Parallel Processing is a commonly used strategy in sensory systems of the mammalian brain. In the primate visual system, information is relayed from the retina to primary visual cortex (V1) along three parallel pathways: magnocellular (M), parvocellular (P), and koniocellular (K). These three pathways remain anatomically and physiologically distinct as they pass through M, P, and K layers of the lateral geniculate nucleus (LGN) of the thalamus and into V1, with the M pathway terminating primarily in layer 4C[alpha] the P pathway in layer 4C[beta], and the K pathway in the cytochrome oxidase (CO) blobs of layer 2/3. Beyond V1, visual information is processed in relatively independent dorsal and ventral streams specialized for computations related to spatial vision and object recognition respectively. Understanding the relationship between early parallel pathways and dorsal and ventral cortical processing streams has proven difficult because of the substantial convergence of M, P, and K pathways outside of layer 4C of V1. We used rabies virus as both a mono- and trans-synaptic retrograde tracer to determine the contributions of M and P pathways to dorsal stream cortical area MT in macaque monkey. MT is specialized for motion and depth processing and is thought to be dominated by the M pathway, with little or no contribution from the P or K pathways. We first injected rabies virus into MT with a 3 day survival time, allowing virus to cross one synapse and infect cells disynaptic to the injection site. We found large numbers of parvalbumin- positive, calbindin-negative neurons retrogradely labeled in the M and P layers of the LGN, providing evidence for a disynaptic M and P input to MT, likely mediated by layer 6 Meynert cells in V1. We next analyzed V1 after the very same injections into MT and found disynaptic label in layer 4C to be confined almost exclusively to M-dominated layer 4C[alpha]. This indicated that the most direct ascending input through layer 4C of V1 to MT is dominated by the M pathway. In order to determine if MT receives indirect P input through layer 4C of V1, we made injections into MT once again, this time with a 6 day survival time, allowing the virus to spread up to four synapses past the cells that project directly to MT. In this case, we found transynaptic label throughout all layers of V1, including P-dominated layer 4C[beta], indicating that MT does receive indirect P inputs through layer 4C of V1. In order to determine the likely relay of this P input to MT, we made 3 day survival injections of virus into V3 and V2, areas that provide indirect inputs from layer 4B of V1 to MT. Only after certain injections into V2, but not V3, did we observe disynaptic label in both M and P sublayers of 4C. These results suggest that while the most direct ascending input through layer 4C of V1 to MT is dominated by the M pathway, within a few more synapses the P pathway contributes as well, likely through the CO thick stripes of V2. In a final set of experiments, we used a modified rabies virus that expresses green fluorescent protein and doesn't cross synapses to compare the detailed morphology of neurons projecting directly from V1 to MT or V2. We found that cells projecting from layer 4B of V1 to MT were a majority spiny stellate and those projecting to V2 were overwhelmingly pyramidal. Additionally, MT-projecting cells had larger cell bodies, more total dendritic length, and were located deeper within layer 4B. Finally, pyramidal cells projecting to MT were located preferentially underneath CO blobs, where they could receive strong M inputs onto their apical dendrites. These results suggest that specialized and distinct cell populations in layer 4B of V1 mediate an M- dominated signal to MT and a mixed M and P signal to V2. All together, these studies provide strong evidence for the existence of multiple circuits between LGN and MT. Each pathway receives a different combination of M and P inputs and is uniquely suited to convey visual information with varying degrees of spatial and temporal resolution, contrast sensitivity, and color selectivity. Distinct cell types underlie many of these circuits, overcoming the lack of spatial compartmentalization of V1 outputs. Functional studies that can target these specialized cell types and circuits will be necessary to elucidate the contributions of each pathway to response properties in MT and, ultimately, to visual perception and behavior
