Topography and topology of parts of the cerebral cortex

Spatial relations between parts of the nervous system (or any other organ) can be described from two points of view. First, topography studies the distribution of parts on the surface of or within an organ; in the present article, we study the topographical distribution of cortical types across the cerebral cortex in adult brains of rats and primates. Second, topology studies the invariant neighborhood relations between parts on the surface of or within an organ unaffected by the continuous change of shape or growth of this organ; in the present article, we study the topological distribution of cortical types (rings) unaffected by the expansion of the cerebral cortex across phylogeny [for definitions of topography and topology applied to brain structures see Nieuwenhuys (2011, 2017)].

The topographical arrangement of cortical types in concentric rings was identified by Sanides in Nissl and myelin-stained sections of adult brains. The outer location of allocortex, agranular neocortex (periallocortex) and dysgranular neocortex (proisocortex) rings is easy to spot in coronal sections of rodent brains that show dorsolateral isocortical areas flanked, on the medial side, by cingulate agranular and dysgranular areas and, on the ventrolateral side, by insular agranular and dysgranular areas (Fig. 1a). In primates, the topographical analysis of cortical types shows that agranular neocortex (periallocortex) and dysgranular neocortex (proisocortex) are partially or totally covered by isocortical areas (Fig. 1b). This is due to the extensive expansion of isocortical areas in primates that expand over limbic areas like cupcake batter flows over a cupcake mold (Fig. 1c–f). Thus, topographically, agranular and dysgranular areas may appear more “central” than “peripheral” in primates compared to rodents but, topologically, agranular and dysgranular limbic areas flank isocortical areas both in rodents and primates.

Fig. 1

Topography and topology of cortical types. a Coronal section of the rat brain through the olfactory tubercule. Dorsolateral isocortical (eulaminate with well-developed layer IV, white and light shades of gray) areas are flanked on the medial side by the allocortical precommissural hippocampus (Hipp, black) and cingulate agranular and dysgranular areas (dark shades of gray); Isocortical areas are flanked on the ventrolateral side by allocortical olfactory areas (Pir, black) and insular agranular and dysgranular areas (dark shades of gray). b Coronal section of the human brain through the nucleus accumbens. Dorsolateral isocortical (eulaminate with well-developed layer IV, white) areas are flanked on the ventromedial side by the allocortical precommissural hippocampus (Hipp, black) and cingulate agranular and dysgranular areas (dark shades of gray); isocortical areas are flanked on the ventral side by allocortical olfactory areas (AON, black) and insular and posterior orbitofrontal (pOFC) agranular and dysgranular areas (dark shades of gray). c Lateral view of a small cupcake. d Lateral view of a large cupcake. e, Vertical section of the small cupcake in c; the gray scale in c and e resembles the distribution of cortical types on the rat brain in a. f Vertical section of the large cupcake in d; the gray scale in d and f resembles the distribution of cortical types on the human brain in b. The topological neighborhood relations of cortical types (shades of gray) in a, c, and e are preserved in b, d, and f, despite the extensive expansion of isocortical areas in primates that expand over limbic areas like cupcake batter flows over a cupcake mold. ac anterior commissure, accum nucleus accumbens, AON anterior olfactory nucleus in the primary olfactory cortex, cc corpus callosum, Cd caudate nucleus, Cl claustrum, Endo endopiriform nucleus, Hipp anterior extension of the hippocampal formation, LOT lateral olfactory tract, Pir piriform cortex in the primary olfactory cortex, pOFC posterior orbitofrontal cortex, Put putamen, sm stria medullaris, TOL olfactory tubercule. a, b are modified from Fig. 3 in Garcia-Cabezas and Zikopoulos (2019)

In this article we describe the topographic distribution of cortical types in rats, macaques, and humans. Then, we identify the invariant neighborhood relations (topological relations) of cortical types across rats, macaques, and humans.

Terminology of cortical types

We first define the terms that will be used through the present article to label parts of the cerebral cortex in the three species analyzed (Table 2).

Table 2 Nomenclature of cortical types

The cerebral cortex can be divided in areas and types according to its microscopic structure as seen in Nissl-stained sections. Cortical areas are parts of the cortex with characteristic cytoarchitectonic features (or a combination of features) that are absent in other parts. In contrast, Cortical types are defined based on “the constant variations that one observes in each of the layers in different regions” (von Economo and Koskinas 1925/2008; von Economo 1927/2009). For cortical type analysis, instead of looking for particular cytoarchitectonic features, the observer looks across the cortical quilt for gradual and systematic variation of laminar features, like number and prominence of layers, or size of largest neurons in layers, to name a few (García-Cabezas et al. 2020). Cortical types are topologically arranged in concentric rings across the surface of the cortex (Sanides 1970); thus, we will use Cortical ring as synonym of cortical type. Pallial sectors are parts of the cortex that are patterned during early stages of development due to the action of morphogenetic proteins; four pallial sectors (medial, dorsal, lateral, and ventral pallium) are defined using genoarchitectonic techniques (Puelles 2017).

Several cortical types can be defined across gradients of laminar elaboration. We will briefly describe the laminar features of cortical types and their topological relations across the gradient of laminar elaboration. Allocortex is the part of the cortex with the simplest laminar elaboration. The hippocampal formation (archicortex) and the primary olfactory cortex (paleocortex) are the 2 parts of the allocortex and have areas with 2 or 3 layers. In contrast to allocortical areas, the Isocortex, is composed of areas with 6 well-developed layers. We use Eulaminate area as synonym of isocortical (6-layer) area. We also use Neocortex sensu stricto as synonym of isocortex. The isocortex is not homogeneous and several eulaminate types of progressive laminar elaboration can be defined. At the end of the isocortical gradation in primates are Koniocortical areas, which are eulaminate areas with the most complex laminar elaboration.

Between allocortical and isocortical areas is the Mesocortex, which is composed of areas that have more layers than the allocortex, but fewer than the isocortex; thus, mesocortical areas either lack or have rudimentary granular layer IV. We use Neocortex non sensu stricto as synonym of Mesocortex. Mesocortical areas that lack layer IV are Agranular and those which have rudimentary layer IV are Dysgranular. Agranular areas are adjacent to allocortical areas and are also called Periallocortical. Dysgranular areas are between agranular areas and isocortical areas and are also called Proisocortical. Limbic areas are those at the limit of the hemisphere; thus, allocortical and mesocortical areas are limbic areas.

In the present article we analyze cortical types in isocortical (neocortex sensu stricto) and mesocortical (neocortex non sensu stricto) areas. We do not analyze allocortical areas [primary olfactory cortex: piriform cortex and anterior olfactory nucleus; hippocampal formation: dentate gyrus, fields in Ammon’s horn, subiculum, presubiculum, parasubiculum, taenia tecta, indusium griseum; Nieuwenhuys et al. (2008); Puelles et al. (2019)]. Recent genoarchitectonic data suggest that the entorhinal cortex, which traditionally has been considered periallocortical, is allocortex (Puelles et al. 2019); therefore, we also excluded the entorhinal cortex for cortical type analysis.

Principles of cortical type analysis

We will briefly describe the general principles of cortical type analysis. For more details, the reader is referred to our previous article on the human cerebral cortex (García-Cabezas et al. 2020).

Laminar architecture varies systematically across the entire surface of the neocortex in gradients of laminar elaboration with increasing development of granular layer IV and sharper definition of the other layers. We have traced cortical gradients in primates (García-Cabezas and Barbas 2017; García-Cabezas et al. 2017, 2019, 2020; Zikopoulos et al. 2018) and have identified the most useful laminar features for cortical type analysis using Nissl-stained sections. These features, summarized in Table 3, include development of layer IV, prominence (denser cellularity and larger neuron bodies) of deep (V–VI) or superficial (II–III) layers, definition of sublayers (e.g., IIIa and IIIb), sharpness of boundaries between layers, and presence of largest pyramids in layers III and/or V (García-Cabezas et al. 2020).

Table 3 Laminar features of cortical types in the neocortex

A key aspect of cortical type analysis is that transitions from one type to another are gradual without abrupt jumps along the gradients of laminar elaboration. Another aspect to consider is that some cortical areas defined in classical studies [e. g., Brodmann (1909/1999); von Economo and Koskinas (1925/2008); von Bonin and Bailey (1947)] will have just one cortical type, others may be in the middle of transitions from one type to another, and others may be large enough to span through several cortical rings.

Equivalence of cortical types across studies

Finally, we will make some comments regarding the objectivity of cortical type analysis. It must be taken into account that progressive elaboration of laminar structure of the cerebral cortex across gradients is an objective fact revealed by microscopic examination of Nissl-stained sections of mammalian brains. In contrast, the definition of cortical types across gradients of laminar elaboration is subjective, although not arbitrary. We defined 1 allocortical type, 2 mesocortical types (agranular and dysgranular) and 4 isocortical types (eulaminate I, II, II, and koniocortex) in the human cortex (García-Cabezas et al. 2020), other researchers defined 5 types in prefrontal areas (Barbas and Pandya 1989; Barbas and Rempel-Clower 1997), 6 types across the entire neocortex (Barbas 1986; John et al. 2021), or defined 8 types in visual areas of the Rhesus macaque (Hilgetag et al. 2016); others defined 3 types in the human cortex (Zhang et al. 2020), but other researchers may define fewer or more types (e.g., 4 types: allocortex, mesocortex, eulaminate, and koniocortex).

In Table 2, we summarize cortical type equivalences across the most relevant parcellation studies of the macaque and human cortex. The authors of most of these studies performed cortical type analysis based on qualitative features of laminar architecture following the seminal scheme of Helen Barbas for the macaque neocortex (Barbas 1986). The major exception to this rule is the study by Hilgetag et al. (2016), in which cortical type definition is based on laminar features and neuron density because eulaminate types have higher neuron density than mesocortical types. Hilgetag et al. (2016) used unbiased stereological methods to quantify neuron density across visual areas of the macaque and sorted these areas into types according to their laminar features and neuron density. They found that area V1 differed by about three ordinal scales from the densest eulaminate III prefrontal area and, thus, extended the classification from five cortical types for prefrontal areas of Helen Barbas (Barbas and Pandya 1989; Barbas and Rempel-Clower 1997) to eight types for the cortical visual system: V1 was assigned to cortical type 8 and several unimodal visual areas were assigned to cortical type 7 (e.g., V2) and cortical type 6 (e.g., V4). Thus, cortical type 8 defined by Hilgetag et al. (2016) based on neuron density is the equivalent of our Koniocortical type, cortical types 6 and 7 are subsumed in our eulaminate III, and some of the areas assigned to cortical type 5 fall in our eulaminate II. Also, Hilgetag et al. (2016) included the entorhinal cortex (area 28), which is part of the allocortex (Puelles et al. 2019), in their cortical type 1.

In summary, subjective interobserver differences are not arbitrary as long as the topological order of types is preserved. That means one can move across the cortical surface from one type to the next type up or down the gradient with increments no higher than 1. The topological distribution of cortical types across gradients of laminar elaboration from allocortex to koniocortex allows to relate classification schemes of cortical types of different studies to each other.

The rat neocortex is mostly mesocortical

We performed cortical type analysis on Nissl-stained sections across the entire cortical quilt in rats. For that purpose, we took micrographs across all layers of neocortical areas described by Zilles (1985) for analysis of laminar features summarized in Table 3.

Mesocortical areas adjacent to hippocampal and olfactory allocortical areas showed overall simple lamination without layer IV, indistinct layers V–VI, and irregular boundary between layers I and II. These areas had more prominent and denser layers V–VI than layers II–III and the largest bodies of pyramidal neurons were in layer V (Fig. 2A–E). We categorized these areas as Agranular.

Fig. 2

Cortical types across the rat cerebral neocortex. a–e, Micrographs of agranular mesocortical areas (Nissl staining). f–j, Micrographs of dysgranular mesocortical areas (Nissl staining). k–o, Micrographs of eulaminate areas (Nissl staining). Cortical areas are indicated according to Zilles (1985); see Table 4 for abbreviations of areas. Roman numerals indicate cortical layers. WM: white matter. Calibration bar in o applies to a–o

Other mesocortical areas adjacent to agranular areas also had simple lamination, but an incipient layer IV could be identified. In these areas, layers V–VI were more prominent than superficial layers II–III and the largest bodies of pyramidal neurons were in layer V (Fig. 2F–J). Layers V–VI were slightly more distinct than in agranular areas. We categorized these areas as Dysgranular.

Finally, few areas adjacent to dysgranular areas and apart from agranular areas had well-developed layer IV. In these areas, layers V–VI were still more prominent but superficial layers II–III were denser than in agranular and dysgranular areas. The largest pyramids, which in Hind Limb (HL) motor area were the largest bodies of pyramidal neurons across the entire cortex of the rat, were in layer V. Layers V–VI were well differentiated (Fig. 2K–O). These areas did not meet two criteria for eulaminate areas described in primates: equal prominence of superficial II–III and deep V–VI layers and largest pyramids equally distributed in layers III and V. In spite of this, they had overall the best laminar elaboration across the entire cortical quilt of the rat and had well-developed layer IV. We categorized these areas as Eulaminate.

The areas of the rat neocortex according to Zilles (1985) and their corresponding cortical types are summarized in Table 4. We also show cortical types (in grayscale) and areas in coronal maps redrawn from Zilles (1985) (Fig. 3). Agranular, dysgranular, and eulaminate areas in the rat neocortex could be traced across 2 gradients (paraolfactory and parahippocampal) of laminar elaboration. The paraolfactory gradient (dashed arrow in Fig. 3A–E) started adjacent to the anterior olfactory nucleus and the piriform cortex and expanded in dorsal and medial direction. The parahippocampal gradient (solid arrow in Fig. 3A–F) started adjacent to the hippocampal formation; in rostral and middle levels, this gradient expanded from the taenia tecta and indusium griseum in dorsal and lateral direction (Fig. 3A–E). In caudal levels, this gradient expanded from hippocampal areas in dorsal and lateral direction and from entorhinal areas in dorsal and medial direction (Fig. 3F). Both paraolfactory and parahippocampal gradients converged in dorsally located eulaminate areas. Coronal maps in Fig. 3 show that most of the rat cerebral neocortex consists of agranular (52% of all areas) and dysgranular areas (32% of all areas); thus, most of the rat neocortex is mesocortical with fewer eulaminate areas, which constitute about 16% of all cortical areas.

Table 4 Areas in the cerebral cortex of the rat according to Zilles (1985) and their corresponding typesFig. 3

Distribution of cortical types across the rat cortex. a–f, Coronal maps of the rat brain according to Zilles (1985). a is the most rostral map and f is the most caudal map. Allocortical areas are colored in black; agranular mesocortical areas are colored with the darkest gray; dysgranular mesocortical and eulaminate areas are colored in progressively lighter grays. g, Grayscale of cortical types in a–f. Cortical areas are indicated according to Zilles (1985); see Table 4 for abbreviations of neocortical areas. AON anterior olfactory nucleus, cc corpus callosum, IG indusium griseum, LOT lateral olfactory tract, Pir piriform cortex, TT taenia tecta. Solid arrows indicate the parahippocampal gradient of laminar differentiation. Dashed arrows indicate the paraolfactory gradient of laminar differentiation

The Rhesus macaque neocortex is mostly isocortical

We performed cortical type analysis on the micrographs of Nissl-stained sections of the Atlas of the Rhesus macaque (Macaca mulatta) cortex of von Bonin and Bailey (1947). These authors adapted the terminology used by von Economo and Koskinas for areas in the human cortex (von Economo and Koskinas 1925/2008) to Rhesus macaque cortical areas. Accordingly, von Bonin and Bailey identified each cortical area of the Rhesus macaque with a symbol that comprises a Roman capital letter from the initial of the respective lobe (F for frontal, O for occipital, T for temporal, P for parietal, L for limbic cingulate, I for insular) and a calligraphic capital for the sequence of gyri within each lobe. The symbol of some areas also comprises a Greek subscript for the microscopic features of the area (Triarhou 2007).

von Bonin and Bailey (1947) surveyed 58 sites across the cortical quilt in Rhesus macaques and showed micrographs of each site in high quality micrographs. We applied the criteria summarized in Table 3 for cortical type analysis to the 58 micrographs and categorized the areas depicted as agranular, dysgranular, eulaminate I, eulaminate II, eulaminate III, and koniocortex. We will briefly describe the laminar features of each type.

Some areas had overall simple laminar differentiation, lacked layer IV, and had irregular boundary between layers I and II. Neurons in the deep layers (V–VI) of these areas were densely packed and superficial layers (II–III) were sparsely populated; the largest bodies of pyramidal neurons were in layer V. We categorized these areas as Agranular.

Other areas had rudimentary layer IV and their deep layers (V–VI) were slightly more prominent than superficial layers (II–III). The largest bodies of pyramidal neurons were in layer V and layers I and II were separated by a slightly irregular boundary. We categorized these areas as Dysgranular.

A third category of areas had thin but continuous and regular layer IV and deep (V–VI) and superficial (II–III) layers were equally prominent. The largest bodies of pyramidal neurons were equally distributed in layers III and V and the boundary between layers I and II was sharper than in agranular and dysgranular areas. Layers V and VI had sharper differentiation. We categorized these areas as Eulaminate I.

In other areas, layer IV was thicker than in Eulaminate I areas, superficial (II–III) layers were more prominent than deep (V–VI) layers, and the largest bodies of pyramidal neurons were in layer III. Layers V and VI had sharp differentiation and, in some areas, they were divided in sublayers. The boundary between layers I and II was also sharp. We categorized these areas as Eulaminate II.

Finally, some areas had laminar features comparable to Eulaminate II areas but with larger and more prominent pyramidal neurons in layer III. We categorized these areas as Eulaminate III. Only one koniocortical area, the primary visual area, was photographed by von Bonin and Bailey. This area had the thickest layer IV and superficial (II–III) layers were densely populated with small neurons. This was the only area photographed by von Bonin and Bailey that we categorized as Koniocortex.

The areas of the Rhesus macaque neocortex according to von Bonin and Bailey (1947), the plates in which cortical type analysis was performed, and their corresponding cortical types are summarized in Tables 5 and 6. We also show the 58 points on the surface of the cerebral cortex of the Rhesus macaque at which micrographs were taken (Fig. 4A–D) and represent cortical types at each of these points in grayscale (Fig. 4A´–D´). We also represent approximate locations of the other two koniocortical areas (primary auditory and primary somesthetic areas), based on analysis of archival Nissl-stained sections used in previous studies by our group. The distribution of agranular, dysgranular, eulaminate I, eulaminate II, eulaminate III, and koniocortical areas across the cortical quilt of the Rhesus macaque follows two gradients of laminar elaboration, paraolfactory and parahippocampal, previously described in this species (Barbas and Pandya 1987, 1989; Pandya et al. 1988, 2015; Barbas and García-Cabezas 2015; Pandya et al. 2015; Hilgetag et al. 2016; García-Cabezas and Barbas 2017; García-Cabezas et al. 2019). Cortical type maps (Fig. 4A´–D´) show that isocortical (eulaminate) areas account for about 75% of all cortical areas and cover more cortical surface of the Rhesus macaque than mesocortical (agranular and dysgranular) areas. Of note, all agranular areas are in continuity forming a closed ring at the limit of the hemisphere. All dysgranular areas also form a continuum, and so do eulaminate I, and eulaminate II areas. In contrast, eulaminate III areas (1) in the DLPFC; (2) in frontal motor, anterior parietal, and middle part of the temporal lobe; and (3) in the occipital lobe are surrounded by eulaminate II areas. The three koniocortical areas are also each surrounded by eulaminate III areas.

Table 5 Frontal (F) and parietal (P) areas in the cerebral cortex of the Rhesus macaque according to the Atlas of von Bonin and Bailey (1947) and their corresponding typesTable 6 Temporal (T), occipital (O), cingulate (L), and insular (I) areas in the cerebral cortex of the Rhesus macaque according to the Atlas of von Bonin and Bailey (1947) and their corresponding typesFig. 4

Distribution of cortical types across the Rhesus macaque cortex. Medial (a), lateral (b), ventral (c), and dorsal (d) views of the Rhesus macaque cerebral cortex; Roman numerals indicate the location of plates in the Atlas of von Bonin and Bailey (1947). a´–d´, Medial, lateral, ventral, and dorsal views of the Rhesus macaque cerebral cortex; dots in grayscale indicate cortical types for cortical areas photographed in the plates of the Atlas of von Bonin and Bailey (1947). Allocortical areas are colored in black; agranular mesocortical areas are colored with the darkest gray; dysgranular mesocortical and eulaminate areas are colored in progressively lighter grays. Koniocortical areas are colored in white. e, Grayscale of cortical types in a´–d´

Areas in the human ventromedial prefrontal cortex (VMPFC) are mesocortical and isocortical

We have charted cortical types across the human cortex in a previous article (García-Cabezas et al. 2020). Here, we analyzed cortical types and laminar gradations across the human VMPFC (Fig. 5A) according to laminar features of Table 3 to provide direct comparison of cortical gradients of laminar elaboration between primate (human) and rodent (rat) species. These gradients are described in the VMPFC of humans spanning from the allocortical precommissural part of the hippocampal formation (taenia tecta) to eulaminate areas in the frontal pole (Mackey and Petrides 2014; García-Cabezas et al.