A Connectomic Hypothesis for the Hominization of the Brain

Abstract

Cognitive abilities of the human brain, including language, have expanded dramatically in the course of our recent evolution from nonhuman primates, despite only minor apparent changes at the gene level. The hypothesis we propose for this paradox relies upon fundamental features of human brain connectivity, which contribute to a characteristic anatomical, functional, and computational neural phenotype, offering a parsimonious framework for connectomic changes taking place upon the human-specific evolution of the genome. Many human connectomic features might be accounted for by substantially increased brain size within the global neural architecture of the primate brain, resulting in a larger number of neurons and areas and the sparsification, increased modularity, and laminar differentiation of cortical connections. The combination of these features with the developmental expansion of upper cortical layers, prolonged postnatal brain development, and multiplied nongenetic interactions with the physical, social, and cultural environment gives rise to categorically human-specific cognitive abilities including the recursivity of language. Thus, a small set of genetic regulatory events affecting quantitative gene expression may plausibly account for the origins of human brain connectivity and cognition.

Keywords: brain hominization; brain phenotype; connectomic fundamentals; human genome.

Figure 1

Enlargement of the brain leads to a reconfiguration of brain wiring with a relative decrease of connection density and increasing modularity. In relatively small mammalian brains, such as that of the macaque monkey, connectivity between neurons or brain areas is denser than in larger brains, such as the human brain. This is due to the fact that the average number of synaptic connections per neurons stays largely constant across mammalian brains (left bottom), rather than scaling up with the number of neurons in the network (left top), which would result in an impossible increase in white matter (Striedter 2005). Moreover, as shown on the right, when transitioning from smaller to larger brains, there is a shrinkage of the “horizon of connectional possibilities,” defined by the distance between neurons (d1 and d2), leading to increased network sparsity due to a receding horizon of connectional opportunities between neurons, thus offering a parsimonious wiring constraint merely imposed by brain size changes. Note that the brain network is characterized by modules (blue circle), that is, sets of areas/neurons that are more connected in-between them when compared with the rest of the network.

Figure 2

Modularity and working memory capacity. (A) Schematic depiction of the hierarchical modular organization of the brain’s connectome (Hilgetag and Hütt 2014). (B) Networks with high or low modularity forming the “reservoir” of an artificial neuronal “echo state” network (Rodriguez et al. 2019). (C) Functional consequences of the modular architecture of the network. The network was tested for working memory-like capacities, that is, the duration that the network could retain a sequence as well as the number of sequences that could be recalled. Note that a network configuration situated very close to high modularity exhibits the highest performance, memorizing a larger amount of sequences and retaining such memory for prolonged durations (Rodriguez et al. 2019).

Figure 3

Evolution of the multilevel connectional architecture for neural representations in biological and artificial brain networks. (A) Multilevel artificial neural network architecture (left). A synthetic agent with a multilevel visual system can navigate a natural environment. Activity in the artificial neural network in higher levels allows a more accurate reconstruction of the location of the artificial agent (right). Note the decrease of the error of location reconstruction with increased level. Thus, a serial, convergent processing of activity from the sensorium to higher levels of the network enables abstract representations. (B) Enlargement of the brain and expansion of the association cortex can lead to the overall sparsification of the network (cf. Fig. 1), and, in addition, to an expanded multilevel structure of the human brain. The increased number of levels, or processing stages, defined as synaptic steps between neurons, is due to the expansion of the association cortex in humans in relation to monkeys and presumably other primates. In humans, sensory areas drift apart in physical space and, thus, do not directly connect with each other, but integrate information through a multilevel connectomic architecture toward the network core. The presence of more hierarchical levels may bestow the human brain with increased capacity for more refined and abstract representations of the sensorium. (A) Modified from Wyss et al. (2006). Brains in (B) from Krubitzer and Seelke (2012). Modality: 1 = somatosensory; 2 = auditory; 3 = visual.

Figure 4

The GNW and Core-Periphery network architecture. (A) GNW model. The model postulates that the brain possesses a central connectional and functional component, the global workspace, composed of distributed and heavily interconnected neurons with long-range axons, in which the conscious integration of peripheral sensory input, such as visual, and emotional content takes place, giving rise to the “ignition” process (modified from Dehaene and Changeux 2011). (B) Network architecture of the macaque monkey cortex with a tightly interconnected and central network component (“network core”), encompassing association areas, and a less central “periphery” part of the network, encompassing mostly sensorimotor areas. Thus, the network core can be conceived as the connectomic backbone of the global workspace (adopted from Markov et al. 2013). (C) Situating the core–periphery network architecture within the cytoarchitectonic gradients of the cortex. A species-specific relation to the gradients of microstructural features of the cortex is observed. In progressively larger brains, core areas differ from periphery areas in terms of their cytoarchitecture, with the more topologically central core areas encompassing association areas with less laminar differentiation compared with the periphery areas, which encompass primarily sensorimotor areas with a high degree of laminar differentiation.

Figure 5

Laminar-wise “reafference shift” from monkey to human. (A) Laminar origin of connections is related to the cytology of the areas, specifically to soma size of the projection neurons in upper and deep layers (Goulas et al. 2018). Areas with connections emanating predominantly from deep layers (e.g., rostral temporal pole) tend to host projection neurons with larger soma size in deep compared with upper layers (interno-pyramidal areas). Areas with connections emanating predominantly from upper layers (such as peripheral visual areas) tend to host projection neurons with larger soma size in upper compared with deeper layers (externo-pyramidal areas). Areas with a more laminar-balanced origin of connections (e.g., frontal pole) also exhibit a more balanced soma size of projection neurons in upper and deeper layers (equipyramidal areas). (B) Qualitative observations indicate that the human cerebral cortex, relative to the monkey cortex, and presumably to other primates, exhibits a higher proportion of externo-pyramidal to interno-pyramidal areas (Sanides 1962, 1970; Sanides and Krishnamurti 1967). Due to the relation of cytology and laminar origin of connections, such cytological changes may denote a shift of the origin of long-range connections to upper layers in the human brain. Drawings modified from Goulas et al. (2018).

Figure 6

Development of dendrites and soma size of layer IIIc projection neurons and its plausible contribution to higher cognitive functions such as theory of mind and dendrophilia including language recursivity. (A) Model of epigenesis by selective stabilization of synapses. A nesting of many such elementary steps occurs in the course of development resulting in a hierarchical foliation of the growing networks. For a given set of developing neurons (e.g., thalamocortical or neuromuscular junction), the growing axon terminals branch exuberantly at first. But then, depending on the state of activity of the target neuron—both intrinsic spontaneous firing and evoked by external inputs—some synapses are eliminated (pruned), while others are strengthened and stabilized. In postnatal life, an important part of the activity in the network results from inputs from the environment and so the epigenetic selection of synapses represents an internalization of the outside world. (B) Total synapse density during the development of the monkey and human brain cortex (region V1). Note the extended time window in humans where multiple waves of synapse selection take place. Also note the sharper decrease of total synaptic density in humans before puberty, reflecting the more prominent elimination than formation of synapses. (C) Development of soma size of projection neurons in upper (layer IIIc) and deep layers (layer V) of the prefrontal cortex (Petanjek et al. 2008, 2019). Note that the soma size of layer IIIc neurons increases rapidly and matches or exceeds the soma size of layer V. (D) Dendrites of layer IIIc projection neurons have 2 phases of development. The first phase occurs perinatally, during approximately the initial 2.5 months. This initial phase is succeeded by a dormant phase. However, after the dormant phase, a second growth spurt takes place approximately at 2.5 years. Importantly, the second growth spurt characterizes upper layer (layer IIIc) projection neurons and not deep layer (layer V) projection neurons (Petanjek et al. 2008, 2019). This is approximately the age where cognitive skills like theory of mind and language recursivity start to develop (approximate span: 2–5 years), and thus, the described developmental epigenetic processes might contribute to the neurobiological basis of the “cultural brain.” (A) Adapted from Changeux et al. (1973). (B) Adapted from Bourgeois (1997). (C) and (D) adapted from Petanjek et al. (2019).