Received 12 August 2010 | Accepted 26 December 2010 | Published 31 December 2010
Vittorio Porciatti, Tsung-Han Chou, William J. Feuer
Bascom Palmer Eye Institute, University of Miami Miller School of Medicine, Miami, FL
Correspondence to: Vittorio Porciatti, Bascom Palmer Eye Institute, University of Miami Miller School of Medicine, Miami, FL, 33136; Phone: (305) 326-6050, FAX: (305) 482-4567; email: email@example.com
Purpose: To characterize differences in retinal ganglion cell (RGC) function in mouse strains relevant to disease models. C57BL/6J (B6) and DBA/2J (D2) are the two most common mouse strains; D2 has two mutated genes, tyrosinase-related protein 1 (Tyrp1) and glycoprotein non-metastatic melanoma protein B (Gpnmb), causing iris disease and intraocular pressure (IOP) elevation after 6 months of age that results in RGC degeneration, and is the most widely used model of glaucoma. DBA/2J.Gpnmb+ (D2.Gpnmb+) is the wild type for the Gpnmb mutation and does not develop IOP elevation and glaucoma.
Methods: Young (2–4 months of age) B6, D2, and D2.Gpnmb+ mice (n=6 for each group) were tested with pattern electroretinogram (PERG) in response to different contrasts and spatial frequencies. PERG amplitude and latency dependencies on stimulus parameters (transfer functions) were established for each mouse strain, together with corresponding thresholds for contrast and spatial resolution.
Results: PERG analysis showed that B6, D2, and D2.Gpnmb+ mice had comparable contrast threshold and spatial resolution. Suprathreshold spatial contrast processing, however, had different characteristics in the three strains. PERG amplitude and latency changes with increasing contrast were different between B6 and D2 as well as between D2 and D2.Gpnmb+.
Conclusions: B6, D2, and D2.Gpnmb+ mice have different characteristics of PERG spatial contrast processing consistent with different mechanisms of contrast gain control. This may imply differences in the activity of underlying PERG generators and synaptic circuitry in the inner retina.
The two most common inbred mouse strains C57BL/6J (B6) and DBA/2J (D2) differ in several specific functions. These include differential sensitivity to nociceptive stimuli , taste , alcohol, barbiturates, and cocaine [3,4]. Visual behaviors, such as visual detection, pattern discrimination, and visual acuity, are reported to be similar in young (within 4 months of age) B6 and D2 mice . The electroretinogram (ERG) is also reported to be similar in young B6 and D2 mice [6,7]. However, retinal ganglion cell (RGC) population is reported to be significantly larger in D2 mice than in B6 mice . It is possible that there are differences in RGC function between B6 and D2 strains that are not reflected in measures of either visual behavior or ERG and that probe primarily the preganglionic retinal activity . As mouse models of RGC death, glaucoma, and optic neuropathy using B6 and D2 genetic backgrounds are increasingly used [10-12], we wanted to determine if there is a basic difference in RGC function between the two control B6 and D2 strains. We also wanted to determine if there is a difference between the most widely used D2 mouse model of intraocular pressure (IOP) elevation and glaucoma [13-15] and its control DBA/2J.Gpnmb+, which does not develop glaucoma , at ages before the development of high IOP in D2.
We used the pattern electroretinogram (PERG) to systematically investigate the physiologic characteristics of RGC response in 2–4-month-old mice. There is a large body of evidence that PERG reflects RGC electrical activity in mammals [9,17], including mice [18-21]. The PERG is currently used to probe abnormalities of RGC function in mouse models of glaucoma [22,23] and optic nerve disease [24,25].
Results show that the PERG spatial contrast gain control characteristics differ between B6 and D2 mice. PERG spatial contrast gain control characteristics also differ between D2 and D2.Gpnmb+ mice. Altogether, results suggest that neural processing involving RGC differs among these genotypes. Preliminary results of this study have been previously published in abstract form .
All procedures were performed in compliance with the Association for Research in Vision and Ophthalmology (ARVO) statement for use of animals in ophthalmic and vision research. The experimental protocol was approved by the Animal Care and Use Committee of the University of Miami. A total of 18 mice (B6, n=6; D2, n=6; D2.Gpnmb+, n=6; Jackson Labs, Bar Harbor, ME) were tested in the age range 2 to 4 months. Mice were maintained in a cyclic light environment (12 h:12 h light [50 lux]–dark) and fed ad libitum.
Detailed description of the PERG technique is reported elsewhere [19,20,27]. In brief, mice were weighed and anesthetized with intraperitoneal injections (0.5–0.7 ml/kg) of a mixture of ketamine (42.8 mg/ml) and xylazine (8.6 mg/ml). Mice were then gently restrained in a custom-made holder that allowed unobstructed vision. The body of the animal was kept at a constant body temperature of 37.0 °C using a feedback-controlled heating pad (TCAT-2LV; Physitemp Instruments, Inc. Clifton, NJ).
A PERG electrode (0.25 mm diameter silver wire-World Precision Instruments, Sarasota, FL-configured to a semicircular loop of 2 mm radius) was placed on the extrapupillary corneal surface by means of a micromanipulator. A small drop of balanced saline was topically applied every 30 min to prevent corneal dryness. Reference and ground electrodes were stainless steel needles (Grass, West Warwick, RI) inserted under the skin and scalp (reference) and tail (ground).
Visual stimuli consisted of contrast-reversing (1 Hz, 2 reversals) horizontal bars generated by a programmable graphic card (VSG-; Cambridge Research Systems, Rochester, UK) on a cathode-ray tube (CRT) display (Sony Multiscan 500, Sony Electronics Inc., San Diego, CA) with the center aligned with the projection of the pupil. The pupils were not dilated, and eyes were not refracted for the viewing distance since the mouse eye has a large depth of focus [28-30]. At the viewing distance of 15 cm, the stimulus field covered an area of 69.4×63.4°. Patterns had fixed mean luminance of 50 cd/m2 and variable contrast (0.1 to 1 in ten steps) and spatial frequency (0.05 to 0.8 cycles/degree in five steps). The luminance of the CRT display was γ-corrected using a photometer (OptiCal OP200-E; Cambridge Research Systems Ltd., Rochester, UK). Contrast was defined as C=(Lmax–Lmin)/(Lmax+Lmin), where Lmax=luminance of the bright stripes and Lmin=luminance of the dark stripes .
Three consecutive PERG responses to 600 contrast reversals each were recorded. The responses were superimposed to check for consistency and then averaged (1,800 sweeps). The waveform of averaged PERGs to high-contrast (1.0) gratings of low spatial frequency (0.05 cycles/deg) consisted of a major positive peak at around 90–120 ms (defined as P100) followed by a slower negative wave with a broad trough at around 200–300 ms (defined as N250, examples in Figure 1). Note that the human transient PERG also consists of a positive–negative complex. However, the human positive wave peaks at about 50 ms (P50), and the trough of the subsequent negative wave occurs at about 95 ms (N95). It is commonly thought that the N95 wave is more specifically related to RGC function and is more affected than the P50 wave in optic nerve disease . In contrast, the P50 wave is thought to have a preganglionic origin and be affected in macular diseases . In the mouse transient PERG, the positive (P100) and negative components (N250) do not appear to dissociate in disease models; both the P100 and the N250 components are altered in glaucoma  as well as after selective RGC degeneration induced by optic nerve crush . The PERG responses represented in Figure 1 were obtained under conditions that maximize response amplitude (0.05 cycles/deg, 1.0 contrast), thereby yielding a robust response—defined here as maximal PERG—that has been used in several studies on mouse models of optic neuropathies [19,22-24,33-35]. Both P100 and N250 components were evaluated.
In the present study, the entire dynamic range of the PERG response to spatial contrast was investigated. As the level of PERG signal progressively decreased with decreasing contrast and increasing spatial frequency, manual identification of P100 and N250 components would have potentially introduced operator bias in waveforms close to response threshold. To prevent this, maximal voltage in the expected time window for P100 (50–200 ms) and minimal voltage in the expected time window for N250 (201–350 ms) were automatically identified using a simple macro written in Sigmaplot language (version 11.2; Systat Software, Inc., San Jose, CA). For the analysis of contrast transfer function and spatial transfer function, response amplitude was defined as the peak-to-trough voltage (P100–N250); response latency was defined as the time-to-peak of the P100 wave. The latency of the N250 component was not systematically investigated since, in many instances, was rather broad, precluding accurate peak-time measurement of this component. The time-to-peak of the negative trough (N95) of the human transient PERG is not currently evaluated for the same reason .
For statistical analysis, responses of the two eyes were averaged and used as a single entry. Strain differences in absolute amplitude and latency of P100 and N250 components of maximal PERG (0.05 cycles/deg, 1.0 contrast) were analyzed with Students t- tests. To compare transfer functions, peak-to-trough (P100-N250) response amplitudes and P100 latencies were first normalized to the maximal PERG. The normalized PERGs of B6 and D2.Gpnmb+ mouse strains to different contrast- and spatial frequency stimuli were then each compared to D2 mice with a two-factor subject (mouse strain) by repeated measures (stimulus levels) analysis of variance (ANOVA) with orthogonal polynomial decomposition, followed by post hoc t tests. A p value of <0.05 was considered significant.
Examples of maximal PERGs in response to contrast reversal gratings (temporal frequency=1Hz, spatial frequency=0.05 cycles/deg, contrast=1.0) for the three mouse strains are displayed in Figure 1 as group averages±standard error of the mean. It is apparent in Figure 1 that in B6 mice the PERG tended to have a shorter latency compared to both D2 and D2.Gpnmb+, whereas waveforms were similar in D2 strains. Evaluation of P100 and N250 components was performed on individual waveforms and their mean displayed in Figure 2. The amplitude of the P100 component tended to be smaller in B6 than in D2, but the difference was not significant (t test, p=0.19). The P100 component had a similar amplitude in D2 and D2.Gpnmb+. The N250 component had virtually identical amplitude in all strains. On average, the latency of the PERG P100 component was substantially shorter in B6 mice than in D2 mice by about 22.7 ms (t test, p=0.001), whereas the latency of D2 and D2.Gpnmb+ was similar.
Figure 3 shows how the PERG amplitude and latency change as a function of stimulus contrast for a fixed spatial frequency of 0.05 cycles/degree. To appreciate differences in the function among strains, all data were expressed as relative changes compared to the maximal PERG, waveforms and absolute values of which are shown in Figure 1 and Figure 2. With decreasing contrast, the PERG amplitude progressively decreased while the latency progressively increased in all strains. However, there were notable differences among strains. As shown in Figure 3A, in B6 mice the contrast function of amplitude was approximately linear over the entire contrast range, whereas in D2 the contrast function had a more complex shape. In particular, the function was approximately linear between 0.2 and 0.6 contrast, displayed a local minimum (notch) at 0.8 contrast, and a second linear branch at 0.8–1.0 contrasts. The response latency (Figure 3C) increased approximately linearly with decreasing contrast in both B6 and D2. The slope of latency increase with decreasing contrast tended to be steeper in B6 compared to D2. At contrast of 0.1, PERG responses of both B6 and D2 were indistinguishable from a control response obtained with the stimulus occluded (noise) and were not included in the figure. We considered the contrast threshold being located at some point between contrasts of 0.1 and 0.2.
As shown in Figure 3B, the form of the contrast amplitude appears to be different between D2 and D2.Gpnmb+ mice. In particular, the amplitude notch at 0.8 contrast visible in D2 mice was not present in D2.Gpnmb+ mice. At low (0.2–0.3) contrast the PERG amplitude was relatively higher in D2.Gpnmb+ mice compared to D2. The response latency (Figure 3D) increased approximately linearly with decreasing contrast in both D2 and D2.Gpnmb+ mice. The slope of latency increase with decreasing contrast tended to be steeper in B6 compared to D2. At contrast of 0.1, PERG responses of both D2 and D2.Gpnmb+ mice were indistinguishable from a control response obtained with the stimulus occluded (noise) and were not included in the figure. We considered the contrast threshold being located at some point between 0.1 and 0.2 contrast.
B6 versus D2-- There was a statistically significant strain by contrast level interaction in normalized PERG amplitude means (p=0.041, repeated measures ANOVA; see panel A). Post-hoc t-tests revealed no significant differences between strains for contrast levels 0.2 thru 0.6; however, there were significant differences between strains for contrast levels 0.8 (p=0.050) and 0.9 (p=0.029). Latency became significantly greater with decreasing contrast in B6 mice compared to D2 mice (p=0.015, panel C), while no interaction was observed (p=0.25). In summary, the form of PERG amplitude contrast function significantly differed between B6 and D2 strains at high contrasts. PERG latency significantly differed between B6 and D2 strains, but the form of the latency transfer function was similar.
D2 versus D2.Gpnmb+-- There was a statistically significant strain by contrast level interaction in normalized PERG amplitude means (p=0.031, repeated measures ANOVA; see panel B). Post hoc t tests revealed differences that were significant at contrast levels 0.2 (p=0.031), 0.3 (p=0.004), and 0.8 (p=0.015). There was a highly significant strain by contrast level interaction in normalized PERG latency means (p=0.001, repeated measures ANOVA; see panel D). Post hoc t tests on PERG latency could have been performed; however, orthogonal polynomial decomposition revealed that the interaction was due to differences in the slopes of the linear relation of PERG response to contrast between the two strains of mice (p=0.001). In summary, the form of PERG amplitude contrast function significantly differed between D2 and D2.Gpnmb+ strains at both low and high contrasts. PERG latency significantly differed between the two strains, but the form of the latency transfer function was similar.
Figure 4 shows how the PERG amplitude and latency change as a function of spatial frequency (range 0.05–0.8 cycles/deg) for a fixed temporal frequency of 1 Hz and contrast of 1. As for the contrast functions shown above, all data were expressed as relative changes compared to the maximal PERG, waveforms and absolute values of which are shown in Figure 1 and Figure 2. With increasing spatial frequency, the PERG amplitude progressively decreased, while latency increased, in all strains. At 0.8 cycles/degree, the PERG amplitude was just above the noise level (the amplitude of a response with the stimulus occluded) in all strains. We considered this spatial frequency as an index of retinal visual acuity. The PERG latency at 0.8 cycles/degree was not included in the figure since at this spatial frequency the signal was very close to the noise level and the automatic peak evaluation produced unreliable estimates.
Statistical comparisons were performed with the same approach used for the contrast function shown in Figure 3. That is, the normalized PERG means of B6 and D2.Gpnmb+ mouse strains to different spatial frequencies were each compared to D2 mice with a two-factor subject (mouse strain) by repeated measures (contrast level) ANOVA.
B6 versus D2-- There was no statistically significant difference in PERG amplitude between strains (p=0.26) and strain by spatial frequency interaction (p=0.75, panel A). There was no statistically significant difference in PERG latency between strains (p=0.37) and strain by spatial frequency interaction (p=0.55, panel C). In summary, there were no significant differences in the form of both amplitude and latency spatial functions between B6 and D2 strains.
D2 versus D2.Gpnmb+-- There was a statistically strain by spatial frequency interaction (p=0.011, panel B) in PERG amplitude. Post hoc t tests revealed significant strain differences at 0.2 cycles/deg (p=0.032) and at 0.4 cycles/deg (p=0.028). There was a statistically significant difference in PERG latency between strains (p=0.001) but no strain by spatial frequency interaction (p=0.89, panel D). In summary, there was a significant difference in the form of amplitude function between D2 and D2.Gpnmb+ strains for intermediate spatial frequencies. PERG latency significantly differed between the two strains, but the form of the latency transfer function was similar.
The PERG is a specialized kind of ERG that reflects inner activity and it represents an effective tool to assess normal and abnormal RGC function. In mouse models of optic nerve degeneration, the PERG may help to understand how genetic diversity relates to specific differences in RGC function and susceptibility to stress . In this study we have used the PERG to characterize the spatial contrast properties of RGC response in the two most common inbred mouse strains, C57BL/6J (B6) and DBA/2J (D2), which are used in several disease models. We also tested a relevant DBA/2J substrain, D2.Gpnmb+ that has a wild-type glycoprotein non-metastatic melanoma protein B (Gpnmb) allele but no other known differences to modern D2 mice .
Our results show that the PERGs of B6, D2, and D2.Gpnmb+ displayed many similarities but some notable differences. The waveform of PERG obtained under conditions that maximize the signal (spatial frequency 0.05 cycles/degree, max contrast 1.0)  differed between B6 and D2 strains. In D2 strains the PERG had a substantially longer latency (about 20 ms) compared to B6. No obvious differences in amplitude and latency between D2 and D2.Gpnmb+ strains were observable. The major difference among the three mouse strains was the way PERG amplitude and latency changed as a function of spatial contrast (contrast transfer function). In B6 mice, the contrast transfer function of PERG amplitude was linear, whereas in D2 there was a clear notch in PERG amplitude at a contrasts of 0.8. In addition, the slope of latency increase associated with decreasing contrast was shallower in D2 compared to B6. The amplitude notch at 0.8 contrast occurring in D2 mice was not present in D2.Gpnmb+ mice, which had a more robust response at lower contrasts. In addition, the slope of latency increase associated with decreasing contrast was shallower in D2 compared to D2.Gpnmb+ mice. Amplitude notches at high contrast have been reported before for the visually evoked potentials (VEP) [38,39], but their origin is still a matter of speculation. One possibility is that the notch originates from the interaction between different underlying neural generators that respond with different latency, resulting in amplitude cancellation.
Overall, differences in contrast transfer functions of amplitude and latencies can be understood in terms of different mechanisms of contrast gain control in the PERG generators . Photoreceptors do not adapt to contrast, whereas RGC typically display substantial gain control . Contrast gain control is a mechanism whereby RGC adjust their responsiveness (both in terms of amplitude and latency) through feedback conductances, thereby allowing more efficient use of their dynamic range. Contrast gain control mechanisms are expected to play a major role at sites where there is a large convergence of neural inputs to a target neuron . As in the mouse retina there is a large convergence between photoreceptors and RGC, this might explain the remarkable changes of the PERG signal latency with changing contrast that we found. Altogether, our results suggest that neural processing in the inner retina for suprathreshold contrast stimuli differs between B6 and D2 mice. D2 and D2.Gpnmb+ also displayed differences for suprathreshold contrast stimuli. At threshold contrasts, however, contrast gain control mechanisms are expected to play a lesser role. In all strains, the PERG contrast threshold was very similar, in the order of 10%–20%. This value is in keeping with previous reports in B6 mice obtained with PERG , VEP , and optomotor [43-45] studies but somewhat lower than that obtained with optokinetic response [46,47] and intrinsic optical imaging .
The spatial frequency function of PERG amplitude (obtained at maximum contrast) was similar in B6 and D2 mice but displayed subtle differences between D2 and D2.Gpnmb+. The spatial frequency threshold (acuity) was about 0.8 cycles/degree in all strains. This value is in keeping with previous reports on visual acuity in B6 mice obtained with PERG [18,49], VEP [42,49-51], and behavior  but somewhat higher than that reported for optomotor response [45,53], intrinsic optical imaging , and swim tasks .
With increasing spatial frequency, the PERG latency increased dramatically over the spatial frequency range in all strains. This space–time association lends further support to the notion that considerable convergence and spatial summation is at play in inner retinal circuitry [54,55]. Convergence and spatial summation are mechanisms whereby more synapses are simultaneously activated with increasing stimulus size (decreasing spatial frequency), leading to a larger compound synaptic potential that reaches threshold faster .
RGC population in DBA/2J mice is reported to be significantly larger (63,351±1208) than that of C57BL/6J (54,630±874) . This may have a counterpart in a different inner retina circuitry between B6 and D2 mice resulting in different spatial contrast functions. Differences in the PERG spatial contrast function between D2 and D2.Gpnmb+ mice, however, are likely to result from factors other than RGC number. D2 mice have mutations in two genes, the tyrosinase-related protein 1 (Tyrp1, which is linked to iris stromal atrophy) and the transmembrane glycoprotein nmb (GpnmbR150X, which is linked to iris pigment dispersion) [13,57]. The function(s) of the Gpnmb gene are not well known. Gpnmb influences the glaucoma phenotype of D2 mice . D2 mice wild type for the GpnmbR150X mutation (D2. Gpnmb+) develop mild iris disease and modest IOP elevation but not glaucomatous nerve damage . Low levels of GPNMB protein are also expressed in the neuronal retina of DBA/2J mice  and monkeys . Differential expression of GPNMB in the inner retina of D2 and D2.Gpnmb+ mice may have a counterpart in a different RGC function.
In summary, PERG analysis shows that B6, D2, and D2.Gpnmb+ mice have comparable thresholds for contrast and spatial frequency. Suprathreshold spatial contrast processing, however, has different characteristics in these mouse strains, implying different synaptic circuitry in the inner retina. It remains to be established whether these differences have a counterpart in susceptibility to RGC to insult or disease.
This study has been supported by grants NIH R01EY019077, NIH core center grant P30EY014801, and by an unrestricted grant to Bascom Palmer Eye Institute from Research to Prevent Blindness, Inc.