0.115 cycle/deg (log cycle/deg = −2.16), which is consistent with our results. 

These data suggest that the initial part of the OKR in mice encodes the preferred SF of the visual system. In other words, observation of the OKR for 500 ms is sufficient to test the properties of the visual system. The methodological advantages of observing the mouse OKR have been demonstrated elsewhere; e.g., the OKR requires no training, can be elicited repeatedly with minimal fatigue or adaptation, and its neural substrates are thought to be simple (Cahill & Nathans, 2008; Stahl, 2004, 2008). Furthermore, focusing on the initial component of the OKR has a number of benefits. First, this approach allows exclusion of nystagmus. Second, because the eye movement is only recorded for a short period of time in each trial, more data can be collected in a smaller time interval. Third, the pure neural and behavioral response to the visual input can be observed without a visual feedback signal by analyzing the open-loop phase (∼150 ms). Finally, the OFR in primates has been extensively studied. Therefore, comparisons can be made between the properties of the mouse and primate visual systems. 

The mouse OKR has been mainly used to evaluate the effects of adaptation or learning on motor performance (see review by Stahl, 2008). We suggest that the initiation of the OKR can be used to investigate the mouse visual system. 

Optimal stimuli for visually driven smooth eye movements in mice

We determined which type of visual stimuli would cause vigorous transient and sustained eye movements. First, we found that higher contrast values cause larger eye movements in mice. The dependence on contrast was observed during the open-loop and closed-loop phases. Therefore, to elicit vigorous transient and sustained eye movements, a high-contrast visual stimulus should be used. In addition, eye movements elicited by the motion of a square-wave pattern were not significantly different from those elicited by the motion of a sinusoidal pattern. Thus, the effect of the edge of the visual pattern (a component of the higher TF) on eye movement is small, and smooth eye movements are generated primarily by the fundamental Fourier component of the visual stimulus. 

Second, we found that the best spatiotemporal frequency of the visual stimulus included an SF of 0.125 cycle/deg and a TF of 1.5 Hz during the closed-loop phase of the OKR. The range of the spatiotemporal frequencies that caused a vigorous sustained OKR was limited around this peak. The open-loop phase of the OKR showed a preference for higher TF values, however, and the distribution of the color contour plot was broader. The fitting analysis estimated that the best SF and TF were 0.117 cycle/deg and 2.99 Hz for the open-loop phase, and 0.152 cycle/deg and 1.43 Hz for the closed-loop phase. These results indicate that the optimal spatiotemporal frequency of the visual stimulus was different between during the open-loop phase and during the closed-loop phase. 

Third, we investigated the effect of the duration of the visual stimulus presentation. For the preferred visual stimulus (SF, 0.125 cycle/deg; TF, 1.5 Hz), elongation of the stimulus presentation period increased the eye velocity, which reached a sustained value after a 160-ms presentation. In other words, to observe the closed-loop phase of the OKR, the visual stimulus should be presented for at least 160 ms. In contrast, for the visual stimulus with an SF of 0.0625 cycle/deg and a TF of 6 Hz, presentation of the stimulus for 40 ms elicited a transient increase in the eye velocity, whereas the velocity profile was not changed if the visual stimulus was presented for a longer period of time. Accordingly, a short presentation is sufficient to observe a transient eye response to visual stimuli with higher TF values. 

Fourth, we investigated the effect of the MOD and found that the initial eye velocity was weak or delayed when the MOD was 0 ms. Therefore, the MOD is important when attempting to induce a vigorous initial OKR response. It has been reported that the MOD gradually affects the initiation of smooth pursuit eye movements in primates (Krauzlis & Lisberger, 1994; Tabata, Miura, Taki, Matsuura, & Kawano, 2006). The amplitude of the visually driven smooth eye movement gradually increased as the MOD became longer, before it plateaued at MODs between approximately 200 and 400 ms. In the present study, however, there was no significant differences among the results obtained with MODs greater than 0 ms. The difference may be attributed to the nature of the movement, voluntary or involuntary. 

In some mice, the amplitude of the initial OKR elicited by the clockwise visual motion was relatively smaller than that by the counterclockwise visual motion. However, both eye movements showed similar qualitative properties. Therefore, we conclude that the optimal visual stimulus is the same regardless of the eye movement direction. 

Taken together, our results show that visually driven smooth eye movements can be used as a powerful tool to investigate brain function in mice. An appropriate visual stimulus should be chosen to ensure vigorous eye movements, however. 

Two components of OKR initiation

We observed both transient and sustained components of OKR initiation. As suggested by a number of previous OKR studies, sustained eye velocity may play an essential role in compensating for the visual motion on the retina and is thought to be generated through the AOS and NOT. The functional role and neural substrates of transient eye movement, however, are not clear. We found that presentation of a visual stimulus for only 40 ms resulted in transient eye movements, and that these movements were more pronounced with stimuli characterized by higher TFs than those used to induce the largest sustained eye movements. Accordingly, the mouse OKR system detects the onset or acceleration of visual motion with a relatively high TF. This might be related to the visual system functions as an event detector. 

What is the neural substrate of these components? It is reported in wallaby that the NOT includes two types of neurons (Ibbotson & Mark, 1996). The first type of neurons is “slow cells”, which are maximally sensitive to visual stimuli moving at low velocities. The neurons exhibit the rapid increase of the activity and slow decay. The second type of neurons is “fast cells”, which are maximally sensitive to the stimuli moving at high velocities. The neurons exhibit the rapid increase of the activity. After the peak, the activity drops sharply and then declined to a level below the spontaneous level. If the mouse NOT has these types of neurons, the “fast cells” might be contributed to elicit a transient eye movement for the visual stimulus with higher TFs. 

Is the visual cortex relevant to the transient response? In rats, lesion of the visual cortex had no significant effect on the OKR (Harvey, De'Speriati, & Strata, 1997). Therefore, the visual cortex is not included in the generation of the sustained OKR. Whether lesions in the visual cortex affect the initial part of the OKR, however, is still unclear. If such lesions only change the transient component, the visual cortex would be a promising candidate for the neural substrate. 

We also found that the Q values in open-loop and closed-loop phases were −0.414 and −0.012, indicating that the ocular responses to the visual stimulus does not initially show a clear velocity tuning but lately show a velocity tuning. This suggests difference of the tuning property between the transient and the sustained eye movements. However, to clarify how the mouse brain computes the velocity of the visual stimulus, further studies including neurophysiology are necessary. 

Comparison with the OFR

The OKR is a slow compensative smooth eye movement that is thought to have direct (early) and indirect (late) components (Cohen, Matsuo, & Raphan, 1977). The direct component is caused directly by the onset of the visual stimulus and manifests as a rapid increase in eye velocity. The direct component acts synergistically with the translational vestibule-ocular reflex (TVOR) and is thought to correspond to the OFR in primates (Büttner & Kremmyda, 2007; Miles, 1998). In contrast, the indirect component, which exhibits a more gradual increase in eye velocity during continuous optokinetic stimulation, is functionally linked to the rotational vestibule-ocular reflex (RVOR). Although it is unclear whether the OFR that performs as the backup system of the TVOR is developed in lateral-eye animals such as mice, here we compare the initial OKR in mice and the OFR in primates. 

The latencies of the OFR in monkeys and humans are approximately 50 ms and 70 ms, respectively. We found that the visual stimulus (SF, 0.125 cycle/deg; TF, 1.5 Hz) caused smooth eye movements in mice with a latency of 101 ms. The minimum latency was 85 ms. van Alphen et al. (2001) measured a 70-ms latency for the OKR in mice using a search-coil method, which is slightly shorter than the latency measured in the present study. We do not believe that our results contradict those from the previous study, however, because the latency strongly depends on the technique and criteria used to record and calculate the eye movements. Because the mouse brain is much smaller than the primate brain, the physical distance that the neural activity must be transmitted is much shorter in mice than in primates. Nonetheless, even the shortest latency (70 ms reported by van Alphen et al., 2001) is longer than the latency reported for the monkey OFR. 

We then compared the contrast-dependent characteristics of the mouse initial OKR and the primate OFR. Because the OFR is an open-loop response, we examined the contrast dependence during the open-loop phase. For high-luminance stimuli (mean luminance, 100 cd/m ²), c ₅₀ was more than 100 in two of six cases, indicating that the contrast dependence did not saturate. In other cases, the mean c ₅₀ value from the Naka–Rushton equation was 14.28 ( Table 1). For low-luminance stimuli (mean luminance, 25 cd/m ²), the c ₅₀ value was 30.36 ( Table 1). On the other hand, in monkeys and humans, experiments using a CRT monitor (mean luminance, 37.8 cd/m ²) showed c ₅₀ values of 3.24 and 3.9, respectively (Miura et al., 2006; Sheliga, Chen, FitzGibbon, & Miles, 2005). Thus, the value of c₅₀ in mice is much higher than in monkeys and humans. In other words, the Naka–Rushton curve for mice moves toward higher contrast values compared with the results obtained in monkeys and humans. These results suggest that contrast resolution in the mouse OKR system is lower, and that higher contrast is required for a marked initial OKR. The mean n values in mice were 2.94 and 2.15 for high- and low-luminance stimuli, respectively. Although these values are slightly larger than those for humans (n = 2.10) and monkeys (n = 2.09), the difference is not statistically significant (P > 0.25; one-tailed t-test). The Naka–Rushton equation provides a good fit for the contrast-dependence curve for neurons in the lateral geniculate nucleus, V1, and middle temporal areas in monkeys (Albrecht, Geisler, Frazor, & Crane, 2002; Albrecht & Hamilton, 1982; Heuer & Britten, 2002; Sclar, Maunsell, & Lennie, 1990). Therefore, the difference between the mouse initial OKR and the primate OFR may reflect differences in the respective visual systems. 

Finally, we examine the dependency of the response on the spatiotemporal frequency. Gomi, Abekawa, and Nishida (2006) created a color contour plot representing the spatiotemporal frequency relationship of the monkey OFR based on experiments conducted by Miles et al. (1986). The results showed that low SF values (∼0.05 cycle/deg) and high TF values (>10 Hz) are required to elicit the largest eye movements. Our results suggest that the shape and peak location of the color contour plot differ from those observed in data from the monkey. 

Thus, the mouse OKR and the primate OFR are associated with behavioral differences. It has been reported that the MST area in monkeys plays an essential role in generating the OFR (Kawano et al., 1994; Takemura et al., 2007). In contrast, the mouse visual cortex is less developed, and it has been thought that the AOS and the NOT is the main pathway in the mouse OKR system. The differences between the mouse OKR and the primate OFR might result from differences in the neural pathways that generate the movements. 

Acknowledgments

We thank Dr. M. Funabiki for his many useful comments about the initial setup of the equipment, and Ms. Y. Sugita for her helpful support during the experiments. This work was supported by JSPS KAKENHI (16GS0312 and 21240037). 

Commercial relationships: none. 

Corresponding author: Kenji Kawano, MD, PhD. 

Email: [e-mail]. 

Address: Konoe-cho, Yoshida, Sakyo-ku, Kyoto-shi, Kyoto, Japan.