An integrative understanding of comparative cognition: lessons from human brain evolution

Abstract

A comprehensive understanding of animal cognition requires the integration of studies on behavior, electrophysiology, neuroanatomy, development, and genomics. Although studies of comparative cognition are receiving increasing attention from organismal biologists, most current studies focus on the comparison of behaviors and anatomical structures to understand their adaptative values. However, to understand the most potentially complex cognitive program of the human brain a greater synthesis of a multitude of disciplines is needed. In this review, we start with extensive neuroanatomic comparisons between humans and other primates. One likely specialization of the human brain is the expansion of neocortex, especially in regions for high-order cognition (e.g., prefrontal cortex). We then discuss how such an expansion can be linked to heterochrony of the brain developmental program, resulting in a greater number of neurons and enhanced computational capacity. Furthermore, alteration of gene expression in the human brain has been associated with positive selection in DNA sequences of gene regulatory regions. These results not only imply that genes associated with brain development are a major factor in the evolution of cognition, but also that high-quality whole-genome sequencing and gene manipulation techniques are needed for an integrative and functional understanding of comparative cognition in non-model organisms.

Introduction

A comprehensive understanding of animal cognition requires the integration of knowledge from Tinbergen’s four questions: mechanism, adaptation, ontogeny, and phylogeny of behaviors (Tinbergen 1963). Evolution not only shapes different body traits of animals for biodiversity but also gives rise to specialized cognition to cope with complex physical and/or social environments. Through comparison with closely related species, researchers have suggested adaptive values of cognitive abilities in a broad spectrum of species. For example, exceptional odor discrimination of burrowing rodents and moles (Lavenex and Schenk 1998; Catania 2013), sophisticated flexibility of learning to adapt to fast-changing environments (Day et al. 1999; Liu et al. 2016; Logan 2016), outstanding spatial memory in species with high-levels of spatial demands (Shettleworth et al. 1990; Bednekoff and Balda 1996; Pašukonis et al. 2018; Liu et al. 2019), impressive problem-solving abilities in species adapted to cope with novel environments (Heinrich and Bugnyar 2005; Sol et al. 2005; Leal and Powell 2012), and even sex differences of cognition within species (Keagy et al. 2012; Guigueno et al. 2014; Liu and Burmeister 2017; Ventura et al. 2019). However, very few numbers of species and cognitive traits have been studied to address all of Tinbergen’s four questions. Undoubtedly though, human cognition and the brain is one of the most comprehensively studied subjects. Therefore, here, we review recent progress in the study of the human brain and attempt to understand human cognition in an integrative way.

As one of 8.7 million eukaryotic species on planet earth, only humans (Homo sapiens) can dwell in any habitat on this planet (e.g., hot tropical desert, cold Antarctica, underground, and underwater) and even outer space (e.g., on a space station in near-earth orbit and our moon). All of these unique feats can be attributed to our superior cognitive ability to understand and utilize natural resources. Indubitably, human intelligence is a product of the sophisticated human brain. One effective way to understand human specialization is to do a comparison with genetically closely related species, especially the chimpanzee.

In order to understand the genomic and developmental mechanisms underlying human brain function, diseases, and uniqueness, several large-scale endeavors were launched including the Human Genome Project (Venter et al. 2001), Chimpanzee Genome Project (Waterson et al. 2005), ENCODE (Dunham et al. 2012), PsychENCODE (Akbarian et al. 2015), and Roadmap Epigenomics Project (Bernstein et al. 2010) to name a few. These projects have studied human brain specializations at different layers. In the first layer, genome sequence and its epigenetic features sit at the base of brain specialization. It ultimately contributes to the variety, characteristics, and expression levels of proteins. In the second layer, the transcriptome which is controlled by genome properties has been documented to vary across different brain developmental stages. Any changes in a given gene expression profile could result in different fates (e.g., proliferation or differentiation) of cells. Finally, the neuroanatomical structure of the brain in the third layer could be radically changed. The effects of mechanisms of human brain evolution on each layer have been extensively reviewed: genomic (Vallender et al. 2008; Konopka and Geschwind 2010; Florio et al. 2017), epigenomic and transcriptomic (Preuss et al. 2004; Somel et al. 2013; Lein et al. 2017), developmental (Kriegstein et al. 2006; Geschwind and Rakic 2013; Florio and Huttner 2014), and neuroanatomic (Sherwood et al. 2012; Sousa et al. 2017a), thus we will focus our attention on the connections of mechanisms across these different layers rather than details of the mechanisms themselves. In this review, we will start with human neuroanatomical specializations, discuss the developmental mechanisms underlying these anatomic traits, correlate gene expression differences to developmental trajectories, and finally review the genomic alterations that affect changes in gene expression and protein function.

Neuroanatomy

The human brain, which is composed of 86 billion neurons with an almost equal number of glial cells, has become highly specialized in human evolution (Herculano-Houzel et al. 2015; Silbereis et al. 2016; Sousa et al. 2017a). One prominent feature of the human brain is its large size and mass, which is roughly ×3.5 larger than that of a chimpanzee’s and ×2 larger than pre-human hominid brains (Wood and Collard 1999; Carroll 2003). However, the human brain is not the largest on the Earth. After splitting from the common ancestor of mammals about 220 million years ago, brain size enlarged independently in different mammalian lineages (Borrell and Reillo 2012; O’Leary et al. 2013). Compared with human (1.3 kg), humpback whales (6.1 kg), African bush elephants (5.4 kg), and even common bottlenose dolphins (1.6 kg) all possess larger brains (Boddy et al. 2012). However, given that brain size is positively correlated with body size, encephalization quotient, which is calculated by using actual brain size divided by expected brain size with allometric justification, is always applied in brain comparisons (Striedter 2005; Jerison 2012). Perhaps not surprisingly, encephalization quotient is highest in humans (5.72), compared with chimpanzees (1.72), humpback whales (0.29), African bush elephants (1.09), and common bottlenose dolphins (3.51) (Boddy et al. 2012). These data suggest that selection pressure disproportionately acted on the brain compared with the rest of the body in human evolution.

The neocortex, a six-layered structure which covers the top of the mammalian brain, is a hallmark of the evolution of the mammalian brain and is associated with high order cognitive ability (Striedter 2005; Rakic 2009). Neocorticalization, which refers to the disproportionate expansion of the neocortex compared with other brain sections, also occurred independently within different clades (Striedter 2005; Lewitus et al. 2013). Compared with mammals in other clades (e.g., insectivores, carnivores, and rodents), neocorticalization, which is defined as neocortical mass divided by brain mass, is the most pronounced in primates, with the human data point having the greatest positive deviation from the expected value which is based on an allometric best-fit line in primates (Finlay and Darlington 1995; Barton and Harvey 2000; Striedter 2005). As a result, a prominent trait of primates—to compromise the tradeoff between cortical size expansion and cranial volume limitation—is a gyrencephalic brain, compared with the lissencephalic brains of rodents (Kelava et al. 2013; Zilles et al. 2013). These observations suggest that the evolution of the mammalian neocortex follows different morphometric rules to generate a specialized primate neocortex. Such morphometric rules could be linked to a developmental trajectory which will be discussed later.

If the overall expansion of the neocortex in humans can be thought of as an extension of the morphometric rule of primate brains, then region-specific expansion within the human neocortex is a human-specialized trait which occurred within the most recent 200,000 years during the emergence of human-specific cognition and culture (McDougall et al. 2005; Chudek and Henrich 2011; Buckner and Krienen 2013). Neuroanatomists have categorized major brain regions into two groups with respect to their connectivity and cognitive functions. One group, called the primary regions, includes the visual, somatosensory, and motor cortex and mainly controls basic sensory and motor aspects of cognition. The other group is called the association regions, and includes the prefrontal cortex and parts of the temporal and parietal cortex. These regions are associated with higher-order cognitive function, for example, learning and memory and decision making (Buckner and Krienen 2013; Raichle 2015). Compared with other primates, the cortical association regions of the human brain have undergone a greater expansion of areal size after scaling with cortical size, while primary regions show similar areal scaling in humans (Sporns 2013; Reardon et al. 2018; Wei et al. 2019). These results indicate that the allometric expansion of cortical regions associated with higher-order cognitive ability sets humans apart from other primates in evolution.

The human brain has not only evolved with respect to size but also with regards to cell composition and cellular properties. Human pyramidal neurons (e.g., von Economo neurons in layer V), especially those in the prefrontal cortex, are increased in size, have greater complexity of dendritic branching, and increased synaptic spine density compared with chimpanzee pyramidal neurons (Nimchinsky et al. 1999; Allman et al. 2010). Rosehip neurons, which send inhibitory efferences to pyramidal neurons in layer III, were recently identified in layer I of the human cortex, while a similar cell type appears to be absent in rodent (Boldog et al. 2018). However, the existence of rosehip neurons in other primates remains to be examined. Moreover, the increased neuronal connectivity of the prefrontal cortex might promise better computational ability in humans (Elston 2003; Sipser 2012; Reardon et al. 2018). Consistent with this speculation, an interesting finding of the human neocortex is that rather than gray matter, which contains the cell bodies of neurons, white matter (containing the axon fibers) is disproportionately enlarged in human brain evolution (Zhang and Sejnowski 2000; Sousa et al. 2017a). As a result, the glia (e.g., oligodendrocytes and astrocytes) to neuron ratio in humans has increased relative to other primates (Sherwood et al. 2006; Oberheim et al. 2009). Larger pyramidal neurons with more synaptic spines, neuronal fibers, and associated glia cells therefore come together to form a larger and more complex cluster called neuropil (Elston et al. 2001, 2006; Schenker et al. 2008; Spocter et al. 2012). The increased number, size, and complexity of neuropil and corresponding neuronal fibers requires greater cortical space to accommodate them (Elston 2003; Semendeferi et al. 2011; Sousa et al. 2017a; Reardon et al. 2018). These modifications could be one of the underlying reasons for the expansion of the prefrontal cortex in humans.

The evolution of the human brain, or rather neocortex, is comprised of expansion in both tangential and radial directions that correspond to area and thickness (Florio and Huttner 2014). The areal expansion of the human neocortex was discussed above. Here, we will discuss how the human brain has become specialized with respect to thickness and lamination. In the evolution of the mammalian neocortex, cortical thickness generally continues increasing as a function of brain size (Falk and Gibson 2001; Sherwood et al. 2012). However, increasing layer thickness can be accurately predicted by the scaling of the neocortex (Changizi 2001; Sherwood et al. 2012). That means the developmental programs that result in cortical formation in the radial direction are largely conserved in mammals. However, the upper layers (layer II/III) are enlarged in primates compared with other mammals due to “layer shifting” (Hill and Walsh 2005; Hodge et al. 2019). The projection patterns of neurons from different layers have become specialized with the neurons of the upper layers predominately containing intracortical projections while neurons of the lower layers are primarily extracortical projecting (Douglas and Martin 2004; Hutsler et al. 2005). Therefore, the expansion of upper layers in primates is related to increased cortico-cortical connections, ultimately facilitating integration and computation of information (Catani et al. 2012; van den Heuvel et al. 2016; Sousa et al. 2017a).

Another prominent trait of the human brain is its high metabolic cost. The human brain only accounts for about 2% of body mass, but it consumes about 20% of oxygen intake (Kety and Schmidt 1948; Boddy et al. 2012). Meanwhile, the average brain energy consumption rates of other primates and vertebrates are only about 12% and 5%, respectively (Mink et al. 1981). According to Kleiber’s law, which is an empirical model to describe the relationship between mass and energy consumption (Kleiber 1932, 1947), unit energy consumption should decrease as brain mass increases. Therefore, as brain size increased in humans, the unit energy consumption should have decreased in humans compared with other primates. However, energy consumption per unit volume is similar between humans and rhesus macaques (Noda et al. 2002; Karbowski 2007). Therefore, beyond neuroanatomy, the human brain also specializes in metabolic traits. However, given a positive correlation between the complexity of connectivity and metabolic rate across brain regions (Elston et al. 2006; Semendeferi et al. 2011; Sherwood et al. 2012), the metabolic specialization of the human brain could be thought of as a derivative of neuroanatomical specialization.

Development

The expansion of the human brain, especially the neocortex, is largely due to a specialized developmental trajectory. Compared with other primates, humans have longer prenatal and postnatal developmental stages (Kornack and Rakic 1998; Leigh 2004; McNamara 2002; Hill and Walsh 2005). The average human gestation time is about 268 days, while it is only 167 and 226 days for rhesus macaque and chimpanzee, respectively (Peacock and Rogers 1959; Silk et al. 1993; Jukic et al. 2013). For postnatal development, humans take about 11–13 years to reach puberty, while the average juvenile period of other primates is about 4–7 years (Marson et al. 1991; Terasawa and Fernandez 2001; Kail and Cavanaugh 2018). This protracted period of development has been put forth as a major reason for the expansion of the human neocortex (Vrba 1998; Shaw et al. 2008), especially cortical regions such as prefrontal cortex (PFC) (Huttenlocher and Dabholkar 1997; Liu et al. 2012). PFC, one of the most expanded cortical regions in humans, has delayed maturation and hence a longer developmental trajectory than other brain regions (e.g., ×4 longer than cerebellum) (Somel et al. 2011). The periods of prenatal and postnatal development also correspond with neurogenesis and synaptogenesis, respectively, both processes working together to result in greater numbers of more complex neurons in the human brain (Kornack and Rakic 1998; Hill and Walsh 2005; Petanjek et al. 2008; Bianchi et al. 2013). Therefore, in this section, we will discuss how these two developmental mechanisms/stages contribute to human brain evolution and cognition.

Brain size in a particular lineage (e.g., primate) is generally positively correlated with the number of neurons (Striedter 2005; Azevedo et al. 2009). Neuron abundance is mainly determined by the number of progenitor divisions in development, especially the number of divisions of early progenitor cells in the ventricular zone (VZ) that determine the number of cortical columns and hence neuron abundance (Geschwind and Rakic 2013). Consistent with variations in brain size, mouse neural progenitors divide 11 times before differentiating to neurons (Takahashi et al. 1995), while in macaques they undergo at least 28 divisions (Kornack and Rakic 1998). Although the exact number of divisions is still unknown, human neural progenitors are thought to divide many more times than other primates based on the duration of the cell cycle and developmental period (Hill and Walsh 2005; Kriegstein et al. 2006). Hence, in general, the evolution of larger brains in mammals is associated with an increasing number of divisions in neural progenitors.

Although neuronal cell types are diversified in mammalian evolution, the types of neural progenitors are highly conserved (Lein et al. 2017). Neural progenitors can be distinguished into apical progenitors (APs) and basal progenitors (BPs) based on the location of their cell body (Kriegstein et al. 2006). The APs are closer to the ventricle and form the VZ, while BPs are closer to the pia and form the subventricular zone (SVZ) (Florio and Huttner 2014).

Neuroepithelial cells (NECs), one type of AP at the VZ, are the most primitive neural progenitor. NECs undergo symmetric division to initiate a progenitor pool for later proliferation and differentiation. Compared with rodents, primates have a ×10 longer absolute duration of proliferation time of NECs (Caviness et al. 1995; Rakic 1995, 2009; Geschwind and Rakic 2013); this protraction of proliferation time likely contributes to the expansion of neocortex in primates.

NECs give rise to another AP, apical radial glia (aRG), also known as ventricular radial glia (vRG), which undergo different types of cleavage to generate daughter cells (Konno et al. 2008; Florio and Huttner 2014). One type of cleavage happens on the vertical plane to generate two aRGs, which keep the potential for self-replication, while the other cleavage happens on the diagonal plane to give rise to two BPs that move to the SVZ (Konno et al. 2008; Shitamukai and Matsuzaki 2012). The evolution of mammalian brains acts on the developmental program to control the two types of cleavage. Mammals with small brains (e.g., rodents) mainly carry out diagonal cleavage after a few rounds of vertical cleavage, while primates mainly undergo vertical cleavage to expand the aRG pool before diagonal cleavage (Lancaster and Knoblich 2012; LaMonica et al. 2013; Florio and Huttner 2014; Fig. 1A, B).

Fig. 1.

Schematic diagram of prenatal neocortical development in eutherian mammals. (A) Rodent, (B) non-human primate, and (C) human. Primates have a prolonged neocortical developmental, with the longest duration occurring on the human lineage. Compared with rodents, primates undergo more rounds of symmetric division and proliferation in aRGs and bIPs, respectively, to expand the progenitor pool, and hence primates have prolonged durations of the proliferation of both APs and BPs. Genes associated with symmetric division of aRGs and bIPs are upregulated in primates. Compared with non-human primates, human bRGs undergo greater numbers of proliferative division before differentiative division, and hence the duration of proliferation of BPs is prolonged in human. Genes that facilitate bRG proliferation are upregulated in humans, while genes associated with bRG differentiation are upregulated in non-human primates.

After diagonal cleavage, aRG generate two types of BP: basal radial glial (bRG) also known as outer radial glia (oRG) and basal intermediate progenitor (bIP). With brain expansion, the properties and composition of BPs underwent species-specific specialization. bIPs can be classified as neurogenic bIPs and proliferative bIPs. About 90% of bIPs are neurogenic in rodents (Noctor et al. 2004; Arai et al. 2011; Wang et al. 2011), while most bIPs in primates are proliferative (Hansen et al. 2010; Lui et al. 2011). Rodent bRGs mainly undergo self-renewing division to generate one bRG and one neuron (Shitamukai et al. 2011; Wang et al. 2011), while primate bRGs proliferate to form a large bRG pool (Reillo et al. 2011; Betizeau et al. 2013; Fig. 1A, B). As a result, bRGs account for about 10% and 70% of BP pool in rodents and primates, respectively (Hansen et al. 2010; Lui et al. 2011; Florio and Huttner 2014). In primates, the abundant bRGs, which migrate further and are closer to the basal side, form a lineage-specific structure called the outer SVZ (oSVZ). The remainder of the SVZ, called the inner SVZ, is equal to the SVZ of rodents and is mainly composed of bIPs (Smart et al. 2002; Reillo et al. 2011; Florio and Huttner 2014). Expansion of the bRG pool has been associated with cortical gyrification, and manipulation of bRG division can result in gyrification of mouse brain and affect the degree of folding in ferret brain (Nonaka-Kinoshita et al. 2013; Stahl et al. 2013; Matsumoto et al. 2020). Enlargement of the oSVZ in primates, a process absent in rodent, provides a mechanism to explain why the neocortex disproportionately expands with respect to surface area compared with the ventricle area and ultimately results in gyrification in primates (Fietz et al. 2010; Konopka and Geschwind 2010). In humans, although the developmental program that drives expansion of the oSVZ is not essentially different from other primates, the amount of time of proliferation of BPs is prolonged (Kriegstein et al. 2006; Fig. 1B, C). Another feature of human brain development is the slower migration of human neurons than neurons from other primates in a neurosphere migration assay (Marchetto et al. 2019). Therefore, the protracted proliferation of progenitors and slow migration of human neurons could account for the extended duration of prenatal development and expansion of human neocortex relative to other primates. Additionally, the expansion of SVZ in later prenatal development has been associated with the enlargement of the upper layers in primates relative to rodents (Smart et al. 2002; Martínez-Cerdeño et al. 2006).

Similar to prenatal development, the process of human postnatal development is also prolonged relative to other primates and delayed due to protraction of prenatal development (Liu et al. 2012; Zhu et al. 2018). Postnatal development of the brain is mainly associated with synaptic maturation which includes neuropil expansion, synaptic elimination, and myelination. Neurite growth and synaptic formation are the main processes to increase synaptic density and result in neuropil enlargement (Spocter et al. 2012). Synaptic density peaks shortly after birth without any differences across cortical regions in macaque and chimpanzee (Rakic et al. 1986). In contrast, the peak timing of human synaptic density varies from about 2 years in primary areas (e.g., auditory cortex) to about 8 years in associative areas (e.g., PFC) (Huttenlocher and Dabholkar 1997; Huttenlocher et al. 1982). As a result of the protracted time for synaptic formation, the human neocortex, especially the PFC, undergoes further expansion which is associated with more complex neuropil and more space for neuronal fibers (Liu et al. 2012; Somel et al. 2013). Synapse elimination and myelination, which are the final stages of synaptic maturation, are also prolonged in humans. Chimpanzees finish this process with sex maturation around 10 years, while the synaptic maturation of humans is protracted to 30 years (Petanjek et al. 2011; Miller et al. 2012). The protracted maturation of synapse formation could also be associated with a longer time window for synaptic remodeling that underlies and/or facilitates learning. Consistent with neuroanatomical findings, late childhood has been suggested to be a particularly unique aspect of human brain development based on morphophysiological and behavioral evidence (Schultz 1960), and recent transcriptomic data have shown that macaque brains lack the global transcriptomic signatures of late childhood that human brain demonstrate, further confirming the unique contribution of postnatal development to human cognition (Zhu et al. 2018).

Gene expression

Earlier studies comparing biological traits and protein similarity between human and chimpanzee suggested that human evolution might primarily result from alteration of gene expression rather than protein sequence (King and Wilson 1975). Consistent with this hypothesis, gene expression in the human brain has been found to undergo accelerated changes compared with other primates (Enard et al. 2002a; Preuss et al. 2004). A greater number of differentially expressed genes (DEGs) between human and chimpanzee are in the brain relative to other organs (e.g., liver, heart) (Cáceres et al. 2003; Gu and Gu 2003; Khaitovich et al. 2004). PFC is enriched for DEGs compared with other brain regions (e.g., cerebellum and primary sensory areas) (Nowick et al. 2009; Wei et al. 2019); however, other datasets have shown that the striatum has the greatest number of DEGs between human and other primates (Sousa et al. 2017b). Co-expression network analysis, which captures how genes co-vary to form co-expression networks across samples, have identified human-specific co-expression networks, particularly within the PFC (Konopka et al. 2012; Xu et al. 2018). Therefore, in addition to overall brain and region expansion, these results support the idea that the human brain has specialized gene expression.

Further analysis of the brain DEGs between human and other primates (mainly chimpanzee) has discovered that the human brain tends to have more upregulated genes relative to other primates (Cáceres et al. 2003; Gu and Gu 2003; Khaitovich et al. 2004; Kanton et al. 2019), while DEGs of other organs (e.g., heart and liver) show equal numbers of upregulated and downregulated genes (Cáceres et al. 2003). These DEGs are enriched for genes involved in neuronal activity, metabolism, neurotransmitter receptor, and synaptic and extracellular matrix functions (Cáceres et al. 2003; Uddin et al. 2004; Hodge et al. 2019). These findings are consistent with results on the organismal level. For example, the enrichment of upregulated neuronal activity and metabolism-related genes can potentially explain the increased activity and metabolic rate of the human brain (Preuss et al. 2004). Moreover, the more expanded cortical areas (e.g., PFC) have a greater enrichment of metabolic genes (Khaitovich et al. 2008; Somel et al. 2013; Rubinov 2016). A mechanistic explanation for these findings suggests that an overall upregulation of mRNA could prepare human brain cells for a rapid response to external stimuli via a supply of pre-generated transcripts (Preuss et al. 2004). Additionally, overall increased neuronal activity in the human brain could activate genes associated with cytoprotection and hence facilitate neuronal survival (Pelicci et al. 2002; Cáceres et al. 2003; De Sarno et al. 2003).

Gene expression in the human brain has become specialized across developmental time points. Given the prominent difference in cortical and regional size between humans and other primates, gene expression alterations during development can play an important role (Geschwind and Rakic 2013; Bae et al. 2015; Silbereis et al. 2016). Comparison of gene expression along developmental trajectories across primates has found that the overall patterns of their trajectories are conserved (Zhu et al. 2018). This suggests that the overall developmental program has been constrained in brain evolution of primates. However, the expression signatures along these trajectories also suggest that humans have two developmental stages with prolonged durations (Somel et al. 2013; Zhu et al. 2018). Consistent with neuroanatomical development, one stage is during the early prenatal period, while the other prolonged stage is during the postnatal period (from birth to puberty) (Zhu et al. 2018). Moreover, in vitro studies of cultured neurons and brain organoids have implied a delayed maturation of human neurons by showing that maturation-related genes are upregulated in chimpanzee neurons compared with time-matched human neurons (Mora-Bermúdez et al. 2016; Kanton et al. 2019; Pollen et al. 2019).

In the prenatal neocortex, human-specific upregulated genes are enriched for genes associated with neural progenitor identities, while macaque-specific upregulated genes are enriched for genes associated with mature neuron identities (Zhu et al. 2018). Comparison of gene expression across different cortical regions of the human brain shows that the expression profile of the PFC is more enriched for genes related to neural development function, such as neurogenesis, differentiation, and migration (Zhu et al. 2018; Kanton et al. 2019). These results echo more rounds of proliferation and hence more progenitors during the prenatal development of the human brain (Kriegstein et al. 2006; Florio and Huttner 2014). The proliferation marker Pax6 is downregulated in rodent bIPs, which are mainly composed of self-consuming subtypes that differentiate into neurons (Miyata et al. 2004; Noctor et al. 2004; Florio and Huttner 2014), while bIPs, which maintain PAX6 expression level in primates, maintain a proliferative state to generate more bIPs for cortical expansion (Fietz et al. 2010; Betizeau et al. 2013; Fig. 1A, B). As mentioned above, the vertical plane of division in primates results in proliferation while the diagonal plane of rodent division causes differentiation in aRG (Konno et al. 2008; LaMonica et al. 2012). The division plane is determined by the orientation of mitotic spindles (Lancaster and Knoblich 2012), and mitotic spindle associated genes have been found to be upregulated in the human VZ (Miller et al. 2014). In addition, the primate subplate, a region of the cortex just below the six layers that expands during development, shows a diversified gene expression signature compared to rodents (Miller et al. 2014; Hoerder-Suabedissen and Molnár 2015). The increased complexity of the primate subplate might be an additional mechanism for layer expansion and reorganization.

During postnatal development, more than 70% of genes show variation in expression across development, yet their developmental trajectories show similar patterns between human and other primates (Somel et al. 2009). However, the developmental duration is substantially protracted in humans (Somel et al. 2011, 2013; Zhu et al. 2018). Interregional variation of gene expression trajectories has been found in human but not in macaque (Somel et al. 2011). This suggests that the regional expansion of the human neocortex might result from gene expression trajectories specialized to maintain a longer developmental stage. Consistent with findings in neuroanatomy, the PFC is one of the cortical regions with the most protracted gene expression trajectory (Somel et al. 2011; Liu et al. 2012; Geschwind and Rakic 2013). Enrichment analysis on genes varying along trajectories has uncovered gene functions for axon guidance, synaptogenesis, synaptic activity, myelination, and synapse elimination (Konopka and Geschwind 2010; Somel et al. 2013; Zhu et al. 2018). Synaptic genes in the human PFC peak around 5 years compared with only a couple of months in other primates (Somel et al. 2013). The prolonged expression of synaptic genes corresponds with a longer duration of synaptogenesis and hence more synaptic connections in human (Somel et al. 2011, 2013; Liu et al. 2012; Geschwind and Rakic 2013). The more abundant synapses and protracted synaptic elimination process in the human PFC provide a higher level and longer time of plasticity for learning (Geschwind and Rakic 2013; Sousa et al. 2017a). This might be a major mechanism that drives the uniqueness of the human brain and cognition.

Genome and genes

Each genome including sequence, epigenetic, and structural information encodes all species-specific biological traits, and thus the comparison of genomic and genetic signatures is a fundamental approach to understand the mechanisms which result in species-specific cognitions. The human genome includes more than 3 billion base pairs of which about 2% are protein-coding regions and about 15% regions are functional non-coding regions, including regions of regulatory elements and non-coding RNAs (Ponting and Hardison 2011; Venter et al. 2001). Sequence comparisons between human and chimpanzee identified about 35 million nucleotide substitutions and 90 million base pairs of structural alteration, including insertion or deletion (indels), inversions, and duplications (Cheng et al. 2005; Waterson et al. 2005).

Mutations that occur in coding areas could alter biological properties of the original protein, create new proteins, or even totally remove proteins (Loewe 2008). Genome-wide comparisons between human and chimpanzee discovered that about 500–1000 protein-coding genes were under positive selection in humans (Waterson et al. 2005; Scally et al. 2012), which was defined as higher nonsynonymous substitution rates than synonymous substitution rates (Scally et al. 2012). These genes are enriched for brain-related functions (Liu et al. 2012). Interestingly, the sequences of DEGs between humans and other primates are under higher levels of positive selection than other genes (Zhu et al. 2018). One of the most famous genes under positive selection might be human FOXP2 which evolved two amino acid substitutions compared with other primates (Enard et al. 2002b, 2009). As a transcription factor (TF), these relatively small modifications substantially changed its downstream targets and resulted in profound effects on neural development and vocal motor control (i.e., a behavior related to human speech) (Lai et al. 2001; Teramitsu et al. 2004; Konopka et al. 2009). With regards to structural mutation, deletion of olfactory receptor genes is pronounced in primates that evolved to switch to visual-dependent information integration from an odor-dependent ancestor (Gilad et al. 2003, 2005). Eight families of protein-coding genes have been detected as having undergone duplication in humans relative to other primates, and some of these duplicated genes acquired novel functions for synaptic formation (Fortna et al. 2004; Goidts et al. 2006; Sudmant et al. 2010; Sousa et al. 2017a).

The majority of nucleotide substitutions between human and chimpanzee occurred at non-protein coding regions (Waterson et al. 2005). Whole-genome comparisons between human and chimpanzee identified human accelerated regions (HARs). Most HARs are located in non-protein coding regions, which may be the regulatory regions of adjacent genes (Pollard et al. 2006). Further analysis of these genes found that they are enriched for brain developmental functions especially for associative areas (Pollard et al. 2006; Doan et al. 2016). Given the extensive differential gene expression between humans and other primates (Konopka et al. 2012; Berto et al. 2019; Sousa et al. 2017b; Zhu et al. 2018), evolutionary differences, including cis-regulatory elements (CREs), TFs, and epigenetic modification, are predicted to be located in gene regulatory regions. The most common gene regulatory mechanism would be a TF interacting with a promotor CRE to activate gene transcription. Studies comparing the contribution of TFs and promotor CREs to cortical DEGs between humans and chimpanzees consistently show that TFs play an important role (Xu et al. 2018). Further analysis has found that these TFs are enriched for genes involved in neural development and brain disorders (Xu et al. 2018; Zhu et al. 2018). Epigenetic modifications that differentiate human and chimpanzee are mainly methylation changes on promotor and enhancer CREs, which account for about 20% and 6% expression differences, respectively (Xu et al. 2018). Methylation usually happens at CpG islands of DNA sequence but non-CpG methylation is common in neurons (Rizzardi et al. 2019) and is mainly associated with the downregulation of gene expression (Wang et al. 2018b). Consistent with gene expression data, the chimpanzee brain shows higher methylation levels than the human brain (Somel et al. 2013; Wang et al. 2018b; Berto et al. 2019; Mendizabal et al. 2019). Again, the differentially methylated genes are enriched for neural function (Houston et al. 2013). CLOCK was recently discovered to be less methylated in its regulatory regions and hence upregulated in human brain (Babbitt et al. 2010; Konopka et al. 2012; Berto and Nowick 2018; Berto et al. 2019; Mendizabal et al. 2019). Consistent with the slower migration of human NPCs, the upregulation of CLOCK in human brain has been associated with slower migration of NPCs (Fontenot et al. 2017). Another genomic mechanism that may be relevant to human brain evolution is an increase in copy number through gene duplication. Such duplicated genes, which will be discussed in detail later, are upregulated in humans (Sudmant et al. 2013; Marquès-Bonet et al. 2004). Therefore, genomic differences in multiple gene regulatory mechanisms can explain the differences in gene expression that account for human brain evolution.

Functional studies on species-specific genes also provide insightful results to understand human brain evolution. In the evolution of the gyrencephalic primate brain, TIS21 was downregulated while PAX6 was upregulated to maintain the proliferative potential of bIPs compared with rodents (Fietz et al. 2010; Arai et al. 2011; Fig. 1A, B). ARHGAP11B, which is only expressed during development of the human brain, has been found to increase the rate of mitosis in bRGs, and forced expression of ARHGAP11B in mouse and the common marmoset resulted in folding of the neocortex and an increased number of bRGs in the oSVZ, respectively (Prüfer et al. 2014; Florio et al. 2015; Heide et al. 2020; Sousa et al. 2017a; Fig. 1B, C). A balance between Robo and Dll1 signaling has been found to control the formation of BPs in neurogenesis, which determines cortical expansion in amniote evolution (Cárdenas et al. 2018). Given that BPs play an essential role in human cortical expansion, Robo and Dll1 signaling might also be involved in human brain specialization. NOTCH2NL, which is duplicated from NOTCH2, prolongs duration of BP proliferation through the Notch pathway, and copy number variation of this gene correlates with human brain size (Fiddes et al. 2018; Suzuki et al. 2018). Furthermore, forced expression of NOTCH2NL in the developing mouse brain could result in an increased number of BPs (Florio et al. 2018; Fig. 1B, C). Positive selection of ASPM and MCPH1 has also occurred in humans (Evans et al. 2005; Mekel-Bobrov et al. 2005). Both genes control spindle activity, which has been associated with the fate of radial glia through the determination of the division plane (Bond et al. 2002; Konopka and Geschwind 2010). The DUF1220 protein domain, which is a component of genes in the Neuroblastoma breakpoint family for neural development, largely expanded its copy number from 28 copies in chimpanzee to 90–125 copies in human (Fortna et al. 2004; Popesco et al. 2006; O’Bleness et al. 2012). Its increasing copy number has been found to promote the proliferation of APs in the VZ (Keeney et al. 2015; Sousa et al. 2017a). Additionally, enhancer deletion of GADD45G in humans has been associated with the human-specific expansion of the SVZ (McLean et al. 2011). Therefore, at least part of the selection pressure to evolve the human brain acts on genes that control cell proliferation during neural development.

Human MCPH1 not only functions to facilitate cellular proliferation during prenatal development but is also responsible for the prolonged development of human neurons (Shi et al. 2019). Transgenic macaques with human MCPH1 have been found to have a protracted postnatal development time with increased synaptic complexity and better performance in a short-term memory test (Shi et al. 2019). A downstream target of MCPH1 called MEF2A, which shows human-specific positive selection on its CRE relative to other hominids, is upregulated in human (Shi et al. 2019). MEF2A functionally postpones brain development in two ways. One way is through the suppression of the function of NR4A1 to eliminate dendritic spines, which have been described as a sign of neuronal maturation (Hawk and Abel 2011). The other way is to activate early growth response genes for synaptic protein synthesis which could facilitate neurite growth and postpone maturation (Flavell et al. 2008; Somel et al. 2013). SRGAP2 is a conserved mammalian gene involved in neuronal maturation (Fossati et al. 2016). Human-specific duplication of SRGAP2 creates a paralog called SRGAP2C that antagonizes the function of SRGAP2 (Geschwind and Rakic 2013; Fossati et al. 2016). As a result, SRGAP2C prolongs synaptic maturation and increases the density of dendritic spines and the length of dendritic shafts (Charrier et al. 2012; Sousa et al. 2017a). Additionally, human FOXP2 also increased dendrite length and improved performance on memory tests in mouse (Enard et al. 2009). Therefore, human brain evolution is also associated with genetic modification of synaptic maturation.

Another human-specific brain trait is its exceptionally high metabolic rate. GLUD2, which evolved from GLUD1, gained its human-specific function in mitochondria (Rosso et al. 2008; Konopka and Geschwind 2010). Human ZNF331 which lacks orthologous genes in rodents plays an important role for activity regulation (Ataman et al. 2016; Hardingham et al. 2018). OSTN, CAMTA1, and TUNAR underwent human-specific alteration of function to respond to neural activity (Qiu et al. 2016; Pruunsild et al. 2017). This novel function facilitates the modification of synaptic connections after receiving neuronal inputs, so it is a potential mechanism for effective learning from experiences (Hardingham et al. 2018).

Synthesis

A comprehensive understanding of cognitive specialization requires integrative comparisons among phylogenetically closely-related species under all of Tinbergen’s four questions. In this review, we started with brain anatomy, and then discussed some of the underlying developmental mechanisms for anatomical alterations. Next, we described gene regulatory mechanisms and ended with genomic/genetic comparisons across species. Compared with our closest relative, chimpanzee, the human brain has expanded ×3 relative to its scaling size. Expansion of the neocortex, especially the PFC, mainly accounts for human brain evolution. Brain and regional expansion have been attributed to alterations of a developmental program. In humans, brain development is protracted at both the prenatal and postnatal stages relative to other primates. In prenatal stages, human NPCs undergo more proliferative cycles for longer developmental times, leading to a greater number of neurons compared with other primates. In postnatal stages, a prolonged development time facilitates neurite growth and synaptic formation. These developmental mechanisms ultimately result in a greater number and higher complexity of neuropils for cortical expansion and the enhanced computational capacity of humans. Consistent with the anatomy and development findings, gene sequence and regulatory regions are under positive selection in human for genes associated with brain developmental functions. Although these insights are notable achievements toward understanding the evolution of the human brain, future work on the functional aspects of genome regulation is needed to further understand the genetic mechanisms of proliferation, differentiation, neurite growth, and synaptic formation. However, these studies provide a blueprint for scientists who are interested in studying comparative cognition in other organisms.

Cognitive specializations are shaped by selection pressures for animals to adapt to physical and social environments. The most classic study compared spatial learning between food-caching and nonfood-caching birds. In the task of using spatial cues to locate hidden food, food-caching birds committed fewer errors than nonfood-caching birds (Shettleworth et al. 1990). Further comparison of neuroanatomy showed that food-caching birds have a larger relative hippocampal volume than nonfood-caching birds (Healy and Krebs 1992). In recent years, comparative cognition has received increased attention from organismal biologists. For example, food-caching chickadees that live in harsher environments evolved better spatial memory than populations in milder environments, and this difference in spatial learning has been associated with a difference in adult seasonal neurogenesis (Roth et al. 2012). An active-foraging lizard species showed higher levels of behavioral flexibility than closely-related species with a sit-and-wait strategy (Day et al. 1999). Frogs with parental care perform better in place learning and behavioral flexibility tests and show mammal-like spatial cognition in a modified Morris Water Maze (Liu et al. 2016, 2019; Pašukonis et al. 2018). Brain development patterns on anterior–posterior axes are associated with different ecotypes of cichlid fishes (Sylvester et al. 2010). Due to the emergence of next-generation sequencing technology (Wang et al. 2009), transcriptomic analyses have been applied to understand the gene regulatory mechanisms for cognitive differences. Consistent with human brain specialization, more advanced cognitive abilities have been associated with the upregulation of genes involved in neurogenesis (Pravosudov et al. 2013; Liu et al. 2020). In the future, high-quality whole-genome sequences will be needed to understand gene regulatory mechanisms. CRISPR-Cas9 and RNA interference technologies could be applied to study the function of candidate genes that are under selection (Doudna and Charpentier 2014; Hannon 2002). Additionally, given the cellular diversity of the brain and newly appreciated cell type-specific gene expression, single-cell/nucleus sequencing could capture a more accurate assessment for species comparisons (Tang et al. 2009; Tosches et al. 2018).

 

From the symposium “Integrative comparative cognition: can neurobiology and neurogenomics inform comparative analyses of cognitive phenotype?” presented at the annual meeting of the Society for Integrative and Comparative Biology January 3–7, 2020 at Austin, Texas.