Primary visual cortex, or V1, in most mammalian species is clearly distinguishable from other cortical areas by the crisp, stratified appearance of its cortical layers (Campbell, 1905; Brodmann, 1909; Hässler, 1967; Garey, 1971; Billings-Gagliardi et al., 1974; Braak, 1984; Jones, 1984; Le Brun Kemper and Galaburda, 1984). In primates, including humans (Allman and Kaas, 1971; Lund, 1973; Braak, 1976; Fitzpatrick et al., 1983; Preuss et al., 1993; Casagrande and Kaas, 1994; de Sousa et al., 2010; Wong and Kaas, 2010), as well as some related non-primate species (Collins et al., 2005; Wong and Kaas, 2008; Wong and Kaas, 2009b), most of these layers have expanded and segregated into sublayers with distinct functional and connectional properties, making V1 even more conspicuous in comparison to other cortical areas. Yet despite its marked appearance and the clear identification of multiple layers and sublayers across species, the classification of cortical layers within V1 of primates remains a controversial topic (Campbell, 1905; Clark and Sunderland, 1939; von Bonin, 1942; Garey, 1971; Billings-Gagliardi et al., 1974; Henry, 1989; Casagrande and Kaas, 1994; Kaas, 2003; Zilles and Amunts, 2010; Molnár and Belgard, 2012; Nieuwenhuys, 2013). Numerous classification systems have been proposed over the history of V1 research; some based on laminar variations in staining intensity for histological markers such as cytochrome oxidase (CO; Livingstone and Hubel, 1982; Horton, 1984), some based on changes in cell density and morphology through the cortical sheet, as seen with Nissl, myelin, or Golgi stains (Lewis, 1880; Cajal, 1899; Campbell, 1905; Brodmann, 1909; Clark, 1925; von Economo and Koskinas, 1925; von Bonin, 1942; Hässler, 1967; Garey, 1971; Valverde, 1977), and still others based on variations in layer-specific gene and protein expression patterns through the areal extent of V1 (Hevner et al., 2003; Watakabe et al., 2006; Yamamori and Rockland, 2006; Bernard et al., 2012; Bryant et al., 2012; Takahata et al., 2012). Within a single primate species such as macaque monkeys, which are a widely utilized primate model for V1 research, anywhere from eight to twelve layers and sublayers can be identified depending on the criteria for classification (Lund, 1973; Billings-Gagliardi et al., 1974; Felleman and Van Essen, 1991; Casagrande and Kaas, 1994), leading to enormous confusion when comparing studies of layer-specific connections or functions in V1. In less specialized primates such as prosimians or lemurs (Preuss et al., 1993; Preuss and Kaas, 1996; Wong and Kaas, 2009a; Wong et al., 2009), where cortical layers are less distinct, only six to eight layers are commonly identified in anatomical studies, making the comparison of laminar-specific properties across primate species rather difficult. Thus, a common system of laminar classification across primates would greatly benefit the interpretation of past and future studies of V1’s structure and function. Of course this laminar system should be consistent with laminar patterns used in other cortical regions in primates, and for primary visual cortex in non-primate taxa.

The most common discrepancy seen between laminar schemes for V1 in primates involves the superficial and middle cortical layers. Under the famous scheme of Brodmann (1909), layer 3 of V1 is a single layer and layer 4 is divided into three distinct sublayers, termed 4A, 4B, and 4C, with 4C further divided into 4Cα and 4Cβ. This definition, originally based on Nissl-stained sections through the border of V1 with the second visual area (V2) in anthropoid primates and humans, stemmed from the observation that all three middle layers merged into a single layer 4 in V2 and thus, must all be derived from a single layer 4 in ancestral primates. Another dominant laminar scheme, that of Hässler (1967), suggests that layer 3 of V1 is expanded into three layers, termed 3A, 3B, and 3C, and layer 4 only contains two subdivisions, 4A and 4B. When compared directly, Brodmann’s 4A, 4B, 4Cα, and 4Cβ are Hässler’s 3Bβ, 3C, 4A, and 4B, respectively, which leads to significant confusion when discussing the origin and function of layers 3 and 4 in primate V1.

Anatomical studies of V1 in mammalian brains have benefited from recent advances in histology, immunohistochemistry, and in situ hybridization techniques that allow for the selective labeling of individual cell populations based on cell- or layer-specific markers (Hevner et al., 2003; Yamamori and Rockland, 2006; Belgard et al., 2011; Yamamori, 2011; Bernard et al., 2012; Molnár and Belgard, 2012; Takahata et al., 2012). Many markers can be variably expressed across layers, within and across species, so converging types of evidence from different markers is more informative (Molnár and Belgard, 2012), but some of them do distinguish the same layers of V1 across multiple species. Thus, a reconsideration of laminar classification in V1 of primates is timely, given the benefits of reliable layer-specific markers and the vast array of studies on layer-specific V1 functions in current research.

Two markers have gained widespread use in recent research by providing new insights on areal, laminar, and cellular functions within V1. The first is neuronal nuclear antigen (NeuN; Wolf et al., 1996), a DNA-binding protein that is exclusively expressed in neuronal cell bodies across the central nervous system (CNS) and highly conserved across a range of mammalian and non-mammalian species. Immunolabeling for NeuN distinguishes neurons from astrocytes, microglia, and endothelial cells, and identifies morphologically distinct classes of neurons in cortical and subcortical brain structures. Thus, NeuN resembles a Nissl stain but avoids the complications of labeling cells other than neurons. The second is vesicular glutamate transporter 2 (VGLUT2), a neurotransmitter transport protein found on the vesicles of many glutamatergic neurons in the CNS (Aihara et al., 2000; Hisano et al., 2000; Herzog et al., 2001). VGLUT2 is expressed in the presynaptic terminals of most thalamic relay neurons that project to primary sensory cortical areas (Kaneko et al., 2002), and immunolabeling for VGLUT2 reliably distinguishes terminal projections from the lateral geniculate nucleus (LGN) to the granular layer 4 of V1. (Balaram et al., 2011, 2013; Bryant et al., 2012; Garcia-Marin et al., 2012; Rovó et al., 2012). When combined with traditional markers such as CO and Nissl, both NeuN and VGLUT2 can provide additional information about anatomically and functionally distinct layers within V1.

In an attempt to define homologous V1 layers, we labeled NeuN and VGLUT2 through the cortical sheet of seven primate species (chimpanzees, macaque monkeys, owl monkeys, squirrel monkeys, marmoset monkeys, galagos, and mouse lemurs), and one highly visual close relative of primates (tree shrews). By then comparing NeuN- and VGLUT2-labeled sections to adjacent CO- and Nissl-labeled sections, we outlined laminar boundaries through V1 based on changes in the density and morphology of NeuN-labeled cells, as well as the discrete boundaries of VGLUT2-labeled geniculostriate terminations in each species. NeuN labeling revealed highly conserved cortical lamination patterns in V1 of every species examined in this study, and a clear visualization of specialized and homologous V1 layers, particularly at the border of V1 with extrastriate visual area, V2. Additionally, VGLUT2 labeling provided consistent demarcations of layer 4 in V1 across species, and identified specialized geniculostriate terminations to discrete sublayers of layer 3 and layer 6 in some primates. When compared across the primate lineage, both NeuN and VGLUT2 distributions provided evidence for multiple subdivisions of the superficial and deep V1 layers, but only one granular layer 4 with two subdivisions in V1. These layers are consistent with Hässler’s (1967) laminar scheme for V1. The laminar boundaries identified in both NeuN and VGLUT2 preparations are comparable to laminar divisions in other cortical areas across primates, as well as laminar divisions within V1 in non-primates, and may highlight specializations in V1 that reflect ethological and behavioral differences between primate groups.

NeuN and VGLUT2 immunoreactivity was examined in V1 of seven primate and one non-primate species: chimpanzees (Pan troglodytes), macaques (Macaca mulatta), marmosets (Callithrix jacchus), squirrel monkeys (Saimiri sciureus), owl monkeys (Aotus trivirgatus), galagos (Otolemur garnetti), mouse lemurs (Microcebus murinus), and tree shrews (Tupaia glis). At least two, and up to five, individual cases for each species were utilized in this study. All procedures involving mouse lemurs followed the guidelines established by the European Communities Council directives. All procedures involving chimpanzees, macaque monkeys, squirrel monkeys, owl monkeys, marmosets, galagos, and tree shrews followed the animal care and use guidelines established by the National Institutes of Health.

NeuN and VGLUT2 immunohistochemistry identified neuronal cell bodies and glutamatergic terminations respectively, in V1 of all species examined in this study. Compared to traditional Nissl and CO stains (Figure 1), NeuN immunoreactivity (IR) revealed clear laminar boundaries in V1 and V2 (Figures 2 and 3), and could distinguish between granule cells and pyramidal cells in all species (Figure 4). Eleven distinct laminar divisions in V1 were identified across species and designated as layers or sublayers of the six classic neocortical layers (best seen in Figure 2). VGLUT2 IR clearly distinguished V1 layers that receive geniculate inputs, layer 4 and upper layer 6 in all species as well as layer 3Bβ in anthropoid monkeys, and better separated thalamic inputs from intrinsic V1 projections compared to traditional CO stains (Figure 1). Additionally, NeuN and VLGUT2 IR revealed the relative shifts of neocortical layers at the boundary of V1 with V2 (Figures 3 and 5), providing evidence for a single layer 4 with two subdivisions in V1 that continues as a single undifferentiated layer in V2. The thalamorecipient sublayer of layer 3, 3Bβ, which was previously characterized as part of layer 4 in some species (Brodmann, 1909; Lund, 1988), ended abruptly at the V1/V2 border and did not continue as part of layer 4 in V2. The layers adjacent to layer 4 of V1 however, layers 3C and 5A, did continue across the border from V1 to V2 along with the remaining superficial and deep neocortical layers. A detailed analysis of NeuN and VGLUT2 labeling in each species, as discussed below, provided strong evidence for a uniform laminar scheme in V1 across primates, which can also be applied to highly visual non-primates such as tree shrews.

The primary goals of this study were to describe patterns of NeuN and VGLUT2 immunoreactivity in V1 of chimpanzees, Old World monkeys, New World monkeys, prosimians, and tree shrews, in order to identify and compare specialized and homologous laminar divisions across primates and closely related non-primate groups. We find that NeuN immunoreactivity identifies a common pattern of lamination in V1 across all these species, regardless of their phyletic origin in the primate lineage. Similarly, VGLUT2 identifies conserved patterns of geniculostriate terminals in layer 4 of all species, as well as specialized patterns of geniculate terminals in layer 3Bβ of anthropoid monkeys, mouse lemurs, and tree shrews. These findings provide anatomical evidence for common laminar architecture in V1, with multiple subdivisions of the superficial and deep layers, but a single layer 4 with two subdivisions, across primate and non-primate species.

Layers 1 and 2 of V1 were clearly homologous across primates and tree shrews, with almost no neurons in layer 1 and small, densely arrayed neurons in layer 2. The faint VGLUT2 IR in layer 2 of chimpanzees and macaque monkeys likely derives from pulvinar projections to layer 2 of V1 (Balaram et al., 2013; Marion et al., 2013) that may also exist in New World monkeys (Tigges et al., 1981; Stepniewska and Kaas, 1997; Soares et al., 2001; Kaas and Lyon, 2007) and prosimians (Raczkowski and Diamond, 1980; Wong and Kaas, 2010). The three subdivisions of layer 3 seen in NeuN preparations also appear to be highly conserved across primates, and are likely related to similar divisions of layer 3 in non-primates as well. The clustered arrangement of medium and large pyramidal neurons in 3A is morphologically similar across primates and tree shrews, as is the dense distribution of smaller pyramids in layer 3B. Both layers appear expanded in anthropoid primates compared to those in prosimians and tree shrews, and this may reflect the greater number of total neurons in V1 of anthropoid compared to prosimian primates (Collins et al., 2010).

Layer 3Bβ appears to be unique to anthropoid monkeys, given the lack of dense CO reactivity or VGLUT2-positive terminations in layer 3 of chimpanzees (Preuss et al., 1999; Bryant et al., 2012), and humans (Bryant et al., 2012; Garcia-Marin et al., 2012). In apes and humans, thalamic inputs to layer 3B of V1 appear to be reduced or absent altogether (Preuss et al., 1999; Preuss and Coleman, 2002; Bryant et al., 2012; Garcia-Marin et al., 2012). The presence of VGLUT2- positive geniculate terminations in layer 3B of tree shrews and mouse lemurs, however, raises the possibility that parvocellular and koniocellular inputs to this region arose early in the ancestors of the arboreal primate lineage. In nocturnal galagos and owl monkeys, where VGLUT2-positive terminations were more diffuse, parvocellular projections are reduced and layer 3 is dominated by koniocellular inputs instead. Thus, parvocellular inputs to layer 3Bβ appear to have evolved early on in the primate lineage, and they were either refined to a single layer in diurnal, arboreal anthropoid monkeys or they gradually receded in nocturnal or terrestrial primates such as galagos, owl monkeys, chimpanzees and humans.

In all primates, koniocellular geniculate terminations are often coincident with the well-known CO blobs of V1 (Livingstone and Hubel, 1982; Horton, 1984; Lachica and Casagrande, 1992; Casagrande et al., 2007), but the colocalization of VGLUT2 IR with CO blobs is not well documented across primates and not all koniocellular geniculate projections utilize VGLUT2 (Balaram et al., 2011, 2013). In prosimians and tree shrews, VGLUT2 IR does identify patchy blob-like structures in layer 3B of V1 that correspond to CO-dense blobs or patches in the same layer (Wong and Kaas, 2009b, 2010). However, in New World and Old World monkeys, as well as humans, VGLUT2 IR only reveals sparse distributions of geniculate terminals in layer 3B that may be periodic and coincident with the CO blobs (Bryant et al., 2012; Garcia-Marin et al., 2012; Balaram et al., 2013). These differences in VGLUT2-positive terminations between prosimians, anthropoid primates, and humans suggest that geniculate terminations to the blob regions of V1 were more diffuse in ancestral primates, were refined to sparser inputs in anthropoid monkeys, and may be further reduced in present-day chimpanzees and humans.

The termination of VGLUT2-positive projections to 3Bβ at the V1/V2 border in most species does suggest that this layer is unique to V1. Although it has been common to conclude that layers 3Bβ and 4 of V1 merge to form a single layer in V2, a close examination of the border region between V1 and V2 in NeuN and VGLUT2 preparations shows that this is not the case. The slight dorsal shift of all cortical layers at the V1/V2 boundary may give the illusion of merging middle layers from V1 to V2, but only one granular layer continues uninterrupted from V1 to V2 in NeuN preparations of each species examined here. Similarly, only one layer of dense VGLUT2-positive terminations continues from V1 to V2 in each species, the single layer 4 as previously delineated by Hässler (1967). Further evidence for this conclusion comes from the clear identification of layer 3C in V1, which also continues uninterrupted into V2. The presence of large pyramidal neurons instead of small granule cells, as well as the lack of VGLUT2-positive geniculate projections, differentiates this layer from the standard definition of layer 4 - a single layer characterized by small granule cells that receive dense inputs from the LGN (Cowey, 1964; Hubel and Wiesel, 1972; Dräger, 1974; Hubel et al., 1975; Ribak and Peters, 1975; Glendenning et al., 1976; Kaas et al., 1976; LeVay and Gilbert, 1976; Cooper et al., 1979; Spatz, 1979; Towns et al., 1982; Rosa et al., 1996). Neurons in layer 3C instead receive intrinsic projections from layer 4, similar to 3A and 3B (Levitt et al., 1996), and are known to project extrinsically to MT (Van Essen et al., 1981; Maunsell and Van Essen, 1983; Shipp and Zeki, 1989) and the thick CO bands of V2 (Federer et al., 2009; Sincich et al., 2010), just as 3A and 3B project extrinsically to V2 and other visual areas (DeYoe and Van Essen, 1985; Rockland, 1992; Sincich and Horton, 2005; Sincich et al., 2010; Federer et al., 2013). Analyses of neuronal density in layers 3C and 4 of V1 and V2 showed that both layers are quite distinct from one another, regardless of the area in question. Other studies on laminar specific gene expression in macaque monkeys show that neurons in Hässler’s 3Bβ and 3C are genetically more similar to layer 3 neurons than layer 4 neurons across multiple areas in the neocortex. Thus, Brodmann’s delineation of this layer as part of layer 4 seems inappropriate given the evidence discussed above. Hässler’s (1967) laminar scheme for V1 however, allows for better comparisons of V1 morphology and function across all primates, as well as tree shrews and other visual mammals.

Layer 4 across primates and tree shrews could be identified by small, densely packed granule cells in NeuN preparations as well as dense VGLUT2 IR in both subdivisions, 4A and 4B. Both sublayers merged with a single layer 4 containing similarly sized granule cells but more moderate VGLUT2 labeling in V2. This layer 4, which corresponds to Hässler’s layer 4 of V1, is directly comparable to layer 4 of V1 in rodents (Krieg, 1946; Peters and Feldman, 1976; Caviness and Frost, 1980; Frost and Caviness, 1980; Simmons et al., 1982; Peters, 1985), lagomorphs (Swadlow and Weyand, 1981; Towns et al., 1982; Weyand and Swadlow, 1986), and lemurs (Clark, 1931; Zilles et al., 1979; Preuss and Kaas, 1996), tarsiers (Collins et al., 2005), and great apes and humans (Yoshioka and Hendry, 1995; Preuss et al., 1999; Preuss and Coleman, 2002; Zilles et al., 2002; Preuss, 2007; Bryant et al., 2012). By excluding the variable sublayer 3Bβ with koniocellular LGN inputs as well as the pyramidal sublayer 3C with no thalamic inputs, layer 4 can be consistently identified as a single granular layer across mammals, thus creating a more unified scheme of V1 lamination across primate and non-primate species.

Layer 5 of V1 surprisingly revealed two consistent subdivisions across species, which have only been previously reported in macaque monkeys (Lund, 1973, 1987; Lund et al., 1988; Balaram et al., 2013). The superficial sublayer, 5A, consists largely of interneurons that project intrinsically in V1, and are identifiable in the present results by comparatively weak NeuN labeling against the surrounding V1 layers. In contrast, the deep sublayer 5B could be differentiated by stronger NeuN IR and sparser distributions of medium and large pyramidal neurons, which are known to project subcortically to the pulvinar and superior colliculus in most primate species (Spatz et al., 1970; Lund et al., 1975; Graham et al., 1979; Raczkowski and Diamond, 1980; Baldwin and Kaas, 2012; Baldwin et al., 2013). The consistent identification of two sublayers in layer 5 across primates and tree shrews suggests that these patterns of intrinsic and extrinsic projections are highly conserved in these closely related species. Further studies in non-primate species will reveal if these sublayers are fundamental to V1 organization across other mammals as well.

Similarly, layer 6 of V1 has been recently subdivided into two layers across multiple primate and non-primate species, largely based on distinct differences in gene expression patterns between cells in the upper portion, 6A, and cells in the lower portion, 6B (Yamamori and Rockland, 2006; Hevner, 2007; Belgard et al., 2011; Bernard et al., 2012). The present results also reveal differences in cellular density between the two layers, as well as the presence of geniculate terminations throughout 6A but not 6B. Prior reports of V1 lamination largely consider layer 6 as a single layer that blends into the ventral white matter (Lund, 1973; Lund et al., 1988; Casagrande and Kaas, 1994), but subtle differences between upper and lower tier cells in this layer have been considered (Levitt et al., 1996; Yamamori and Rockland, 2006; Watakabe et al., 2012, 2006). Our results, in combination with recent studies of layer- and cellular-specific gene expression, suggest that layer 6 is made up of two functional subdivisions; a superficial sublayer with sparse geniculate inputs as well as intrinsic and extrinsic visual projections, and a deep sublayer with separate inputs and functions. Both sublayers are consistently present across primates and tree shrews, and are similar to those seen in rodent species as well (Usrey and Fitzpatrick, 1996; Hevner, 2007; Belgard et al., 2011; Bernard et al., 2012).

The results presented here contribute to a unified understanding of laminar organization in V1 across multiple primates as well as a closely related non-primate, tree shrews. They highlight the inconsistencies of Brodmann’s laminar scheme when comparing V1 architecture across multiple species and augment the growing number of reports that promote the use of Hässler’s laminar scheme in V1 research (Spatz et al., 1970; Spatz, 1977; Tigges and Tigges, 1981; Tigges et al., 1982; Maunsell and Van Essen, 1983; Gould et al., 1987; Allman and McGuinness, 1988; Henry, 1989; Casagrande and Kaas, 1994). In conjunction with reports of physiological and connectional differences between V1 layers, they will drive further research on the homologous and specialized laminar features of primate V1.

Conception, data acquisition, analysis, interpretation, drafting of manuscript, final review and approval, accountability – Pooja Balaram and Jon H. Kaas.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.