Numerous studies have shown, at the behavioral level, that the visual system can adjust itself to different visual environments 
. One of the most well known examples of this is the shift in spatial frequency sensitivity that occurs with the change from night (scotopic) to day (photopic) vision 
. This shift serves presumably as an information-optimizing strategy: at night, i.e., under photon-limited conditions, where the signal-to-noise ratio is low, the visual system is better served by integrating over a large area, so it shifts its tuning toward low spatial frequencies. During the day, when photons are not limiting, the system is better served by integrating over smaller areas, so it can resolve image details; in this case, the shift is toward high spatial frequencies (reviewed in 
How the visual system performs this shift is not clear. A large body of evidence, though, points to the retina as the starting point since the shift is detectable at the level of the ganglion cells 
. What remains to be determined is the mechanism that confers this on the cells. The most likely candidate is a change in the surround component of the ganglion cell's receptive field, as it is the surround that shapes the amplitude of the ganglion cell's response at low spatial frequencies. Changes in surround size cause the cell to shift its response toward or away from low spatial frequencies (see refs. 33, 35, and 6 for detailed quantitative analysis of how center and surround parameters affect the shape of the ganglion spatial tuning curve).
A long-standing proposal for how surround size might change with different light levels is that it might do so through a change in the gap junctional coupling of horizontal cells. The rationale for this hypothesis is that the extent of horizontal cell coupling is dependent on ambient light intensity 
. Thus, a change in horizontal cell coupling can serve as a natural knob for adjusting surround size and, therefore, the spatial tuning of the ganglion cells.
Here we tested this proposal. We used the mouse as a model system. We first measured ganglion cell spatial tuning at scotopic and photopic light levels in wild-type animals. As expected, the tuning shifted from low to high spatial frequencies as light intensities were increased from a lower to a higher level. We then measured the spatial tuning in Cx57-deficient mice in which horizontal cell coupling was reduced by >99%. If horizontal cell coupling plays a critical role in the adjustability of ganglion cell spatial tuning, then the shift from low to high spatial frequencies should be abolished. Our results indicated that it was not (). The shift from low to high spatial frequencies was essentially identical to that observed in wild-type mice. Direct measurements of ganglion cell surround size then confirmed this: If horizontal cell coupling plays a major role in the adjustability of ganglion cell surround size, then the shift from “no surround” to “small surround” should be abolished. It wasn't. The shift was essentially identical to that observed in the wild type (). Finally, behavior measurements provided further confirmation. No difference in spatial tuning sensitivity was observed between the Cx57-deficient and wild-type animals ().
These results thus provide strong evidence that changes in the coupling of horizontal cells is not a dominant mechanism for controlling the spatial tuning of ganglion cells. Most significantly, it does not appear to be a critical player in the adjustability of the tuning that occurs with changes from night to day vision. Other processes must dominate. Our measurements with dopamine confirmed this: dopamine's effects on the spatial tuning of ganglion cells could not have been mediated by a change in horizontal cell coupling since dopamine led to the same shift in spatial tuning in Cx57-deficient mice as in wild type, at least under photopic conditions (). This raises the idea that dopamine's dominant effects with respect to spatial tuning are on other retinal pathways e.g., affecting other electrically coupled networks in the retina 
, most likely amacrine cell networks 
In sum, with the aid of a Connexin57 knock out, we were able to test the long-standing hypothesis that the coupling and uncoupling of horizontal cells serves as a critical knob for adjusting spatial tuning to different light conditions, i.e., to night versus day conditions. The results show that this hypothesis, at least as it currently stands, must be rejected. The evidence for rejection is extremely strong because the same result presented itself at multiple levels–that is, when changes in horizontal cell coupling were prevented, as was the case in the knock out, the shift in spatial tuning that occurs when the retina moves from night to day proceeded normally–as measured at the level of both ganglion cell performance and whole animal behavioral performance. Thus, changes in horizontal cell coupling cannot be the critical mechanism that underlies this shift.
At first glance, it might seem surprising that preventing the changes in horizontal cell coupling–an act that affects lateral signaling in the retina–had no significant effect on ganglion cell spatial tuning, but this result can be reconciled with the many recent reports that this tuning is shaped by more than one set of circuits–that is, it is shaped by circuits in both the outer and inner retina 
. What the results of our experiments suggest is that inner retinal circuits dominate–at least for the problem of adjusting spatial tuning to different light conditions. Whatever occurs when the horizontal cells change from the uncoupled to the coupled state is effectively swamped by stronger circuit actions that occur in the inner retina.
This raises the intriguing question of what the changes in horizontal cell coupling are for. One possibility is that they serve to facilitate signal detection in the time domain, rather than the space domain. A change in horizontal cell coupling, because it is a change in the state of a potential shunt 
, would be expected to affect both spatial and temporal signal detection. If its effects on spatial signal detection are redundant to those produced by the inner retina, then losing the coupling would have minimal effect on spatial processing. If its effects on temporal
signal detection are not redundant, then losing it should affect temporal processing. This work thus creates a new hypothesis for the function of the horizontal cell coupling–that it serves to improve signal-to-noise ratios in the time domain, and, therefore, may be a key player in temporal processing.
Finally, for the sake of completeness, we conclude by stating that we cannot completely rule out the possibility that there is another connexin that links horizontal cells. However, if one exists, the likelihood that it contributes substantially to horizontal cell coupling is very small. The reason we state this is that the effects of knocking out Cx57 on horizontal cell coupling are maximal or near maximal, as measured by changes in both dye coupling and horizontal cell length constant. Dye coupling, using neurobiotin, is >99% abolished 
, and horizontal cell length constants are significantly reduced 
, with a reduction greater than that produced by dopamine application, which also reduces horizontal cell coupling (the hierarchy of length constant reduction is shown in Supp. Info. Fig. S3
). With respect to receptive field evaluations: horizontal cell length constants in the knockout are on average 50 µm, with the mean dendritic tree diameter for individual horizontal cells at 100 µm 
. Taken together, these data provide strong evidence that Cx57 is the primary, or exclusive, mediator of horizontal cell coupling, and that eliminating its ability to function provides a strong test for the role of horizontal cell coupling in retinal processing.