Large-scale network organization in the avian forebrain: a connectivity matrix and theoretical analysis

Abstract

Many species of birds, including pigeons, possess demonstrable cognitive capacities, and some are capable of cognitive feats matching those of apes. Since mammalian cortex is laminar while the avian telencephalon is nucleated, it is natural to ask whether the brains of these two cognitively capable taxa, despite their apparent anatomical dissimilarities, might exhibit common principles of organization on some level. Complementing recent investigations of macro-scale brain connectivity in mammals, including humans and macaques, we here present the first large-scale "wiring diagram" for the forebrain of a bird. Using graph theory, we show that the pigeon telencephalon is organized along similar lines to that of a mammal. Both are modular, small-world networks with a connective core of hub nodes that includes prefrontal-like and hippocampal structures. These hub nodes are, topologically speaking, the most central regions of the pigeon's brain, as well as being the most richly connected, implying a crucial role in information flow. Overall, our analysis suggests that indeed, despite the absence of cortical layers and close to 300 million years of separate evolution, the connectivity of the avian brain conforms to the same organizational principles as the mammalian brain.

Keywords: avain neuroanatomy; brain connectivity; brain network analysis; comparative neuroanatomy; pigeon forebrain.

Figure 1

Transverse sections through the pigeon telencephalon showing the locations of each of the 52 regions used in the study. See Table 1 for abbreviations. Regions are colored according to their module and sub-module membership (see also Figure 4). Color codes: red, associative; blue, cortico-hippocampal; green, visual; brown, viscero-limbic; yellow, auditory. Regions colored white are excluded from the study. While the connections of these white regions have been explored, they have not been systematically clarified nor unequivocally confirmed. Black areas, such as the one labeled “V” at A14.00, are ventricles.

Figure 2

Connections in the pigeon telencephalon. A green cell in row i, column j indicates that a connection exists from region i to region j. Top-level modules are outlined in yellow. Sub-modules are outlined in magenta. See Table 1 for abbreviations.

Figure 3

The seven structural motifs that occur with highest z-scores. Circles denote nodes (brain regions) and arrows denote directed arcs (connections).

Figure 4

The telencephalic connectome of the pigeon forebrain. Network analysis reveals five top-level modules. The associative and cortico-hippocampal modules can be further decomposed. Connections to and from hub nodes are shown in a slightly darker color. See Table 1 for abbreviations.

Figure 5

Pathways of the pigeon forebrain in anatomical co-ordinates (sagittal view). Nodes are colored according to top-level module membership. Note that the modules are spatially distributed rather than localized. See Table 1 for abbreviations. See also Figure 1 for color codes: red, associative; blue, cortico-hippocampal; green, visual; brown, viscero-limbic; yellow, auditory.

Figure 6

Pathways of the pigeon forebrain in anatomical co-ordinates (horizontal view). Nodes are colored according to top-level module membership (Figure 4). Note the spatial distribution of modules.

Figure 7

Sub-shells of the innermost k-core following k-core decomposition. The innermost k-core (i = 10) contains almost half the nodes in the network, but its sub-shell structure reveals a finer level of organization. All five hub nodes (shown in bold) appear in the innermost two sub-shells.

Figure 8

The results of rich club analysis. Nodes are ranked according to their total degree. The rich club coefficient for rank k is the proportion of possible connections between nodes of rank k or higher that are actual connections. This measure is then normalized with respect to the average for an equivalent random network. The three nodes at the rightmost end of the plot (AI, AD, and NCL) are designated a rich club, because their normalized rich club coefficients all lie above the random network average.