| |
The duplex retina and shared neural pathways
Overview
Three aspects of the retinal neural substrate are key for our understanding of rod-cone interactions. (This discussion will focus on primate retinal processing because of its direct relevance for human vision.) First, there appear to be no private rod pathways out of the retina. No retinal ganglion cells (the final retinal neural stage, whose axons form the optic nerve) have been found that transmit only rod-initiated signals. All rod signals appear to converge onto neural pathways that also transmit cone-initiated signals. There are two fundamental consequences of this sharing of neural pathways. First, it is inevitable that rod and cone signals will contribute jointly to visual function, given their shared operating range of light levels. Second, the sharing of pathways provides the occasion for a wide range of combination effects: facilitation to suppression, superadditivity to subadditivity, and so on.
The second key feature is that there are multiple sites of interaction of rod and cone signals. Two sites that will be examined in more detail below are (1) the direct electrical (gap junction) synapses that are formed between rod and cone photoreceptors at the first stage of retinal neural processing and (2) the convergence of rod signals transmitted through the AII amacrine cells onto bipolar cells that receive signals from cones. The spotty evidence that we have already, the apparent multiplicity of subtypes of retinal neurons, and the complexity of their interconnections (especially those involving amacrine cells) all make it likely that there are many other points of convergence or interaction of rod and cone signals that are also important for vision in humans, primates, and probably most vertebrates.
The third key feature is that the strength, presence, and sometimes even direction of rod-cone interactions depend on functional network properties of the retina that can be exquisitely sensitive to states of adaptation and patterns of stimulation. These network properties are dynamic and can't be captured in a static wiring diagram. Gap junctions are likely to be important in the formation and state dependence of these network properties (Smith et al., 1986).
Given these features of retinal processing, it is not surprising that rod-cone interactions are found in most or perhaps all types of visual processing and that the properties of these interactions vary tremendously under different circumstances. Two of the better-documented pathways of rod-cone interaction in the retina are highlighted immediately below.
Best Known Interaction Sites
Rod-cone gap junctions
Although anatomical processes that appeared to connect the terminals of primate rod and cone photoreceptors (Fig. 55.1) had been described earlier (Cohen, 1965; Kolb, 1970), it was not until the 1970s that these were definitively identified as gap junctions (electrical synapses) (Raviola, 1976; Raviola and Gilula, 1973) in several vertebrate species, including primates. Nelson showed that in cat retina, rod signals could be recorded from cone photoreceptors and from horizontal cell bodies, which had direct connections to cones but not to rods (Nelson, 1977; Nelson et al., 1976).
Figure 55.1..
Diagram of two pathways for rod-cone interaction in primate retina. The classical rod pathway extends from rods to the rod bipolar (RB) and AII amacrine (RA) to depolarizing and hyperpolarizing cone bipolars (DCB and HCB). In the second pathway, rod signals enter cones directly via gap junctions. (Reprinted from Trends in Neurosciences, vol. 3, Daw, Jensen, and Brunken, Rod pathways in mammalian retinae, pp. 110–1151, © 1990, with permission from Elsevier Science Publishers.)
More recently, rod signals have also been recorded from cones (Schneeweis and Schnapf, 1995) and H1 horizontal cells (Verweij et al., 1999) in primate retina. As in the cat and other mammalian species, the rod signal recorded from primate horizontal cell bodies apparently reflects indirect rod input to cones and not direct rod input to the horizontal cell itself.
These two studies generally agree on several aspects of the relationship of rod and cone signals in this early retinal pathway. Rod and cone responses combine with the same direction of influence on the peak response of the cone or horizontal cell and are independently adaptable. Cone responses are more transient, while rod responses contribute both to the initial peak response and to a more prolonged after-response (OFF response). Rod responses speed up and contribute more strongly to the transient peak as light level increases. The sensitivity of rod signals in cones and horizontal cells is only 1 to 2 log units greater than that of cone signals. Thus, this pathway seems well suited to transmit rod signals at mesopic light levels but is not sufficiently sensitive to mediate rod vision at low scotopic light levels (see also Chapter 17).
AII output to cone bipolars
As shown in Figure 55.1, the classical or primary pathway for rod photoreceptor signals is via rod bipolar cells and then the AII amacrine cells (Kolb, 1970), labeled RA in Figure 55.1. The outputs of these AII cells are mainly to ON (depolarizing) cone bipolar cells via gap junctions (Strettoi et al., 1994) and to OFF (hyperpolarizing) cone bipolars via conventional inhibitory synapses (Muller et al., 1988; Strettoi et al., 1992). (Thus, not only don't rod signals have private paths out of the retina, they don't even have private pathways to the retinal ganglion cells!) It is this AII pathway that appears to mediate rod vision at the lowest scotopic light levels. Rod signals in the AII pathway can be found over a 6 log unit range of stimulus intensities, including the lowest scotopic light levels (Buck et al., 1997c; Dacheux and Raviola, 1986). This means that the AII rod pathway is at least 3 log units more sensitive than the rod-cone gap-junction pathway (Verweij et al., 1999).
The interaction of rod and cone signals in the cone bipolar cells has not been studied systematically in primates. However, robust rod and cone signals can be recorded from AII amacrine cells in primates (Buck et al., 1997c; Dacey, 1996) and rabbit (Xin and Bloomfield, 1999). Both ON and OFF cone signals are found in AII cells and presumably come from cone bipolar cells. In primates, at least, sinusoidal stimulus modulation produces rod and cone responses in the AII of similar strength, but different latency, at higher light levels. These rod and cone signals can either add or cancel each other, depending on the frequency of stimulus modulation and the resulting phase relationship.
The accounts outlined above make the prediction that the cone bipolars should be the site of convergence of rod signals carried through two different retinal pathways, the rod-cone gap-junction pathway at high scotopic and mesopic light levels and the AII amacrine pathway over the entire range of rod response. This prediction has not been tested by intracellular recordings from cone bipolar cells in primates but is consistent with the results of human electroretinography (Stockman et al., 1995). In addition, Buck et al. (1997c) found that the temporal sensitivity of rod signals was sluggish at low scotopic levels (peak sensitivity at 2 to 3 Hz with a 7 to 9 Hz cutoff) but appeared to speed up at mesopic light levels (allowing rod signals to be in canceling counterphase with cone signals at 10 Hz). Whether these brisker rod signals came through the rod-cone gap-junction pathway or the AII pathway was not clear.
Of great interest will be further work to determine (1) the different properties imposed on rod signals that pass through these two distinct retinal pathways and (2) what conditions favor transmission of rod signals by each pathway. Existing evidence and theory on these questions can be found in Chapter 17. The rod-cone gap-junction pathway and the AII amacrine pathway are currently the retinal pathways of rod-cone interaction that are best documented and of clearest relevance for primate vision. Other pathways have been identified in other species, such as the mouse (Tsukamoto et al., 2001), and may one day be determined also to be important for retinal processing in primates.
Rod Influence on Retinal Output Pathways (Ganglion Cell Types)
Fundamental questions also remain about what types of retinal ganglions transmit rod signals out of the retina, under what conditions each type of ganglion cell does so, how rod and cone signals interact to determine jointly the output of each type of ganglion cell, and how the later stages of the visual system use this information to mediate specific aspects of visual function.
Unfortunately, the literature is varied on the role of rod signals in the three best-understood classes of genic-ulate-projecting retinal ganglion cells in primates: the parvocellular-projecting midget ganglion cells, the magnocellular-projecting parasol cells, and the koniocellular-projecting small bistratified cells.
What seems most certain is that there is strong rod input to parasol cells, with the rod input of the same sign as the cone input to the cell's receptive field center (e.g., Virsu and Lee, 1983; Virsu et al., 1987; Wiesel and Hubel, 1966). Gouras and Link (1966) reported that rod and cone inputs interact at mesopic levels in phasic ganglion cells (presumed parasol cells). They also reported that rod signals arrive with longer latency than cone signals, so that the response of these phasic cells becomes cone dominated at mesopic and photopic light levels. Lee et al. (1997) also reported a longer latency for rod signals compared to cone signals in physiologically identified parasol cells and rod domination of the response of dark-adapted cells at all light level below 20 photopic trolands. Lee and colleagues found that rod and cone signals interacted in temporal-frequency-dependent ways which were consistent with superposition or summation of signals arriving with different latency. Similarly, Enroth-Cugell et al. (1977) showed that, in cat retina, separately initiated rod and cone responses always summed, regardless of whether there was a latency difference.
Less certain is the role of rod signals in primate midget ganglion cells. Wiesel and Hubel (1966) reported finding rod input to 4 of 17 parvocellular lateral geniculate nucleus (LGN) cells (presumably reflecting inputs from midget ganglion cells). Rod input to these cells was always in the same direction of effect as the cone input to the receptive field center. Virsu and Lee (1983) and Virsu et al. (1987) also reported strong excitatory rod input to some parvocellular LGN cells, including most M-cone-center ON cells, and inhibitory rod inputs in some light-inhibited cells. L-cone-center cells rarely showed rod input, which was always weak.
In contrast, other studies have found little or no rod influence on the midget/parvocellular pathway. Purpura et al. (1988) found only a weak response of parvocellular-projecting (presumably midget) ganglion cells at mesopic and scotopic light levels. They concluded that primate pattern vision is mediated by magnocellular-projecting (presumably parasol) ganglion cells at scotopic levels: although midget cells may be able to convey information about coarse gratings (<0.6 c/deg), parasol cells display higher contrast gain even under these conditions at scotopic light levels. Most curiously, a sophisticated recent attempt to measure rod influence at scotopic and mesopic light levels in carefully identified macaque retinal ganglion cells found only weak rod input to midget ganglion cells (Lee et al., 1997). Overall, rod input was found in about 65% of midget cells at a scotopic light level of 2 trolands, but typically it never contributed more than a few spikes per second, even at maximum stimulus contrast. At a mesopic light level (20 trolands), rod influence may or may not have been present but was estimated to be less than 5% of the strength of the cone influence on these midget cells. In agreement with the general finding, rod influence, when found, was always in the same direction as the cone input to the center response of the ganglion cell. Lee et al. (1997) found no differences in frequency or strength of rod input between L-cone- or M-cone-center midget cells or between ON- or OFF-center midget cells, in contrast to the results of Virsu and Lee (1983) and Virsu et al. (1987).
We have the least information about the possibility of rod input to the small-bistratified ganglion cells and any other ganglion cells that also receive input from S cones. Virsu and Lee (1983) and Virsu et al. (1987) reported that most parvocellular LGN neurons with response patterns corresponding to B+Y-cells displayed a substantial rod excitatory input at about 2 trolands. D. M. Dacey (personal communication) has also found small-bistratified cells in the far periphery that display strong excitatory rod input at scotopic light levels. However, Lee et al. (1997) found no evidence of any rod input to any of 10 physiologically defined “blue-ON” cells that correspond to the small-bistratified ganglion cells (Dacey and Lee, 1994).
The uncertain picture of rod influence on midget and bistratified ganglion cells that emerges from the abovementioned studies stands in contrast to the considerable body of psychophysical evidence (surveyed later in this chapter) of rod influence on human color vision and the growing body of physiological and anatomical evidence that a mixture of rod and cone inputs is more the rule than the exception in earlier retinal pathways. At this point, one can only speculate that perhaps the variability among the ganglion cell studies is a result of the sensitivity of functional rod influence to adaptation states, light levels, and other stimulus variations that is evident in both the psychophysics surveyed below and the preganglion cell retinal physiology surveyed above.
| |
