Notes on Visual Cortical Feedback and Feedforward Connections

摘要

Introduction Feedforward (FFD) – feedback (FBK) cortical processing ultimately needs to be considered in the context of whole-brain activation, including interactions with cortico-thalamo-cortical, callosal, and the excitatory and inhibitory intrinsic cortical circuits. For the non-human primate (NHP) brain, however, identifying cell types and deciphering the patterns and metrics of axon convergence and divergence is challenging (cf. Rockland, 2019 , 2020 ) and, at the level of detail approachable in the mouse brain, may still be years away. Many of the comments put forth here are not novel and echo previous reports (including my own, Rockland, 1997 ). My goal has been to briefly re-consider what have become key features of FFD-FFK connections in the early visual cortical pathway, with emphasis on the cellular and dendritic circuitry components. Owing to sparsity of data in NHP concerning the role of interneuron subpopulations in microcircuitry, these are not discussed. For detailed reports on visual cortical connectivity and physiological response properties (see Bullier, 2004 ; Douglas and Martin, 2007 ; Shipp, 2007 , 2016 ; Markov et al., 2014a , b ; Angelucci et al., 2017 ; Vanni et al., 2020 ; Vezoli et al., 2021 , among others). Although area V1 is a canonical “start point” for discussing FFD-FBK cortical processes, it is actually something of an outlier; that is, there are cortico-thalamic, but not cortico-cortical FBK projecting neurons in V1; and FFD terminations are of thalamic, but not cortical origin. There are few or no callosal connections. Thus, a strict comparison of cortical FFB and FBK connections is better addressed in extrastriate areas V2, V3, V4, MT, or TEO. Much of the following discussion is written as applying to V2. Neurons of Origin As repeatedly summarized in the literature (e.g., Kennedy and Bullier, 1985 : Rockland, 1997 , 2019 ; Douglas and Martin, 2007 ; Markov et al., 2014a , b ; Anderson and Martin, 2016 ; Angelucci et al., 2017 , among others), FFD projecting neurons from V2 (to V4 and MT) and FBK projecting neurons (to V1) are differentially located in deeper layer 3 (FFD) or layers 2, 3A, 5, and 6 (FBK). The FFD-FBK laminar dissociation, despite a minor degree of laminar intermingling, has been largely confirmed by injections of two distinguishable retrograde tracers in V1 and V4 ( Markov et al., 2014b ; and see Figure 1), where less than 1% of cortically projecting cells in V2 (and 2.2% in V3) were double labeled (i.e., had branching collaterals to both V1 and V4). The further characteristics of these bifurcating, link neurons, and their postsynaptic targets, are unknown. Are they more frequent in the less investigated peripheral visual representation of V2, or for other combinations (e.g., injections in V1 and MT)? The most numerous FBK population is in layer 6 (see estimates in Table 3 in Rockland, 1997 ; Markov et al., 2014a , b ). Along with the smaller number of layer 5 FBK neurons, this infragranular distribution overlaps with that of several cortico-subcortical projecting populations (cortico-collicular, cortico-striatal, or cortico-thalamic projecting neurons in layer 5, and cortico-thalamic or cortico-claustral neurons in layer 6; summarized in Shipp, 2007 ). Appropriate double retrograde tracer experiments have not been done to probe for collateralization of cortico-cortical and cortico-subcortical axons. Whether these neuronal subpopulations are spatially clustered or distributed in a salt-and-pepper pattern has not been established (but see Hawken et al., 2020 for “functional clusters” in V1). Neuronal Subtypes Feedback and feedforward neurons are excitatory pyramidal neurons, although a small number of GABAergic FBK neurons, probably positive for nitric oxide or somatostatin, are found in the supragranular layers of V2 after viral infection in V1 ( Tomioka and Rockland, 2007 ). Pyramidal subtypes can be more finely distinguished, in part by dendritic morphology. Supragranular neurons extend their apical dendrite into layer 1. For layer 6 neurons and many layer 5 FBK neurons, the apical dendrite extends only into layer 3. A subset of layer 5 neurons send apical dendrites to layer 1 (Golgi stains: Lund et al., 1981 ); and intracellular fills of tracer identified FFD projection neurons demonstrated about half (4 of 9 neurons) having apical dendrites that extend to layer 1 ( Markov et al., 2014b ). Soma depth is significant, in that shorter apical dendrites, even of neurons in the same layer, are reported to be less excitable (in mice: Galloni et al., 2020 ). By comparison, five subtypes of morphologically distinct cortico-geniculate (CG) neurons in V1 and at least three subtypes in V2 have been identified ( Briggs et al., 2016 ). Heterogeneity of CG neurons is supportive of some degree of parallel processing (“…not one circuit, but rather a collection of distinct circuits conveying unique [visual] feature information and operating on a corresponding variety o