Distinct patterns of corticogeniculate feedback to different layers of the lateral geniculate nucleus

In primates, feedforward visual pathways from retina to lateral geniculate nucleus (LGN) are segregated to different layers. These layers also receive strong reciprocal feedback pathways from cortex. The degree to which feedforward streams in primates are segregated from feedback streams remains unclear. Here, we asked whether corticogeniculate cells that innervate the magnocellular (M), parvocellular (P), and koniocellular (K) layers of the LGN in the prosimian primate bush baby (Otolemur garnettii) can be distinguished based on either the laminar distribution or morphological characteristics of their axons and synaptic contacts in LGN, or on their cell body position, size, and dendritic distribution in cortex. Corticogeniculate axons and synapses were labeled anterogradely with biotinylated dextran injections in layer 6 of cortex. Corticogeniculate cell bodies were first labeled with fluorescent dextran injections limited to individual M, P, or K LGN layers and then filled with biotinylated Lucifer yellow. Results showed that feedback to the M or P LGN layers arises from cells with dendrites primarily confined to cortical layer 6 and axons restricted to either M or P LGN layers, but not both. Feedback to K LGN layers arises from cells: 1) whose dendrites distribute rather evenly across cortical layers 5 and 6; 2) whose dendrites always extend into layer 4; and 3) whose axons are never confined to K layers but always overlap with either P or M layers. Corticogeniculate axons also showed distributions that were retinotopically precise based on known receptive field sizes of layer 6 cells, and these axons mainly made synapses with glutamatergic projection neurons in the LGN in all layers. Taken together with prior physiological results, we argue that the morphological differences between the three corticogeniculate pathways show that the M and P feedback pathways could rapidly and specifically enhance local LGN activity, while we speculate that the K feedback pathway is organized to temporally synchronize activity between LGN and cortex.