Cytoarchitectural differences are a key determinant of laminar projection origins in the visual cortex

Abstract

Regularity of laminar origin and termination of projections appears to be a common feature of corticocortical connections. We tested three models of this regularity, originally formulated for primate cerebral cortex, using quantitative data on the relative supragranular layer origins (SGN%) of 151 projections from 19 areas (∼ 145,000 neurons) to four areas of cat extrastriate cortex. Predictive variables in the models were: hierarchical level differences (Barone et al., 2000), structural type differences (Barbas, 1986), and distances (Salin and Bullier, 1995) between areas. Global and local hierarchies of cat visual cortex were used to evaluate the hierarchical model. Ranking of areas by their cytoarchitectural differentiation (e.g., relative prominence of layer IV) allowed testing of the structural model, while the distance model was tested for the number of borders separating areas. Laminar projection origins correlated moderately with hierarchical differences, and poorly with border distances, but were strongly and consistently correlated with area differences in cytoarchitectural rank. Moreover, projection densities were moderately and negatively correlated with area distances and structural differences. Our findings suggest that the relative cytoarchitectural differentiation of cortical areas is the main determinant of laminar projection origins in cat visual cortex, and may underlie a general laminar regularity of mammalian cortical connections.

Introduction

Laminar origins and terminations of connections between areas of the mammalian cerebral cortex show a remarkable regularity. This regularity is most clearly demonstrated by the fact that the areas of sensory (e.g., visual, somatosensory) and motor cortex can been arranged into largely consistent global schemes, in which laminar connectivity patterns believed to signify projections in one of three possible directions (“forward,” “backward” or “lateral”) are repeated across the whole arrangement (Felleman and Van Essen, 1991, Hilgetag et al., 2000b, Scannell et al., 1995). The classification of “forward,” “backward,” and “lateral” projections is based on the predominant layers of projection origin and termination. Projections have been labeled “forward” if they originate mainly from neurons in upper cortical layers (II + III) and terminate in the granular layer (IV) of their target, while complementary “backward” projections arise mainly from deep cortical layers (V + VI) and terminate outside layer IV of the target area (Rockland and Pandya, 1979). Finally, “lateral” projections originate in a more bilaminar pattern involving roughly equal contributions from upper and deep layers and terminate across all layers, including layer IV (Felleman and Van Essen, 1991). While this classification was originally applied to connections of primate visual areas 17, 18 and 19 (Rockland and Pandya, 1979), later approaches extended the scheme to additional extrastriate cortices and other types of primate sensory cortex (Felleman and Van Essen, 1991), as well as visual cortex in the cat (Felleman and Van Essen, 1991, Scannell et al., 1995) and rat (Coogan and Burkhalter, 1993).

Quantitative studies have demonstrated that there are also gradual shifts in the laminar projection origins within each of these categories. For example, the relative contribution of upper (supragranular) layers to a “forward” projection may vary from 100% to little more than 50% (Barone et al., 2000, Grant and Hilgetag, 2005, Vezoli et al., 2004) with the remainder coming from the deep layers. Moreover, there exists considerable variability of projection patterns at the level of individual neurons (Rockland, 1997, Rockland, 2003).

What explains the systematic but varied patterns of laminar projection origins? Three principal models have been proposed. First, the graded distribution may depend on the hierarchical level difference between the linked areas in global cortical hierarchies (Barone et al., 2000, Vezoli et al., 2004). Such hierarchical schemes (Felleman and Van Essen, 1991) are popular summaries of the organization of cortical association fibers and their putative function, despite caveats for their construction (Hilgetag et al., 1996) and interpretation (Hegde and Felleman, 2007). In the hierarchical model (Barone et al., 2000) it is argued that projections with a strong laminar bias of origin (i.e., close to 100% upper or deep layer) occur between areas occupying the top and bottom levels of a global hierarchy, while projections of more equal laminar origin (i.e., approximately 50% bilaminar) occur between areas on the same or adjacent hierarchical levels.

In the second, structural model it is proposed that graded laminar patterns derive from differences in the cytoarchitectonic differentiation of the cortical areas forming the source and target of a given projection (Barbas, 1986, Barbas and Rempel-Clower, 1997, Medalla and Barbas, 2006, Rempel-Clower and Barbas, 2000). Specifically, projections from structurally more differentiated to less differentiated areas arise predominantly from the upper cortical layers, while projections in the reverse direction originate mainly from deep layers, and projections between architecturally similar areas possess a balanced bilaminar character (Barbas, 1986).

Finally, it has been an influential idea that the existence or absence of inter-connections between cortical areas (Klyachko and Stevens, 2003, Young, 1992) as well as laminar projection features (Salin and Bullier, 1995) vary systematically with physical distance and the number of intervening area borders.

So far, these different models have been primarily tested for connections of the visual cortex (hierarchical model) or prefrontal cortex (structural model) or both (distance model) in the primate brain. Here an opportunity arose to test all three models on the same quantitative material concerning both the relative densities (i.e., strength) and laminar origins of widespread projections to cat extrastriate visual areas (Grant and Hilgetag, 2005, Hilgetag and Grant, 2000). This allowed us to directly evaluate the selective applicability of these concepts for graded laminar connections in another class of mammal with a well developed cerebral cortex.