Functional connectivity between retina and LGN: anatomical considerations

Despite the central importance of physiological experiments in the study of functional connectivity, anatomy is of course necessary to put physiological studies in context. The anatomy of a feedforward pathway can be studied at multiple levels of detail. At the broadest level, anatomical techniques can simply demonstrate the existence of a projection pathway, as has been known since classical times for the retinogeniculate pathway (Wade, 1998). Next, slightly finer techniques can be used to trace the projections of smaller groups of neurons, as was done in early studies that demonstrated the retinotopic mapping in the LGN and visual cortex (Le Gros Clark and Penman, 1934; Polyak, 1933). At the next level, single-cell labeling techniques (Golgi, horseradish peroxidase, biocytin) can be used to elucidate the structures of dendritic trees or of axonal arbors. Finally, individual synapses between neurons can be identified either with electron microscopy (Famiglietti and Peters, 1972; Guillery, 1969; Hamos et al., 1987; Mason et al., 1984) or, increasingly, with multiple-label confocal microscopy (e.g., Mills and Massey, 1999).

Although it would be difficult to use anatomy to study functional connectivity, as we have defined it, in the retinogeniculate pathway, our current knowledge of anatomy allows us to approximate the range of convergent and divergent connections that might be made. Specifically, anatomy allows us to define an upper limit of the number of ganglion cell inputs to an LGN neuron or the number of LGN neurons receiving input from a single ganglion cell. These numbers can be derived in several different ways.

In visual neuroscience, the coverage factor has generally been defined as the number of neurons of a certain class whose receptive fields overlap a given point in visual space. A second use of the term is in the context of anatomy: the number of axons or dendrites of a given class that overlap in a given position within a neural structure. For retinal ganglion cells, anatomical and physiological coverage factors are closely related, because receptive-field centers are roughly equal in area to a ganglion cell's dendritic arbor (Wässle et al., 1983; see DeVries and Baylor, 1997).

The anatomical coverage factor is derived by first counting the number of retinal ganglion cells of a given class in 1 mm² of retina and then determining the average area occupied by each dendritic arbor. The product of these values yields the total dendritic area of all ganglion cells per square millimeter of retina, which is equivalent to the average number of dendritic arbors that overlap any point. The anatomical coverage factor for β cells in the cat retina (which correspond to X cells) is roughly 3 (Wässle et al., 1981).

Although the retinal coverage factor will in fact prove important for our discussion of the pathway between LGN and cortex, coverage factors in the LGN are more pertinent to our discussion of the retinogeniculate pathway. Although less has been written specifically about geniculate coverage factors (both by ganglion cell arbors and by geniculate dendritic trees), they can be calculated from values in the literature.

The total surface area of the A laminae in the LGN (A plus A1) is roughly 80 mm² (10 by 4 mm × 2; Sanderson, 1971), and there are approximately 100,000 X cells in the retina (reviewed in Peters and Payne, 1993); therefore, there are roughly 1250 retinal arbors per square millimeter in each lamina. The axonal arbor of each ganglion cell terminates throughout the thickness of its appropriate layer, but the cross-section is roughly circular, with a diameter of 0.1 mm (area = 0.008 mm²) (Sur et al., 1984, 1987). The coverage factor of retinal axonal arbors in the LGN is therefore 1250 × .008 = 10. The dendritic trees of X-cell relay neurons in the LGN have the same dimensions as the retinal axonal arbors, but because there are roughly 2.5 more relay neurons in the X-cell system than retinal afferents (Peters and Payne, 1993), the coverage factor is 2.5 times greater. In other words, a single dendritic spine of an X cell in the LGN is within the terminal arbors of ∼10 retinal ganglion X cells, while a single presynaptic bouton is within the dendritic trees of ∼25 relay neurons. Therefore, simply from the morphology of single neurons, the total area of the LGN, and the number of neurons, we have bounded the possible values of divergent and convergent inputs in the retinogeniculate system. Each LGN neuron could receive convergent input from more than 10 retinal neurons (because the dendritic arbor is not a single point). Each retinal neuron could diverge to form synapses onto at least 25 relay neurons. Although these morphological features would allow each relay neuron to receive inputs from many different afferents, only a few specific connections are made onto each cell. This specificity is seen physiologically (Cleland et al., 1971a, 1971b; Lee et al., 1977; Usrey et al., 1999; see below), and has also been demonstrated in a heroic ultrastructural study (Hamos et al., 1987) that found that a given retinal afferent makes synapses on fewer than 10% of potential targets in the LGN.