INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The electroretinogram (ERG) recorded at the corneal surface provides a useful tool with which to measure outer retinal activity (Fishman et al. 2001). In recent years, ERG recordings have found application to the mouse retina, where the ERG has been used to characterize retinal function following the manipulation of gene expression in photoreceptors and other retinal neurons (e.g., Ball et al. 2000; Biel et al. 1999; Calvert et al. 2000; Dhingra et al. 2000; Humphries et al. 1997; Lyubarsky et al. 2000; Masu et al. 1995; Ng et al. 2001; Ripps et al. 2000; Xu et al. 2000). Because the ERG can be recorded noninvasively, the response also provides a useful means to monitor the severity and progression of retinal degeneration (Goto et al. 1995; Olsson et al. 1992; Ren et al. 2000) as well as to evaluate the effects of experimental treatment strategies in animal models of hereditary retinal disease (Bush et al. 2000; Li et al. 2001; Naash et al. 1996; Sieving et al. 2001).

To fully understand the functional changes induced by gene manipulation and treatment strategies in mouse, it is important to define the basic properties of the mouse ERG. Most studies to date have examined the response to brief flashes of light in terms of the major ERG waveform components (e.g., the a and b waves) and how these vary with stimulus properties, age, and/or following gene manipulation. A complementary approach that can provide useful information about retinal status is to measure the ERG temporal response function. In this procedure, the ERG is recorded in response to sinusoidal stimuli that encompass a wide range of temporal frequencies, and the amplitude and phase of the response fundamental are derived from a spectral analysis of the waveform at each stimulus frequency. This method of analysis is particularly useful at high temporal frequencies at which the ERG response can be of low amplitude and difficult to discriminate from noise. Derivation of the response harmonics by this technique is also useful in characterizing nonlinearities in the ERG response (Burns et al. 1992; Odom et al. 1992).

The temporal response function for the cone ERG has been defined previously for human (Burns et al. 1992; Odom et al. 1992; Seiple et al. 1986) and monkey (Kondo and Sieving 2001). For both species, the fundamental response functions have a similar shape with a response minimum near 12 Hz, a peak near 40 Hz, and a high-frequency roll-off that extends to approximately 100 Hz. The shape of the cone ERG temporal response function for monkey (and presumably human) has been attributed to the vector summation of massed responses from the photoreceptors, depolarizing bipolar cells (DBCs) and hyperpolarizing bipolar cells (HBCs), based on pharmacological manipulation of these response components (Kondo and Sieving 2001). At frequencies near 12 Hz, Kondo and Sieving (2001) reported that the responses of the DBCs and HBCs are nearly 180° out of phase and speculated that cancellation between these response components would account for the response minimum. At frequencies near 40 Hz, these authors found that the responses of the DBCs and HBCs are more nearly in phase and suggested that they tended to combine so as to augment the measured fundamental response amplitude. The decline of response amplitude at high temporal frequencies appears to be governed by the temporal response properties of the cone photoreceptors (Burns et al. 1992), although the photoreceptors themselves make little direct contribution to the ERG response (Kondo and Sieving 2001).

Whereas the temporal response properties of the primate cone ERG have been well defined, less is known about the temporal response properties of the mouse cone ERG. A recent study approximated the temporal response function by recording mouse ERG responses to pairs of strobe-flash stimuli in which the two flashes were separated temporally by varying amounts (Ekestein et al. 1998). In contrast to the primate ERG, the mouse ERG showed relatively low response amplitudes at high temporal frequencies (i.e., short interflash intervals). In the present study, we used stimuli consisting of sinusoidal flicker to better define the temporal response characteristics of the normal wild-type (WT) mouse cone ERG. We have compared these results to those of humans recorded under identical stimulus conditions. In addition, we also made recordings from mice carrying the nob (no b wave) gene defect (Pardue et al. 1998). In the nob mouse, which is a model for human CSNB1, the DBCs do not appear to contribute to the ERG, but the responses of photoreceptors and HBCs appear to be spared (Pardue et al. 2001). Therefore a comparison of WT and nob ERG responses provides an opportunity to examine the relative contribution of DBCs and the associated visual pathway to the temporal response function of the mouse cone ERG.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

WT and nob mice, ages 1-4 mo, were studied. All were agouti in coat color, derived from a cross between C57BL/6 and other strains (cf. Candille et al. 1999). All procedures involving animals were approved by the local institutional animal care and use committee. For comparison purposes, ERG recordings were also obtained from two visually normal male humans, ages 44 (S1) and 22 (S2) years, from whom informed consent was obtained.

Stimulation

Light stimuli were derived from two optical channels, each with a 100-W tungsten halogen light source, that were combined using an Oriel bifurcated fiber optic bundle. Because the positions of the individual fibers within the fiber optic bundle are random, the inputs from the two channels were spatially co-extensive at the output end, which subtended 4.5 mm and was placed directly in front of the test eye. One channel provided a steady adapting field of 5 cd/m² to desensitize the rod system. For the mouse eye, this luminance corresponds to 706 photoisomerizations per rod/s, based on the assumption that 1 photopic cd/m² equals 1.4 scotopic cd/m² for the tungsten halogen light source (Wyszecki and Stiles 1982) and that 1 scotopic cd/m² is equivalent to 100 photoisomerizations per rod/s (Hetling and Pepperberg 1999). The second stimulus channel was modulated in a sinusoidal fashion using a ferroelectric liquid crystal shutter, driven by a Displaytech DR-95 driver. The driver was controlled in turn by a Keithley DAS-801 signal processing board housed within a microcomputer. The shutter was driven at a constant frequency of 1 kHz and was pulse-width modulated under computer control, with the duty cycle controlled by a linearized lookup table, to define the stimulus waveforms used here. Across trials, sinusoidal temporal frequencies ranged from 2 to 52 Hz, and modulation depth was held constant at 99%, with the exception of trials using a 0% modulation depth to measure the noise level. In different experimental sessions, the mean luminance of the stimulus channel was set at either 800 or 3,200 photopic cd/m², controlled with neutral density filters. Luminances were calibrated with a Minolta LS-110 photometer.

Recording

Mice were anesthetized with ketamine (80 mg/kg) and xylazine (16 mg/kg) and placed on a heating pad. The mouse ERG was recorded using a stainless steel electrode that made contact with the corneal surface through a thin layer of 1% methylcellulose. Needle electrodes placed in the cheek and the tail served as reference and ground leads, respectively. Human ERGs were recorded using a bipolar Burian-Allen contact lens electrode with the earlobe grounded using an earclip electrode. For both mouse and human recordings, the pupil of the test eye was fully dilated with 1% mydriacyl and 2.5% phenylephrine HCl. Responses were differentially amplified (1-1,000 Hz) and stored using a Diagnosys Espion signal-averaging system that was triggered by a TTL signal generated by the DAS-801 board and synchronized with the stimulus cycle, which was in sine phase. Each stimulus condition was repeated two or three times, and each trial averaged 25 response epochs that were each 2 s in duration.

Analysis

Fundamental response amplitudes (full peak-to-trough amplitudes) were derived from the power spectral densities of the ERG recordings, and response phases were derived from fast Fourier transforms, using the Matlab Signal Processing Toolbox. Recordings made to a stimulus of 0% contrast were used to provide an estimate of the noise level at each temporal frequency. The response to a sinusoidal stimulus at a given temporal frequency was considered to represent an actual ERG response if the amplitude derived from the power spectrum exceeded three times the noise amplitude at that temporal frequency, which provides a signal-to-noise ratio of 2.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Figure 1 presents a representative series of cone ERGs obtained from each type of subject at temporal frequencies ranging from 4 to 40 Hz. The waveforms in the left column were recorded from a WT mouse. The largest responses were obtained at low stimulus frequencies, and response amplitude declined systematically as the temporal frequency increased. For comparison, ERGs were also obtained from two human subjects using the same stimulus and recording system. The middle column of waveforms of Fig. 1 presents a series of responses recorded from S1; the second subject showed a similar pattern of findings. In contrast to the mouse ERG, the human ERG was maximal in amplitude near 40 Hz and showed a more complex waveform at the lower temporal frequencies. Figure 1, right, presents a series of cone ERGs obtained from a representative nob mouse. There were distinct differences between the waveforms of the nob and WT mice that were particularly apparent at low temporal frequencies. Nevertheless, both showed the largest response at the lowest temporal frequency and a systematic decrease in response amplitude at the higher temporal frequencies. For all subjects, the cone ERGs were nonlinear in shape but were more so for the human than for the mouse responses. At 10 Hz, for example, the amplitude of the second harmonic was 96% of the fundamental amplitude for the human response but was only 32% and 23% of the fundamental amplitude for the WT and nob recordings, respectively. At the highest stimulus frequencies, the contribution of the higher harmonics was negligible for all subjects, as the responses became more sinusoidal.

View larger version (43K): [in this window] [in a new window]   Fig. 1. Cone electroretinograms (ERGs) obtained to sinusoidal stimuli from representative a wild-type (WT) mouse (left), a visually normal human subject (middle), and from a nob mouse (right). Each record represents the mean of 25 successive response epochs. For clarity, only 750 ms of the 2,000-ms record is displayed. Stimulus waveforms are represented by the gray traces under each ERG waveform. Stimulus mean luminances for these recordings were 800 cd/m2 (WT mouse), 800 cd/m2 (human), and 3,200 cd/m2 (nob mouse).

Figure 2A plots the mean amplitude of the response fundamental for four WT mice, using stimulus mean luminances of 800 cd/m² () and 3,200 cd/m² (). In both cases, response amplitude increased slightly as stimulus frequency increased from 2 to 6 Hz. As temporal frequency was further increased, however, the amplitude of the ERG response fundamental declined steadily. The dashed line in Fig. 2A indicates the criterion response level that was used to distinguish between signal and noise. This line represents three times the average noise level obtained from all WT mice tested. The noise level was highest at the lowest frequency and then declined systematically with increasing temporal frequency. We restricted our subsequent analysis of the mouse cone ERG to temporal frequencies below 44 Hz, at which the response fundamental exceeded the noise level by a factor of 3 or more.

View larger version (21K): [in this window] [in a new window]   Fig. 2. Amplitude (A) and phase (B) of the response fundamental as a function of temporal frequency, using stimulus mean luminances of 800 cd/m2 () and 3,200 cd/m2 (). The dashed line in A indicates 3 times the noise level obtained to a stimulus of 0% contrast, averaged across all WT mice tested. Data points indicate the mean (±1 SE) response of 4 (800 cd/m2) or 3 (3,200 cd/m2) mice; error bars are omitted when smaller than the plotted point. In B, the solid lines indicate least-squares regression lines fit to the 800 and 3,200 cd/m2 data on linear coordinates, with slopes of 14.6 and 17.8°/Hz, respectively.

Figure 2B plots mean response phases at the same two stimulus luminances. Phases have been "unwrapped" to extend beyond 360° and are plotted in cosine phase, as per convention. For both luminance levels, the fundamental response phase decreased systematically with increasing temporal frequency above 6 Hz. The solid lines represent least-squares regression lines fit to the phase data on linear coordinates. The data obtained at 800 cd/m² were well fit (r² = 0.98) by a regression line with a slope of 14.6°/Hz, which is consistent with a simple response delay of 41 ms. The fit to the data obtained at 3,200 cd/m² was less satisfactory, suggesting that a simple response delay did not account for the phase data at this luminance.

The fundamental response function for the human ERG (S1) is shown in Fig. 3A () together with the mean response function of four WT mice replotted from Fig. 2A (). In agreement with previous studies (Burns et al. 1992; Odom et al. 1992), the human temporal response function displayed two peaks, occurring at approximately 4 and 40 Hz, which were separated by a response minimum near 12 Hz. As noted in the INTRODUCTION, this amplitude minimum for the response fundamental is thought to reflect a relative cancellation of out-of-phase signals generated by DBCs and HBCs (Kondo and Sieving 2001). The absence of such a response minimum in the response function for WT mice indicates that there may be less cancellation between DBC and HBC responses in this species.

View larger version (17K): [in this window] [in a new window]   Fig. 3. Amplitude (A) and phase (B) of the response fundamental for human () and mouse () ERG responses as a function of temporal frequency. Data points indicate the mean (±1 SE) response; error bars are omitted when smaller than the plotted point. Stimulus mean luminance: 800 cd/m2. The mouse data and least-squares regression line are replotted from Fig. 2B. The solid line through the human data represents a least-squares regression line fit to data points at 16 Hz and above on linear coordinates, with a slope of 9.9°/Hz.

Figure 3B presents the corresponding phase results for S1 () and the mean of 4 WT mice, replotted from Fig. 2B (). At low stimulus frequencies, the response phase of the human ERG was relatively constant. After a small phase increase near 10 Hz, the response phase declined steadily with increasing stimulus frequency. These results are similar to those reported previously for the human ERG (Alexander et al. 2000; Burns et al. 1992). The solid line through the human phase data in Fig. 2B represents a least-squares regression line fit to the data at 16 Hz and above on linear coordinates. The fit of this regression line is quite good (r² > 0.99) over this range, indicating that the phase data correspond to a response delay of 28 ms over this range. By comparison, the slope of the regression line fit to the phase data for the WT mice corresponded to a response delay of 41 ms, as noted in the preceding text.

Figure 4A compares the mean fundamental response amplitudes for three nob mice () and for three littermate WT animals () obtained at a stimulus luminance of 3,200 cd/m². At low temporal frequencies, the fundamental response amplitudes of the nob mice were smaller than those of the WT mice. As temporal frequency increased, however, the two amplitude functions approached one another, and overlapped at the highest temporal frequencies. A complementary pattern of findings was apparent for fundamental response phases (Fig. 4B). At low temporal frequencies, the fundamental responses of nob and WT mice were nearly 180° out of phase. At the highest temporal frequencies, however, the phases of nob and WT mice came into register.

View larger version (16K): [in this window] [in a new window]   Fig. 4. Amplitude (A) and phase (B) of the response fundamental for WT () and nob () ERG responses as a function of temporal frequency. Data points indicate the mean (±1 SE) responses of 3 WT and 4 nob mice; error bars are omitted when smaller than the plotted point. Stimulus luminance: 3,200 cd/m2.

To estimate the contribution of the DBC pathway to the WT cone ERG, we derived the vector difference between the nob response fundamental and that of the WT mouse, using the averaged data obtained at 3,200 cd/m² (Fig. 4A). Examples of the vector subtraction analysis for temporal frequencies of 4 and 16 Hz are shown in Fig. 5. At 4 Hz (Fig. 5, top), the responses of WT and nob mice have nearly opposite phases. As shown by the difference vector (WT-nob), subtraction of these opposite-phase responses yields a difference that is larger than either original response. At 16 Hz (Fig. 5, bottom), however, the WT and nob responses are more in phase. As a result, the difference vector (WT-nob) has an amplitude that is intermediate between the WT and nob responses.

View larger version (13K): [in this window] [in a new window]   Fig. 5. Representative vector plots obtained at 4 Hz (top) and 16 Hz (bottom). The length and direction of the arrows indicates, respectively, the mean amplitude and phase for nob and WT responses, as well as for their vector difference (WT-nob).

Figure 6 plots the amplitude of the vector difference obtained at each temporal frequency (), together with the fundamental response amplitudes of nob () and WT () mice, replotted from Fig. 4A. At low temporal frequencies, the vector difference (representing the inferred DBC contribution) was larger in amplitude than either the WT or the nob response. As illustrated by the vector plot for the 4-Hz data shown in Fig. 5, this reflects the nearly 180° phase difference between the fundamental responses of WT and nob mice. At higher temporal frequencies, however, the vector difference declined more quickly than did the functions for the WT and nob mice. This analysis suggests that the DBC contribution to the mouse ERG has a different dependence on temporal frequency than do the other contributors to the ERG response fundamental.

View larger version (15K): [in this window] [in a new window]   Fig. 6. Amplitude of the vector difference between nob and WT response fundamentals () as a function of temporal frequency, with the mean WT () and nob () response fundamentals shown for comparison. Error bars are omitted for clarity. Stimulus mean luminance: 3,200 cd/m2.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study examined the properties of the temporal response function of the mouse cone ERG in relationship to that of the human ERG. We observed that the temporal frequency response functions differed in several respects. First, the temporal frequency range for the mouse response was quite limited, and approached the noise level at frequencies that corresponded to the high-frequency peak of the human temporal response function. This finding is in agreement with a recent study of the temporal properties of the mouse ERG by Ekestein et al. (1998), who used a quite different approach, involving pairs of strobe flashes. Second, the mouse temporal response function did not exhibit the amplitude minimum normally seen in the human response function at temporal frequencies near 12 Hz. In primates, this minimum appears to reflect a frequency-selective degree of cancellation between DBC and HBC components (Kondo and Sieving 2001), and this does not appear to extend to the mouse retina. Third, the overall waveform of the mouse cone ERG was much less complex than was the human response. This was particularly evident at low temporal frequencies where the mouse response included a much smaller contribution from the higher response harmonics. Finally, the mouse response phase was linear with temporal frequency, corresponding to a simple response delay, whereas the relationship between response phase and temporal frequency was more complex for the human cone ERG. Overall, these differences indicate that the mouse cone ERG behaves in a more linear fashion than does the primate response.

The cellular origins of the temporal response function for the primate cone ERG have been studied recently by Kondo and Sieving (2001), using a pharmacological approach. According to their report, the ERG response fundamental represents the vector summation of the response fundamentals of at least three cell types: cone photoreceptors, DBCs and HBCs. Further, the relative contribution of these ERG generators varies with temporal frequency. At high temporal frequencies, the response is thought to be dominated by the activity of HBCs and DBCs with little direct contribution from the cone photoreceptors (Kondo and Sieving 2001), although the photoreceptors do appear to govern the shape of the high-frequency roll-off of the temporal response function (Burns et al. 1992). In addition, the DBCs and HBCs of the primate retina appear to have different temporal response characteristics (Kondo and Sieving 2001), so that their relative contribution to the response fundamental varies systematically with stimulus frequency. Finally, the phase relation between the DBC and HBC contribution also depends on stimulus frequency, so that the manner in which these contributions combine at the recording electrode produces distinct peaks and troughs in the response function where the DBC and HBC responses add together or tend to cancel one another.

The relative contributions of the components underlying the temporal response function of the mouse cone ERG are less clear. Nevertheless, some insight into this issue is provided in the present study by comparing the temporal response functions of WT mice and nob mice. The nob mutant mouse, which is a model for human CSNB1, appears to lack the DBC contribution to the ERG (Pardue et al. 2001). Therefore by vector subtraction of nob from WT responses, one can obtain an estimate of the contribution of the DBC pathway to the murine temporal response function, and identify how that contribution varies with temporal frequency. Vector subtraction indicated that at low temporal frequencies, the mouse cone ERG is dominated by the DBCs, but that the relative DBC contribution decreases systematically with increasing temporal frequency. The fact that the ERG of the nob mouse, which lacks the DBC response, was smaller than that of the WT mouse at low frequencies is in contrast to the results reported for the primate retina (Kondo and Sieving 2001), where a blockage of the light-evoked DBC response by 2-amino-4-phosphonobutyric acid (L-AP4) increased the amplitude of the response fundamental. This difference likely reflects species differences in the relative contributions of the DBCs and HBCs to the cone ERG.

The possibility that the DBC contribution to the mouse cone ERG might decline more quickly with increasing temporal frequency than the responses of the other ERG generators finds support from a study of single cell recordings made from bipolar cells of cold-blood vertebrates (Ashmore and Copenhagen 1980). In that study, the temporal frequency range of the DBCs was found to be substantially lower than that of the HBCs. This conclusion is also supported by Kondo and Sieving (2001), who found that L-AP4 had a profound effect on the amplitude of the primate cone ERG at low, but not high, temporal frequencies. A recent study examined the intensity-response functions and receptive field properties of murine DBCs and HBCs recorded in a slice preparation (Bernston and Taylor 2000). The temporal response properties of these cells were not investigated in that study, however, so the possibility that DBCs and HBCs differ in their temporal response properties remains to be tested directly.

The explanation for the reduced responsivity of the murine ERG at high temporal frequencies is unclear at present. Vector subtraction of nob and WT temporal response functions indicated that DBCs likely make a relatively small contribution to the response function at higher temporal frequencies. Therefore the high-frequency attenuation of the mouse ERG compared with the primate ERG could represent a relative attenuation of either the cone photoreceptor or the HBC contribution. Although a number of studies have examined the response characteristics of individual primate cone photoreceptors (Kraft et al. 1993; Schneeweis and Schnapf 1995, 1999), comparable recordings from mouse cone photoreceptors have not yet been reported. Further, because the light responses of mammalian bipolar cells has only been initially characterized (Bernston and Taylor 2000), the role of the HBCs in defining the high-frequency attenuation of the mouse cone ERG remains to be determined. In the future, pharmacological studies of the mouse cone ERG will be useful in defining further the relative contributions of the cone photoreceptors, HBCs, and DBCs, as would parallel studies of other mutant mice with defects in cone-to-bipolar cell communication (e.g., Ball et al. 2000; Dhingra et al. 2000; Masu et al. 1995).