Some observations have been made regarding the nature of cone signal inputs to mouse ganglion cells. Although the results from a pair of single-unit recording studies [36, 37] are not entirely consistent, a majority of mouse ganglion cells are alleged to receive a summative input from the two pigment types with their relative contributions varying from cell to cell. In addition, signals derived from stimulation of UV and M cone pigments are combined in an opponent fashion in a small minority of the ganglion cells (somewhere between 2% and 13% of all cells). What is unusual is the observation that 20% or more of all ganglion cells seem to receive exclusive inputs from either UV or M pigment. Given the substantial convergence of cones to ganglion cells, these results would imply that there must be large patches in the receptor mosaic where one pigment type is expressed exclusively.

In sum, although some progress has been made in understanding the nature of cone signal pathways in the mouse retina and their functional implications, much obviously remains to be learned.

CONE-BASED VISION IN MICE

In one way or another, most of the studies that use mice as models for human retinal disease attempt to draw inferences about the linkages between retinal mechanisms and seeing. To do this in any realistic fashion requires an understanding of the nature of vision in the mouse. Here, I consider what has been learned about cone-based vision in the mouse.

Assessment Techniques

As the mouse became a principal target for examining linkages between genes and the nervous system, numerous assays were developed to assess the emergent effects of gene alteration. Included in these are various tests of the mouse visual system [38], and among these are applications of electrophysiological indices as well as approaches that involve a number of different behavioral responses. Although there are often strong correlations between the results obtained using these different indices, there can also be significant differences. Such differences are hardly surprising since various tests can index quite different levels of visual system analysis. For example, a basic measure of visual capacity is spectral sensitivity. As described next, measurements of mouse spectral sensitivity can vary greatly depending on the assay technique, with retinal electrophysiological studies yielding results quite different from those embodied in behavioral responses.

Behavioral assays of visual capacity in mice fall into two categories. One involves the use of appetitive procedures in which the animal is reinforced, usually by the transient availability of food or fluid, for correctly selecting one visual stimulus from another (or of one among several) [15, 39]. In these experiments, stimulus choice is indicated by the performance of an operant response, such as touching a panel or selecting a particular path to approach the stimulus. The differences between reinforced and nonreinforced stimuli are then progressively changed to obtain threshold measurements of visual capacity. A second general approach employs escape behavior to assess visual capacity. For instance, mice can be trained to escape from water by swimming to a raised platform, the location of which is indicated by a visual stimulus. As in the appetitive

reinforcement paradigms, the nature of visual cue can then be varied across presentations to obtain a measure of visual capacity. A variety of different configurations of the water escape task have been described [40–42]. In addition to testing behavioral discrimination capacities, examination of mouse vision has also been accomplished by exploiting reflex responses that are elicited by visual stimulation. For instance, animals often make reflexive tracking movements of the eyes, the head, or the whole body to patterned stimuli that are drifted across the visual field [43–45]. The value of the stimulus at which these optokinetic nystagmus (OKN) responses disappear is interpreted as defining a detection threshold.

Each of these methods to assess visual capacity in mice has advantages and disadvantages. By exploiting the motivation of animals to seek resources required for survival, the appetitive tasks come closest to tapping natural visually guided behaviors. However, such paradigms require extensive pretraining by someone skilled in behavioral methodologies and are thus not appropriate for screening programs in which it is necessary to examine large numbers of animals. Escape behaviors such as the various swimming tasks require little training and thus provide a more efficient way to examine mouse visual capacity. There are potential downsides to this procedure as well since this task cannot be used in mice with memory deficits or panic reactions (as may be common in some mouse strains), and it has also been observed that 5–10% of mice in numerous strains do not for one reason or another perform well in swimming tasks [42]. Evaluations of mouse vision using optomotor responses can be done rapidly and reliably [44]. A potential disadvantage of such tests is that the visual pathways subserving OKN responses are largely different from those underlying discriminative visual behaviors; consequently, they may provide a quite different picture of the relationships between retinal organization and vision.

All mammals feature duplex retinas in which rods subserve vision under low light levels, rods and cones function jointly at intermediate light levels, while at still higher light levels rod signals saturate, and cones become the sole functional photoreceptor at higher light levels. A problem is that the illumination levels defining the boundaries between these separate functional ranges depend on the organizational features of any particular visual system; thus, for example, what defines an illumination level that is sufficient to ensure cone-based vision in one species may or may not apply to another species or, indeed, even for animals of the same species carrying genetic manipulations that influence retinal organization.

How can one ensure that behavioral tests of cone vision in mice actually reflect the operation of cones? One solution is purely empirical: Obtain a measurement of spectral sensitivity in the test situation to show that spectral sensitivity function is consistent with what is expected from mouse cones (i.e., predictable from knowledge of the functions described next). For many investigations, this strategy is either too complicated or too time consuming. An alternative is to make use of what has been learned about rod and cone thresholds in the mouse.

One starting point for accomplishing this is a recent tutorial that provides useful instruction on how to calculate the effectiveness of standard human luminance measurements in terms of their ability to photoisomerize mouse photopigments [46]. Direct measurements of animals with that retinas have been engineered to lack cone photoreceptors showed that it requires a light sufficient to produce 104–105 photoisomerizations/rod/s to effect rod saturation [47], and measurements made on mice in which

rod and cone physiology had been manipulated suggested that cone-mediated vision in the mouse is some 10,000 times less sensitive than rod-mediated vision [42].

A potential complicating feature in behavioral studies is the responsiveness of the mouse pupil to ambient illumination. Mouse pupils are unusually sensitive to dim illumination and over their range of constriction can serve to reduce retinal illumination by a factor of about 20 [48]. One effect of pupil dynamics is to keep retinal illumination below the level required for rod saturation in many laboratory situations employing relatively dim stimulus sources (e.g., many computer monitors). There is one other possible complication. An investigation of C57/BL6 mice revealed the surprising fact that pupillary function may be effectively lost in animals that are greater than about 4 months of age [48]. A consequence is that, for any specific lighting circumstance, the effectiveness of the light may be considerably greater for older than for younger animals.

Spectral Sensitivity

Spectral sensitivity functions show the relative weighting of photic inputs as a function of their wavelength composition and thus serve to index a basic feature of visual capacity. In general, these functions reflect the characteristics of the underlying photopigments, the presence of any ocular screening filters, and the manner in which the photopigment signals are processed in the visual system.

Cone-based spectral sensitivity functions derived for the mouse are illustrated in Fig. 3A. The function shown at the top is based on ERG recordings. In line with the observation given here that mouse retinas expresses, overall, more UV than M pigment, sensitivity is highest in the UV wavelengths. Over the wavelengths tested, the mouse lens provides fairly modest differential absorption of light (curve in Fig. 3B). After correction for lens absorption, the ERG data were best fit with a linear summation of absorption curves appropriate for the two classes of cone pigment found in the mouse retina (continuous curve). The lower curve in Fig. 3A plots spectral sensitivity functions obtained from three mice in a behavioral test that involved a visual discrimination task in which animals were required to detect the presence of briefly presented monochromatic test lights that were superimposed on a steady achromatic background. That function was fitted with photopigment absorption curves in the same manner as for the ERG results.

Although contributions from both mouse cone photopigments can be detected in outer retinal signals and in behavior, the relationship between the two is unusual. Typically, short-wavelength-sensitive cones make only small contributions to ERG spectral sensitivity curves and then show a greatly enhanced contribution to increment threshold spectral sensitivity determined in the same fashion as the data of Fig. 3A. The reason for this is that increment thresholds tend to favor contributions made by spectrally opponent mechanisms, while signals recorded in the outer retina do not [49]. As was indicated, cone pigment coexpression in mouse cones may result in attenuated cone opponency in the mouse visual system, and this may explain why the spectral sensitivity functions measured in these different ways in the mouse have what seems an odd relationship. Whether that explanation is correct or not, these measurements of mouse spectral sensitivity functions make an important point: Estimates of spectral sensitivity derived from retinal measurements give an inadequate prediction of how a behaving mouse weights the spectral distribution of retinal illumination.