Somato‐dendritic morphology and axon origin site specify von Economo neurons as a subclass of modified pyramidal neurons in the human anterior cingulate cortex

Abstract

Von Economo neurons (VENs) are modified pyramidal neurons characterized by an extremely elongated rod‐shaped soma. They are abundant in layer V of the anterior cingulate cortex (ACC) and fronto‐insular cortex (FI) of the human brain, and have long been described as a human‐specific neuron type. Recently, VENs have been reported in the ACC of apes and the FI of macaque monkeys. The first description of the somato‐dendritic morphology of VENs in the FI by Cajal in 1899 (Textura del Sistema Nervioso del Hombre y de los Vertebrados, Tomo II. Madrid: Nicolas Moya) strongly suggested that they were a unique neuron subtype with specific morphological features. It is surprising that a clarification of this extremely important observation has not yet been attempted, especially as possible misidentification of other oval or fusiform cells as VENs has become relevant in many recently published studies. Here, we analyzed sections of Brodmann area 24 (ACC) stained with rapid Golgi and Golgi‐Cox in five adult human specimens, and confirmed Cajal's observations. In addition, we established a comprehensive morphological description of VENs. VENs have a distinct somato‐dendritic morphology that allows their clear distinction from other modified pyramidal neurons. We established that VENs have a perpendicularly oriented, stick‐shaped core part consisting of the cell body and two thick extensions – an apical and basal stem. The perpendicular length of the core part was 150–250 μm and the thickness was 10–21 μm. The core part was characterized by a lack of clear demarcation between the cell body and the two extensions. Numerous thin, spiny and horizontally oriented side dendrites arose from the cell body. The basal extension of the core part typically ended by giving numerous smaller dendrites with a brush‐like branching pattern. The apical extension had a topology typical for apical dendrites of pyramidal neurons. The dendrites arising from the core part had a high dendritic spine density. The most distinct feature of VENs was the distant origin site of the axon, which arose from the ending of the basal extension, often having a common origin with a dendrite. Quantitative analysis found that VENs could be divided into two groups based on total dendritic length – small VENs with a peak total dendritic length of 1500–2500 μm and large VENs with a peak total dendritic length of 5000–6000 μm. Comparative morphological analysis of VENs and other oval and fusiform modified pyramidal neurons showed that on Nissl sections small VENs might be difficult to identify, and that oval and fusiform neurons could be misidentified as VENs. Our analysis of Golgi slides of Brodmann area 9 from a total of 32 adult human subjects revealed only one cell resembling VEN morphology. Thus, our data show that the numerous recent reports on the presence of VENs in non‐primates in other layers and regions of the cortex need further confirmation by showing the dendritic and axonal morphology of these cells. In conclusion, our study provides a foundation for further comprehensive morphological and functional studies on VENs between different species.

Introduction

Von Economo neurons (VENs) were first observed by Betz in the human anterior cingulate cortex (ACC) as remarkably large spindle‐shaped neurons (Betz, 1881). This observation was followed by Von Economo's detailed description of their elongated soma shape and regional distribution using Nissl and Bielschowsky silver staining methods (von Economo, 1926). By systematically analyzing all cortical areas of the human brain, von Economo found VENs only in the ACC and fronto‐insular cortex (FI). Von Economo first noted VENs in ‘diseased brains’, and initially concluded that, due to their peculiar shape and size, they represented a pathological alteration of pyramidal neurons (von Economo, 1918). He later realized that these cells were a unique neuron subtype with specific regional distribution, and referred to them as ‘rod cells’ or ‘corkscrew cells’ (von Economo, 1926; Seeley et al. 2012). Cajal gave the first description of VEN morphology using Golgi staining. He demonstrated VENs in the FI of a 1‐month‐old girl, and stated that these neurons had a distinctive dendritic morphology (Cajal, 1899, 1995). This unique morphology clearly differentiates VENs from other types of modified pyramidal neurons. Modified pyramidal neurons represent a spectrum that includes a wide variety of morphological forms, with some forms deviating only slightly from typical pyramidal cells, and others having a bipolar and spindle‐like morphology commonly found in layers V and VI of the entire human cerebral cortex (Braak & Braak, 1985).

After these initial descriptions, VENs came back into the focus of research when it was reported that their number in ACC decreased by 60% in Alzheimer's disease (Nimchinsky et al. 1995). Alterations of VENs have also been described in various other neuropathological and neuropsychiatric disorders, such as autism spectrum disorder (Allman et al. 2005; Santos et al. 2011; Uppal et al. 2014), amyotrophic lateral sclerosis (Braak & Del Tredici, 2018), schizophrenia (Brüne et al. 2010, 2011; Krause et al. 2017), Parkinson's disease (Fathy et al. 2018), Alzheimer's disease (Gefen et al. 2018), behavioral variant of frontotemporal dementia (Seeley, 2008; Kim et al. 2012; Santillo et al. 2013; Santillo & Englund, 2014; Yang et al. 2017), agenesis of the corpus callosum (Kaufman et al. 2008) and alcoholism (Senatorov et al. 2015). In addition, a higher density of VENs in the ACC was proposed to be related with a higher memory capacity in advanced old age (Gefen et al. 2015).

A comparative study found VENs in the ACC of great apes (Nimchinsky et al. 1999). In this study, VENs were analyzed using Nissl staining, and it was established that the density and size of VENs increased with the evolutionary complexity of the analyzed primate species (Nimchinsky et al. 1999). VENs were shown to form clusters only in humans and bonobos, while in other primates they were found as single cells. Furthermore, VENs were abundant in bonobos and chimpanzees, less abundant, but still frequently observed in gorillas and rarely present in orangutans (Nimchinsky et al. 1999). In this study, VENs were not found in the ACC of a number of different primate species, including the gibbon, several New World (Ceboidea) and Old World (Cercopithecidae) monkey species and prosimians. This suggested that VENs appeared in the ACC during hominid (Hominidae, great apes) evolution, and that they become more numerous and possibly more complex with an increase in encephalization level (Nimchinsky et al. 1999). Notwithstanding the initially described lack of VENs in the ACC of prosimians, a more recent study confirmed their presence in the FI of the macaque monkey using Golgi staining (Evrard et al. 2012). This suggests that VENs may have evolved in the FI before the division of Old World monkeys and hominids, and that they appear in the ACC only in the hominid line (Allman et al. 2010; Evrard et al. 2012).

Other studies, however, proposed that VENs might also be present in non‐primate species. Spindle cells resembling VENs were described on Nissl‐stained sections in elephants (Hakeem et al. 2009) and cetaceans (Butti et al. 2009). It was then proposed that VENs could have separately evolved in phylogenetically distant species that have a large absolute brain size (> 300 g) and a sophisticated social organization in common (Nimchinsky et al. 1999; Hof & van der Gucht, 2007; Butti et al. 2009; Hakeem et al. 2009; Allman et al. 2010). In subsequent studies, similar cells were identified on Nissl‐stained sections in the insula of the manatee (Butti & Hof, 2010), pygmy hippopotamus (Butti et al. 2014), artiodactyls and perissodactyls (Raghanti et al. 2015). Using a stereological approach on Nissl‐stained sections taken from the frontal pole, ACC, FI, and occipital pole of the bowhead whale, cow, sheep, deer, horse, pig, rock hyrax and human, spindle cells resembling VENs were found in all the mentioned species, not only in layer V, but in layers II and III of several species as well (Raghanti et al. 2015).

Studies on the molecular characteristics of VENs yielded no specific marker for their identification, though it was established that VENs express biochemical and transcription factors typical for glutamatergic projection neurons, such as SMI‐32 (Nimchinsky et al. 1995), MAP2 (Fajardo et al. 2008), FEZF2 and CTIP2 (Cobos & Seeley, 2015), and VMAT2, GABRQ and ADRA1A (Dijkstra et al. 2018). Their particularly high immunoreactivity for ATF3, IL4Rα and NMB suggests that VENs are involved in pain processing (Stimpson et al. 2011).

The aforementioned findings demonstrate that the morphological and molecular characterization of VENs is yet to be completed. Therefore, the aim of this study is to establish a comprehensive description of the morphological features of VENs in the human ACC using Golgi, Nissl and NeuN staining.

Materials and methods

Brain tissue samples

The brain samples of nine human male subjects (18–59 years) with post mortem delays between 6 and 10 h were processed using Golgi and Nissl methods, and immunohistochemistry (Table 1).

Table 1.

Origin of human brain tissue used in this study

Subject	Age (years)	Post mortem delay (h)	Cause of death	Staining methods
CO 300	50	6	Polytrauma	Golgi Nissl
CO 301	37	6	Polytrauma	Golgi Nissl
CO 302	59	10	Sudden cardiac death	Golgi Nissl
CO 310	18	6	Polytrauma	Golgi Nissl
CO 311	40	6	Sudden cardiac death	Golgi Nissl
CO 384	46	6	Sudden cardiac death	NeuN Nissl
CO 385	29	8	Polytrauma	NeuN Nissl SMI‐32/MAP2
CO 386	37	6	Sudden cardiac death	NeuN Nissl
CO 387	44	6	Methadone/benzodiazepine overdose	NeuN Nissl

None of the cases had medical histories of neurological or psychiatric disorders, and autopsy reports found no neuropathological deviations in the brains. All relevant medical history was obtained from both medical records and autopsy reports. All analyzed subjects died without preagonal state, and the post mortem delay represents the maximum interval in which neuron death took place. The brain tissue is a part of the Zagreb Neuroembryological Collection (Kostovic et al. 1991; Judaš et al. 2011). The brains were obtained with the approval of the Ethics Committee of Zagreb University School of Medicine. The procedure is currently regulated by Ethics Committee approval number 380‐59‐10106‐14‐55/152 from 1 July 2014. The information on the subject's identity and history is stored in secure records, and the brain tissue is given a code indicating only the age of the subject.

Golgi staining

Two Golgi staining methods were used in this study – rapid Golgi and Golgi‐Cox (Zeba et al. 2008; Petanjek et al. 2011; Kang et al. 2017). The Golgi methods were used alternately on adjacent blocks from a total of five human brains (Table 1). The tissue blocks had a volume of 1 cm³ and were taken from the anterior part of the cingulate gyrus of the left hemisphere, dorsal to the genu of the corpus callosum, corresponding to Brodmann area 24.

One block from each brain was used for the classical chrome‐osmium rapid Golgi staining method and was first placed in 4% paraformaldehyde for 12–18 h. It was then immersed in rapid Golgi solution consisting of 0.3% osmium tetroxide and 3% potassium dichromate for 7 days. The solution was then replaced with 1% silver nitrate in which the tissue was immersed for 2 days. These steps were performed in a darkened room. The tissue blocks were then dehydrated in an ethanol cascade (70%, 96%, absolute ethanol) and put in alcohol‐ether (1 : 1). After dehydration, the tissue segments were rapidly embedded in celloidin and serially cut on a microtome into coronal sections at a thickness of 200 μm, after which mounting of the sections was performed.

A second tissue block from each brain was used for the Golgi‐Cox staining method, and was immediately immersed in Golgi‐Cox solution (0.17% potassium chromate, 0.2% potassium dichromate and 0.2% mercuric chloride) for 3 weeks. The solution was replaced once after 24 h. Afterwards, the tissue segments were dehydrated, embedded in celloidin and serially cut on a microtome. This part of the procedure was the same as for rapid Golgi staining. The sections were then immersed in 20% ammonium hydroxide for 5 min and transferred into 15% ammonium hydroxide for 25 min. After rinsing they were further processed through 1% thiosulfate for 7 min followed by mounting of the sections (Bicanic et al. 2017).

The mounting procedure, for both Golgi methods, included brief dehydration of the sections (50%, 70%, 96% ethanol, butanol‐ethanol) and placement into Histoclear (National Diagnostics, USA) after which they were mounted with Histomount (National Diagnostics).

The success of Golgi impregnation was determined according to relevant literature (Williams et al. 1978; Braak & Braak, 1985; Petanjek et al. 2011, 2019). No staining artifacts due to post mortem delay were detected in the analyzed cases.

Immunohistochemistry – single labeling (NeuN)

NeuN single labeling was performed on a total of four human brains (Table 1). One tissue block from each brain was obtained from the left hemisphere. The cut went perpendicular to the axis of the cingulate gyrus and extended to the dorsal part of the frontal lobe. The block encompassed the anterior part of the cingulate gyrus, dorsal to the genu of the corpus callosum as well as the superior frontal gyrus. The block included Brodmann areas 24, 32 and 9 (Rajkowska & Goldman‐Rakic, 1995; Vogt et al. 1995; Triarhou, 2007), which were identified and delineated based on their cytoarchitectonic features (Petrides & Pandya, 1999; Ongür et al. 2003; Petrides et al. 2012; Vogt et al. 2013).

The tissue was first fixed by immersion in 4% paraformaldehyde for 24 h, then cryoprotected in solutions of increasing sucrose concentrations (12, 16 and 18%) in phosphate‐buffered saline, and finally frozen and cut on a cryostat into coronal sections at a thickness of 60 μm. Immunohistochemistry was performed according to standard protocols (Hsu et al. 1981) on free‐floating sections. Following rehydration in phosphate‐buffered saline (PBS), all sections were pretreated in 1% H₂O₂ for quenching of endogenous peroxidase, and washed for 30 min in PBS and 2 × 30 min in KPBS. To prevent non‐specific background staining, sections were immersed in blocking solution (3% horse/goat serum albumin and 0.3% Triton X‐100 in PBS; all from Sigma, USA) for 1 h at room temperature (RT), and then incubated overnight at 4 °C with a primary antibody: monoclonal mouse anti‐NeuN (1 : 4000; Chemicon International, USA; Cat. No. MAB377). Following a 3 × 30 min wash in KPBS, a biotinylated anti‐rabbit/mouse secondary antibody (Vectastain ABC kit, Vector Laboratories, USA) was applied (1 : 200 in blocking solution) for 1 h at RT. The sections were then washed for 3 × 30 min and immersed in Avidin‐Biotin‐Peroxidase Complex (Vectastain ABC kit, Vector Laboratories) for 1 h at RT. The peroxidase activity was visualized using Ni‐3,3‐diaminobenzidine (Sigma). Sections were then dehydrated in graded series of alcohol, cleared in xylene, and coverslipped using Histomount (National Diagnostics). Negative controls were included in all experiments: by either replacing the primary antibody with blocking solution or preimmune goat or horse serum, or by omitting the secondary antibody or replacing it with another secondary antibody.

Immunohistofluorescence – double‐labeling (SMI‐32/NeuN and MAP2/NeuN)

Double‐labeling was used to verify previous findings that VENs express SMI‐32 (Nimchinsky et al. 1995) and MAP2 (Fajardo et al. 2008). A small tissue block of 1 cm³ in volume from one human brain (Table 1) was taken from the anterior part of the cingulate gyrus of the left hemisphere, corresponding to Brodmann area 24. The block was processed by the same protocol as those used for single labeling, and immunohistofluorescence was performed according to standard protocols on free‐floating sections. Following rehydration in PBS, all sections were washed for 30 min in PBS and 2 × 30 min in KPBS. To prevent non‐specific background staining, sections were immersed in blocking solution (3% donkey serum albumin and 0.3% Triton X‐100 in PBS; all from Sigma) for 1 h at RT, and then incubated overnight at 4 °C with primary antibodies: monoclonal mouse anti‐SMI‐32 (1 : 3000; BioLegend, USA; Cat. No. 801701), monoclonal anti‐MAP2 (1 : 3000; Sigma; Cat. No. M4403) and polyclonal rabbit anti‐NeuN (1 : 1000; Abcam, UK; Cat. No. ab104225).

Following a 3 × 30 min wash in KPBS, appropriate secondary antibodies were added: Alexa 488 (Invitrogen, USA), donkey anti‐rabbit (1 : 100; Vectastain ABC kit, USA) and Cy3 donkey anti‐mouse (1 : 100; Vectastain ABC kit, USA) diluted in 0.2 m KPBS + normal donkey serum (1 : 30; Sigma) for 1 h at RT. The sections were then washed for 3 × 30 min and immersed in TrueBlack^® Lipofuscin Autofluorescence Quencher (Biotium, USA) for 30 s to eliminate lipofuscin autofluorescence and reduce autofluorescence from other sources. The sections were then mounted on Superfrost slides (Thermo Fisher Scientific, USA) and dried overnight. After being rapidly rehydrated in double‐distilled H₂O (2–3 s), the sections were coverslipped in Fluoromount G (Invitrogen, USA).

The double‐labeled sections were imaged with a confocal microscope (LSM 510 META; Zeiss, Germany).

Nissl staining

Nissl staining was performed on sections of the cingulate gyrus adjacent to those on which NeuN or Golgi staining was performed (Table 1). Two types of tissue processing were used before Nissl staining – cryoprotection and paraffin embedding.

The cryoprotection protocol was as described for immunohistochemistry. Every 10^th section was put through a series of decreasing ethanol solutions (96, 80 and 70%). After immersion in 96% ethanol, the sections were left in a chloroform solution for better fixation. This was followed by immersion in 80% and 70% ethanol solutions, and quick rehydration in distilled water. Then, the sections were treated with 0.5% Cresyl violet (Sigma) solution for 3 min, followed by differentiation in acetic acid in 70% ethanol for 5 s, and a series of 96% and 100% ethanol until the staining was optimal.

The tissue for paraffin embedding was first dehydrated in an ethanol cascade (70%, 96%, absolute ethanol). It was then embedded in paraffin kept in toluene. After embedding, the blocks were cut on a microtome into coronal sections at a thickness of 20 μm. Then, the sections were treated with 0.5% Cresyl violet (Sigma) solution for 3 min, followed by differentiation in acetic acid in 70% ethanol for 5 s, and a series of 96% and 100% ethanol until the staining was optimal.

Analysis of archive material

Analysis of human Golgi tissue from Zagreb Neuroembryological Collection

Golgi slides from the Zagreb Neuroembryological Collection of a total of 32 subjects ranging in age from newborn to 91 years old were used to verify the presence of VENs in other areas of the human cerebral cortex (Kostovic et al. 1991; Judaš et al. 2011; Petanjek et al. 2011, 2019). The following cortical areas were analyzed: primary motor cortex (Brodmann area 4), primary somatosensory cortex (Brodmann areas 3, 1 and 2), angular gyrus (Brodmann area 39), visual cortex (Brodmann areas 17, 18 and 19), dorsolateral prefrontal cortex (Brodmann areas 9 and 46) and Broca's area (Brodmann areas 44 and 45). The tissue was stained by rapid Golgi and Golgi‐Cox according to the protocol previously described.

Analysis of macaque monkey tissue from archive material

NeuN and Nissl slides of the ACC of three adult (10–12 years old) rhesus macaque monkeys (Macaca mulatta) used in previous studies (Džaja et al. 2014; Sedmak et al. 2017) were reevaluated to determine the presence of cells with somatic morphology resembling VENs. The slides were obtained from 1.5‐cm‐thick tissue blocks of the cingulate gyrus and contained coronal sections of both hemispheres, which passed only through the middle part of Brodmann area 24. Brodmann area 24 was delineated using cytoarchitectonic criteria from relevant literature (Vogt et al. 1987, 2013; Carmichael & Price, 1994; Petrides et al. 2012). The Nissl and NeuN staining used on this tissue was done according to the protocol previously described for human tissue.

Neurolucida three‐dimensional reconstructions

On Golgi sections, neurons of the ACC were selected for three‐dimensional reconstruction with Neurolucida 4.1 (MBF Bioscience, USA) software using a 60 × air objective of an Olympus BX50 microscope connected to a Hitachi 3CCD color video camera HV‐C20M. We randomly selected midsection neurons whose primary dendrites were not cut on the section edge. The qualitative descriptions of the shape and location of VENs were based on a systematic analysis of numerous VENs on Golgi sections.

In the Neurolucida software, the apical dendrite, cell body, basal dendrites and side (oblique) dendrites were traced separately. For dendrite analysis, a centrifugal branch order was used (Uylings & van Pelt, 2002). Centrifugal ordering counts the distance from the root in terms of the number of segments traversed with the advantage that missing portions of the tree do not result in incorrect numbering of the known segments. The Marker tool was used to define the positions of dendritic spines. The analysis of the three‐dimensional reconstructions was performed in Neurolucida Explorer 10 (MBF Bioscience). The Branched Structure Analysis function was used to obtain relevant data for the cell bodies, dendrites and markers. This function calculates the length, area, volume, quantity and complexity of the cell body and dendrites (Uylings & van Pelt, 2002).

For cell bodies, the parameters Area and Aspect Ratio were analyzed. Area is defined as the surface within the boundary of the cell body in μm². Aspect Ratio is defined as the ratio between the minimum and maximum diameters of the contour of the cell body. The Aspect Ratio shows the degree of flatness of the cell body, and its value falls within the range [0, 1]. A lower Aspect Ratio corresponds to a thinner, more elongated cell body.

For dendrites, Segment Analysis (a part of the Branched Structure Analysis function) was used to determine the number of segments, the length of intermediate and terminal segments, and the total dendritic length.

For markers, the Individual Marker and Marker Totals analyses were used to determine the number of markers (both total and per segment), position of the markers on the branch (Distance along branch and Distance from root) and relative to each other (Distance to previous). The final average marker density was calculated as an inverse function of the mean distance between two markers. Using this calculation, only dendritic segments on which markers were present were considered. This ensured that the marker density represented the real spine density on dendritic segments with spines.

Quantitative data analysis

The quantitative data in this manuscript are presented as mean ± standard deviation for normally distributed data and median with range for non‐normally distributed data. Correlation was used to determine the association between cell body Area and total dendritic length with Pearson's correlation coefficient used as a measure for the strength of the association (Mukaka, 2012). The t‐test was used to determine if the correlation coefficient is significantly different from zero, with a P‐value of less than 0.05 considered as statistically significant.

Results

VENs can be identified as a distinct class of modified pyramidal neurons using Golgi staining

Here we present a qualitative description of VENs in the human ACC (Brodmann area 24) supported by quantitative data from a pool of three‐dimensionally reconstructed neurons impregnated by either rapid Golgi (13 neurons) or Golgi‐Cox methods (11 neurons). Both methods were used in order to acquire a more comprehensive overview of VEN morphology. On Golgi sections, VENs were located in the deep part of layer V and the upper part of layer VI (Fig. 1), which corresponds to observations on Nissl sections (Fig. S1). At low magnification (Fig. 2), VENs are rod‐ or stick‐like cells with a main shaft oriented perpendicular to the pial surface and many thin side branches arising from it at a sharp angle. VENs on Golgi sections can be described as antenna‐ or feather‐like cells. The cell body was elongated and gradually, without clear demarcation, continued into two thick extensions (Fig. 2A), one ascending toward the pia mater (‘apical’ extension) and another descending toward the white mater (‘basal’ extension). Due to the thickness of these extensions and a lack of a clear demarcation from the cell body, we defined them as the apical and basal stem. Additionally, we defined the ‘core part’ of a VEN as consisting of the cell body, the apical stem and the basal stem. On Golgi‐Cox sections (Fig. 2B), most VENs had a more spindle‐shaped cell body, and the point of demarcation between the soma and the stems was more apparent.

Figure 1.

Low‐power microphotograph of a rapid Golgi‐impregnated section of Brodmann area 24 in the anterior cingulate cortex (ACC) of an adult human specimen (CO 300). Layers III and Va are comprised of medium‐sized typical pyramidal neurons, whereas in layers Vb and VI, both typical and modified pyramidal cells can be seen. The two main forms of modified pyramidal neurons have either an oval (marked with an ‘o’) or a triangular (marked with a ‘t’) soma with two prominent dendrites perpendicular to the pia mater, travelling in opposite directions. Modified pyramidal neurons also have a medium‐sized soma. In contrast, even on low‐power microphotographs, von Economo neurons (VENs; marked with an asterisk) are more prominent and their soma is more rod‐shaped when compared with other types of modified pyramidal neurons. Note that in layers Vb and VI, medium‐sized pyramidal neurons are frequently seen (marked with arrows), and are characterized by a well‐bifurcated dendritic tree and an elongated conical soma, which gradually continues into the apical dendrite.

Figure 2.

(A) Rapid Golgi‐ and (B) Golgi‐Cox‐impregnated von Economo neurons (VENs) three‐dimensionally reconstructed with Neurolucida and shown in the X–Y plane. VENs have a thick, perpendicularly oriented ‘core part’ and numerous thin side branches that travel horizontally. Note that, on rapid Golgi staining, most VENs are rod‐ or spindle‐shaped, and on Golgi‐Cox most of them are more oval‐shaped. There are also large differences in the number of side branches as well as in their length. In some VENs, the side dendrites extend less than 200 μm in the horizontal plane, whereas in others their horizontal reach exceeds 800 μm (0.8 mm). Note also that the highest concentration of thin branches arises from or around the ending of the basal stem. These branches also tend to travel in a horizontal direction.

Apical stem

The apical stem on VENs was very thick with a diameter of 7–12 μm in the proximal 50–150 μm of the stem's length (Fig. 3A). In one example, the apical stem had a diameter of 8 μm at its origin, and 6 μm after 400 μm of length, showing a very gradual decrease in thickness. The high thickness of the origin of the apical stem and a very gradual decrease in thickness along its course were the key features distinguishing the apical dendrites of VENs from those of other pyramidal neurons, not the apical dendrite topology. The topology of VEN apical dendrites was similar to that of medium‐sized and large pyramidal neurons of layer V (Fig. 3B).

Figure 3.

(A) An enlarged part of the microphotograph shown in Fig. 1 (rapid Golgi impregnation), portraying the cortex between the deep part of layer III and the upper part of layer VI (subject: CO 300). The panel is a composition of three microphotographs of the same slide taken at different section depths. Von Economo neurons (VENs; marked with asterisks) are easily recognizable by their large rod‐shaped ‘core part’ oriented perpendicularly to the pia mater. The thickness of the basal stem is at least half of the maximum soma thickness. The thickness of the core part is at least 5 μm along a length of 150 μm, which means that its thickness is at least one‐third of the maximum soma thickness. In some VENs, this feature is present for more than 400 μm in length. (B) An enlarged, triangular neuron, with two prominent dendrites oriented perpendicularly to the pia mater and located at the border of layers Vb and VI, is shown (indicated by the arrow in A). The arrow in (B) indicates the origin of the axon, which arises from the cell body. (C) An enlarged part of a VEN shown in (D) (corresponding to the region indicated by the arrow in D). The single arrow indicates the axon, while the two arrows are pointing to its origin. (D) A Golgi‐Cox‐impregnated VEN is shown featuring a smaller soma and fewer side branches than VENs impregnated with rapid Golgi.

Basal stem

The basal stem of VENs was 50–60 μm long and 6–10 μm thick. The transition from the cell body to the basal stem was even more gradual than on the apical side (Figs 3C,D, 4, and 5A,B). Sometimes the basal stem diameter reached up to 13 μm with no observable decrease in width toward the end of its course (Figs 4 and 5A). In a vast majority of VENs, the basal stem ended with a unique branching pattern characterized by an abrupt decrease in thickness of the dendrite branches of the next order. In most VENs, one–three thicker dendrites (2–4 μm in diameter) arose from the basal stem and bifurcated once or twice. Their terminal branches were long and decreased in thickness more abruptly than the branches of the previous order. They changed orientation and travelled parallel with the pial surface. Up to 10 dendrites arose from the basal stem ending as well as from the lateral sides of its most distal part. All these features gave the basal stem of VENs a picturesque appearance of a ‘birch tree’. In some cases, the basal stem divided into two thicker dendrites (Fig. 5B), thereby having a branching topology similar to a typical basal dendrite. In our analysis, we found a single example of a VEN in the dorsolateral prefrontal cortex (Brodmann area 9). Here the basal stem was longer and bifurcated at the ending into several dendrites with a topology resembling that of typical basal dendrites (Fig. 5A).

Figure 4.

High‐power microphotograph of one rapid Golgi‐impregnated von Economo neuron (VEN; subject: CO 301). The main part of the figure and most of the panels are a composition of several microphotographs of the same slide taken at different section depths. The smaller panels represent the enlarged parts of the dendrites of the VEN shown here. The real position of the upper right panel (framed by the larger dashed square) is indicated by the smaller dashed square. The magnification is the same for all inserted panels and is indicated by the 10 μm scale bar on the panel in the lower right corner. Note the lack of a clear demarcation between the soma and the stems. On one side, in the middle of the cell body, there is a belly‐like prominence, most likely corresponding to the position of the cell nucleus. The maximum thickness of the cell is just slightly greater than the thickness of the cell nucleus. On the distal dendrites, spines are densely packed, have a long stalk (up to 3 μm) and a small terminal bulb, i.e. they correspond to the long mushroom type of spines. The ending of the basal stem is indicated by an asterisk. The basal stem ends by branching into two thick dendrites. The axon origin is marked with two black arrows on the panel in the lower right corner.

Figure 5.

(A) Microphotograph of a rapid Golgi‐impregnated neuron in Brodmann area 9 with features corresponding to von Economo neurons (VENs; subject: CO 185, archive material, 18 years old). The neuron presented here was the only such neuron found in this region, and it has a slenderer appearance than the VENs found in Brodmann area 24. The basal stem is twice as long (about 150 μm) as the longest found on VENs in Brodmann area 24, while the soma (indicated by an asterisk) and the basal stem are rather thin, comparable to the thinnest VENs of Brodmann area 24. Several primary dendrites arise from the distal point of the basal stem and have a bifurcation pattern similar to the topology of the basal dendritic tree on typical pyramidal neurons. The axon (marked with white arrowheads) arises from one of these dendrites (marked with a white arrow). (a) An enlarged portion of (A), showing the dendrite from which the axon (marked with white arrowheads) arises. The position of the axon origin is indicated by a white arrow. (B) Microphotograph of a Golgi‐Cox‐impregnated VEN in Brodmann area 24 shown at the same magnification as in (A) (subject: CO 310). This VEN has a more oval cell body. The basal stem splits after 70 μm into two thick branches, one of which is longer (marked with black arrows) and continues in the direction of the basal stem. The axon (marked with black arrowheads) also arises from this dendrite, about 30 μm distal to the bifurcation of the basal stem. Note that the other dendrite (marked with black double arrows) that arises from the bifurcation of the basal stem is shorter and has a more irregular (stumpy) form. Also note a side knurl (marked with black double arrowheads) from which several thin dendrites arise. (b) Enlarged view of the bifurcation of the dendrites arising from the basal stem shown in (B). (C) A side branch of the apical dendrite of the VEN shown in (B). A high density of spines can be observed on this dendrite. The magnification for (A) and (B) is the same, and is shown by the scale bar in (B). The magnification for (a) and (b) is the same, and is shown by the scale bar in (a). All the panels are compositions of several microphotographs of the same slide taken at different section depths.

Dendritic side branches

Typical VEN side branches were thin and arose from the cell body and the proximal part of the basal stem. Occasionally, more than 10 thin side branches were present. They mostly diverged in a slightly ascending direction similar to typical apical oblique branches, and either did not branch, or bifurcated once, giving two terminal dendrite branches.

Dendritic spines

Dendritic spine density was high (0.37 ± 0.06 μm⁻¹ on rapid Golgi‐, and 0.43 ± 0.06 μm⁻¹ on Golgi‐Cox‐impregnated VENs) on all side branches and on the branches arising from the basal stem ending (Figs 4 and 5C). The described dendritic spines had a mushroom shape and were relatively long. No mushroom‐shaped dendritic spines were found on the VEN soma.

Axon origin

The key morphological feature of VENs on Golgi sections was the site of the axon origin. The axon originated either from the lower side of the basal stem (Fig. 4), or from one of the thicker dendrites arising from the basal stem (Figs 3C and 5A). In some of the latter cases, the axon arose from a common trunk with one of the thicker dendrites of the basal stem (Fig 5B). The axon was oriented toward the white matter but could not be followed entirely. Even though on most pyramidal neurons the axon arises directly from the cell body (Figs 3B and 6B), on some it arises from the primary dendrite (Fig. 6A,C). However, the axon origin is always located near the cell body, i.e. within 10 μm of the cell nucleus. On VENs, the axon arose at least 100 μm away from the cell nucleus. This is particularly important because it allows distinguishing small, more oval‐shaped VENs from fusiform bipolar neurons also present in layers V and VI.

Figure 6.

Microphotographs of rapid Golgi‐impregnated neurons in Brodmann area 9 of the adult human prefrontal cortex (subject: CO 302). The scale bar is the same for all panels and is shown in the lower right corner. All the panels are compositions of several microphotographs of the same slide taken at different section depths. (A) Two modified spindle‐shaped pyramidal neurons positioned around the border of layers V and VI. The white arrows indicate the axon, which in both cases arises from the thick basal extension (stem), but still in close proximity of the soma. We found that, without clear and detailed definition of their somato‐dendritic morphology, these cells could, even on Golgi sections, be misinterpreted as von Economo neurons (VENs). (B) Two typical pyramidal neurons of layer IIIB. On the larger neuron, there is a more gradual transition between the soma and the apical dendrite. (C) Two typical pyramidal neurons of layer IIIC. On very large pyramidal neurons, the part of the cell extending in the apical direction remains very thick and might be considered as an extension of the soma. However, after giving the first side branch, the thickness decreases abruptly (marked with a black arrowhead) defining this cellular process as an apical dendrite. From the base of the cell body, two dendrites with a short common stem arise. The stem is also a place of axon origin (marked with a black arrow).

Quantitative analysis of somato‐dendritic morphology found VENs to be a diverse neuronal population

The quantitative analysis showed that VENs differ in size and complexity. For example, the ‘core part’ of VENs had a length between 150 and 250 μm. The median soma thickness was 15.5 μm (range: 10.1–22.0 μm) and the median soma length was 54.8 μm (range: 39.5–82.7 μm). VENs were, therefore, characterized by a low Aspect Ratio ranging from 0.14 to 0.40, with a median of 0.27 (Fig. 7A).

Figure 7.

(A) Histogram showing the distribution of von Economo neurons' (VENs’) Aspect Ratio. (B) Distribution of VEN's soma Area shown for both Golgi methods, and separately for rapid Golgi and Golgi‐Cox. (C) Distribution of VENs’ total dendritic length shown for both Golgi methods, and separately for rapid Golgi and Golgi‐Cox. (D) Correlation between VENs’ total dendritic length and soma Area.

The overall range of VENs’ soma Area was 271.3–837.7 μm² with a median of 589.9 μm (Fig. 7B). On Golgi‐Cox sections, the median soma Area was 486.6 μm² (range: 271.3–628.0 μm²), and on rapid Golgi sections, it was 696.3 μm² (range: 556.1–837.7 μm²). Based on their cell body size, almost all VENs belong to the group of large and very large neurons.

Qualitative analysis of Golgi‐stained VENs suggested that there are two groups of VENs based on the size and complexity of their dendritic tree. This observation was supported by quantitative data, which revealed a bimodal distribution of the total dendritic lengths of the analyzed VENs with two peaks – one between 1500 and 2500 μm and the other between 5000 and 6000 μm (Fig. 7C). This bimodal distribution was particularly pronounced on rapid Golgi sections where two distinct groups of VENs could be identified. On Golgi‐Cox sections, the total dendritic length of most VENs was grouped around the lower peak, with only two neurons in the upper peak and a single neuron between the two peaks. It should be noted that there was a weak, albeit statistically significant (P = 0.0316), correlation between the soma Area and the total dendritic length (Pearson correlation coefficient r was 0.44; Fig. 7D). This means that the size of the dendritic tree was largely unrelated to the size of the soma. Therefore, we defined VENs with a total dendritic length closer to the lower peak as ‘small VENs’ and VENs closer to the upper peak as ‘large VENs’. The single neuron between the two peaks was not included in the comparison of the two groups of VENs.

Large VENs had 20–34 primary dendrites arising from the core part (including the ending point of the basal stem). They were characterized by a higher median length (173.0 μm, range: 60.6–498.2 μm) of their terminal segments (Fig. 8A). Terminal segments were very thin, with an average diameter of 0.65 μm. The median length of intermediate segments on large VENs was 29.7 μm (range: 2.8–217.9 μm; Fig.8B).

Figure 8.

(A) Histogram showing the distribution of the terminal segment length of large von Economo neurons (VENs). (B) Histogram showing the distribution of the intermediate segment length of large VENs. (C) Histogram showing the distribution of the terminal segment length of small VENs. (D) Histogram showing the distribution of the intermediate segment length of small VENs.

Small VENs had 12–25 primary dendrites arising from the core part. They were characterized by a lower median length (163.7 μm, range for individual segments: 14.2–371.0 μm) of their terminal segments (Fig. 8C). The median length of intermediate segments on small VENs was 27.0 μm (range for individual segments: 3.0–120.1 μm; Fig. 8D).

The average diameter of intermediate segments was 30% higher on rapid Golgi (1.4 μm) than on Golgi‐Cox sections (1.0 μm) on both large and small VENs.

A total of 62% of intermediate segments on both types of VENs were of the first order, 25% were of the second order, and only 15% were of the third or higher order. Furthermore, more than 30% of the terminal segments were of the first order, and an additional 30% were of the second order. This means that more than 60% of the dendrites arising from the core part did not branch or branched only once. Therefore, dendrites arising from the main trunk contained almost three times more terminal than intermediate segments and, on average, dendrites contained only two segments. This means that 95% of the dendritic tree arising from the core part had a topology typical for the side branches of apical dendrites of typical pyramidal neurons (Sedmak et al. 2018).

Differentiating VENs from other classes of modified pyramidal neurons on Golgi sections

Qualitative and quantitative Golgi analysis indicated that VENs represent a specific subtype of modified pyramidal neurons with typical morphological features. We defined a typical VEN on Golgi staining as a neuron with the following morphological features: an elongated, stick‐like cell body gradually continuing into a thick apical and basal stem, a brush‐like basal stem arborization and an axon origin distant from the cell body.

The cell body size as well as the dendritic length of the largest VENs is comparable to that of large pyramidal neurons in the deep part of layer III of Brodmann area 9 (Petanjek et al. 2008; Zeba et al. 2008; Sedmak et al. 2018). This shows that large VENs belong to the largest and most complex neurons of Brodmann area 24. However, VENs’ basal dendrites show a much lower degree of bifurcation than the basal dendrites of large, small and medium‐sized pyramidal neurons (Fig. S2).

We reevaluated the presence of VENs on Golgi slides of different cortical regions from the Zagreb Neuroembryological Collection. Based on our definition of VENs, no cells with matching morphology were found in Broca's area, the primary motor and primary sensory cortex, the angular gyrus and the visual cortex. In Brodmann area 9 we identified only a single cell that could be defined as a VEN. Still, this cell had additional morphological features not observed on VENs of the ACC (Fig. 5A).

A large amount of modified pyramidal neurons in the lower part of layer V and the upper part of layer VI is triangular with one or two prominent dendrites, which arise from the basal edges (Fig. S3). In some cases, this basal dendrite has the shape of a thick, short stem with thin side branches (Figs S4 and S5) that have a slightly ascending course. This basal dendrite is not as prominent as the basal stem in VENs and does not have rich arborization around its ending. Additionally, such cells have at least one basal dendrite with typical branching. The axon on such types of modified pyramidal neurons usually arises from the basal surface of the cell body. Therefore, on Golgi sections, they can be easily distinguished from VENs.

Some modified pyramidal neurons have a bipolar orientation with a second prominent dendrite oriented perpendicularly and directed toward the white mater. This second prominent, ‘basal’ dendrite has numerous thin side branches and gradually decreases in thickness, often extending up to 500 μm deep into layer VI. Thus, this dendrite resembles the apical dendrite of pyramidal neurons and in some cases even reaches its thickness.

These bipolar cells have either a triangular or oval soma (Figs 9 and S6). Those with an oval soma are difficult to distinguish from smaller oval VENs based on the cell body and proximal dendrite morphology. However, in VENs, the short basal stem has an almost constant thickness until it bifurcates at the ending in a brush‐like manner. On bipolar modified pyramidal neurons, the prominent basal dendrite is longer, and its thickness decreases gradually without terminal brush‐like branching. The axon on these bipolar cells arises mainly from the soma or from the transitional part between the soma and the prominent basal dendrite. Rarely, the axon can arise from a more distal part of the prominent basal dendrite (Fig. 9). In this case (MOD1, Fig. 9), the part of the dendrite between the soma and the axon origin is shaped like a thick stem.

Figure 9.

Three‐dimensional Neurolucida reconstructions in the X–Y plane of two small von Economo neurons (VENs), one large VEN (VENL) and two bipolar modified pyramidal neurons (MOD1 and MOD2). All neurons were stained using the Golgi method. Bipolar modified pyramidal neurons (MOD) are located in layer VI and have an oval or fusiform, perpendicularly oriented soma with two prominent, long and thick dendrites, which gradually decrease in thickness. One is oriented toward the pia mater, and the other toward the white matter. Most oval cells have one side branch, which can be rather short (MOD1) or can extend for several hundred micrometers (MOD2). In such cases, the cell body is triangularly shaped with an apical, a basal and a side angle. Their prominent basal dendrite is thinner and longer than the basal trunk, and the axon arises from the cell body or the most proximal part of the prominent basal dendrite. The arrows indicate the axon, and the arrow with an asterisk points to the axon origin. Dendritic spines are not shown in order to allow a better visualization of the dendritic tree.

Fork cells, which are considered to be evolutionarily close to VENs (Seeley et al. 2012), had, in our sample, morphological features closer to modified pyramidal neurons with two prominent dendrites, where the second one is directed toward the pia mater, instead of toward the white matter (Fig. S7). Both the basal dendrites and the axon arose from the basal angle of the fork cells’ soma.

A comparison between VENs and other pyramidal neurons is given in Table 2.

Table 2.

Comparison of VEN and modified pyramidal neuron morphology

	VENs	Modified pyramidal neurons
Apical stem/apical dendrite (Golgi)	Very thick origin (up to 8 μm), very gradual decrease in thickness	Less thick origin, abrupt decrease in thickness
Basal tree bifurcation degree (Golgi)	Low	High
Basal stem/basal dendrite thickness (Golgi)	Constant or very gradual decrease	Abrupt decrease with length
Basal stem/basal dendrite arborization (Golgi)	High, brush‐like	Low
Axon origin (Golgi)	End of basal stem	Basal part of cell body
Nissl & NeuN	Long (50–120 μm) and stick‐shaped (5–10 μm wide)	Various shapes, including triangular and oval

VEN, von Economo neuron.

Differentiating VENs from other classes of modified pyramidal neurons on Nissl and NeuN sections

Unlike on Golgi sections, it is often challenging to distinguish VENs from other types of modified pyramidal neurons on Nissl sections (Fig. S1). There were a number of Nissl‐stained cells that we were not able to clearly classify either as VENs or as other types of modified pyramidal neurons (Fig. 10).

Figure 10.

High‐power microphotograph of Nissl slides of layer Vb of Brodmann area 24 (subject: CO 311). (A) and (C) are taken from cryoprotected tissue blocks, and (B) is taken from a paraffin‐processed tissue block. The scale bar is the same for all panels. (A) Three von Economo neurons (VENs) are shown grouped together (marked with an asterisk). Layer Vb also contains other types of modified pyramidal neurons. Bipolar neurons are marked with white arrowheads, while triangular or rhomboid‐shaped cells with two prominent perpendicularly oriented dendrites are marked with white arrows. Note that one bipolar cell has a large soma and two prominent dendrites (indicated by small white arrowheads), and that the cell with a triangularly shaped soma has a thick basal stem (marked with a small black arrow). Therefore, it cannot be excluded that those cells might be VENs with a more oval soma. Also, in the middle positioned VEN, the soma is more oval, and the apical stem is not as visible. It is, therefore, possible that this might not be a VEN, but rather another subtype of modified pyramidal neurons. (B) On paraffin sections, VENs are more spiral‐shaped and resemble a corkscrew. (C) The panel shows a VEN where the bifurcation of the basal stem into two thick dendrites is visible (marked with a black arrow). The second neuron is an elongated pyramidal cell with no clear transition between the soma and the thick apical dendrite. However, the basal stem is not visible. Note that on all panels showing Nissl slides, VENs are much smaller than on Golgi slides.

Nissl staining in the human ACC visualized the core part of VENs (soma, apical and basal stems). VENs were located in clusters of 3–5 cells in layer V, and on the border of layers V and VI (Fig. S1). On cryoprotected sections, they were long (50–120 μm), relatively thin (5–10 μm), and spindle‐ or stick‐shaped. On paraffin‐embedded sections they were often corkscrew‐shaped (Fig. 10).

There was no sharp decrease in cell thickness towards the ends of the staining and, therefore, no clear demarcation between the cell body and the apical and basal stems. This feature distinguished VENs from other cells with an oval cell body and bipolar orientation. Furthermore, in most VENs, the spherical nucleus was stained darker than the surrounding cytoplasm, and was shifted laterally from the middle axis of the cell. The axial orientation of the cells was perpendicular to the pia mater. Some of the large and slender pyramidal neurons also lacked a clear demarcation between the cell body and the apical dendrite, but did not have a basal stem and their nucleus was located on the base of the triangular cell body.

Von Economo neurons on NeuN staining had a morphology similar to the one observed on Nissl sections; however, it was more difficult to identify VENs on NeuN sections. The evaluation of NeuN and Nissl slides of the middle part of Brodmann area 24 of the rhesus macaque monkey did not reveal any cells with typical VEN soma morphology. A more comprehensive analysis of the ACC of the rhesus monkey, including Golgi staining, would be of interest, but is out of the scope of this study.

Immunohistochemical analysis in the human confirmed that VENs express both MAP2 and SMI‐32 (Fig. 11), typical markers of principal neurons of the cerebral cortex. Both MAP2 and SMI‐32 signals were found in the cytoplasm of VENs, mostly in the middle part of the cell near the cell nucleus. No difference was observed in the staining pattern of MAP2 and SMI‐32 between VENs and surrounding projection neurons, though VENs gave a qualitative impression of being slightly less intensely stained.

Figure 11.

Microphotograph of fluorescent double‐labeling of von Economo neurons (VENs) from Brodmann area 24 (subject: CO 385). The scale bar is the same for both panels. NeuN labeling is seen as green, while MAP2 and SMI‐32 labeling are seen as red. Co‐localization is seen as yellow. VENs are marked with white arrows. (A) NeuN/MAP2 double‐labeling showing that VENs express the microtubule‐associated protein 2 (MAP2). (B) NeuN/SMI‐32 double‐labeling showing that VENs also express the non‐phosphorylated form of neurofilament H (SMI‐32).

Discussion

Main findings

In this study, we provided a detailed description of the somato‐dendritic features of Golgi‐impregnated VENs in Brodmann area 24 of the human brain. In archive material, we reevaluated the presence of VENs in various cortical areas, including Brodmann area 9, on Golgi slides, and analyzed the reliability of their identification on Nissl and NeuN sections in Brodmann area 24. We found that on Golgi sections, based on specific morphological features, VENs could be clearly identified as a specific cell type within the population of modified pyramidal neurons in the ACC. The distinct morphological features of VENs were: a thick, long and perpendicularly oriented core part composed of the soma, apical stem and basal stem. Thin side branches arose from the core part and were particularly numerous around the ending of the basal stem. The orientation and topology of the side dendrites resembled those of apical dendrite side branches (oblique dendrites), i.e. they had a low bifurcation rate and tended to be directed horizontally (Sedmak et al. 2018). The most distinct feature of VENs was the distant origin of the axon, which arose from the ending of the basal stem, often having a common origin with one of the dendrites arising from the basal stem.

Furthermore, based on their total dendritic length, VENs could be divided into two groups – small and large. This grouping was more pronounced on rapid Golgi than on Golgi‐Cox, which could be attributed to methodological differences between the two staining methods (Koyama, 2013; Kang et al. 2017). The observed differences in VEN soma Areas and in the thickness of intermediate segments between rapid Golgi and Golgi‐Cox could be also explained by a higher shrinkage factor present on Golgi‐Cox staining (Koyama, 2013; Kang et al. 2017). Here we point out that the Golgi method stains only a small percentage of neurons. Nevertheless, Golgi staining is widely considered to be random (Koyama, 2013; Kang et al. 2017) and is likely representative of the VEN population as a whole.

Establishing clear morphological features for identifying VENs on Golgi sections was crucial for their differentiation from other similar modified pyramidal neurons. We found that on Nissl or NeuN sections, only long stick‐shaped cells with thick apical and basal processes could be reliably defined as VENs. Other modified pyramidal neurons with a spindle‐shaped or oval soma could easily be misinterpreted as VENs. Vice versa – it could be difficult to identify VENs with a smaller and more oval soma using methods that do not provide more elaborate dendritic staining.

Somato‐dendritic morphology of VENs

The Golgi studies describing somato‐dendritic morphology of VENs in humans are limited to two human specimens. The first one is Cajal's description of the brain of a 1‐month‐old girl (Cajal, 1995). The cells described by Cajal had the same morphological features as the large VENs described in our study. They had a large core part, numerous thin side branches with high density of dendritic spines and, most importantly, it was clearly shown that the origin of the axon was at the end of the basal trunk (Cajal, 1995). These cells had a more oval shape than the VENs described in our study. One possible explanation for this is regional differences in morphology, as Cajal described these cells in the insular cortex, whereas we analyzed VENs in the anterior cingulate region. Another explanation is developmental changes – Cajal described these cells in a 1‐month‐old girl, and our material originated from adult specimens. The latter is supported by observations made by von Economo who, on Nissl‐ and Bielschowsky‐stained sections, stated that earlier during development, VENs were more spindle‐shaped, whereas in adults they were more rod‐shaped (von Economo, 1926; Seeley et al. 2012).

The second Golgi study of VENs in humans defined them as cells ‘with an elongated, large soma in layer V of the FI or ACC, a prominent basal dendrite, and symmetrical morphology along the horizontal and vertical axes of the cell’ (Watson et al. 2006). They further included ‘only neurons with no additional dendrites or branching for a half‐soma's distance along the length of the proximal dendrites’ (Watson et al. 2006). Therefore, most cells classified as VENs in this study lacked side branches, whereas we showed that even small VENs had numerous side branches arising from the soma. The total dendritic length of these cells was about 800 μm, suggesting that this study did not cover the large VENs typically found in the human brain (von Economo, 1926; Cajal, 1995; Nimchinsky et al. 1995, 1999). Some of the cells analyzed by Watson et al. (2006) could fit our definition of the small VEN subtype (see fig. 4A in Watson et al. 2006), but most of these cells were bipolar oval projection neurons with few dendritic spines (see fig. 4B in Watson et al. 2006). Our analysis of dendritic spines showed that VENs had a substantial amount of spines and a spine density comparable to that of the large long‐projecting pyramidal neurons of layer IIIC of the human prefrontal cortex (Zeba et al. 2008; Petanjek et al. 2011).

Two Golgi studies on VENs were done on non‐human specimens. The first is a study by Evrard et al. (2012), which analyzed VENs in the anterior insular cortex of the macaque monkey. In this study, VENs were described as rather small oval cells with a clear demarcation between the soma and the prominent apical and basal dendrites (Evrard et al. 2012). However, the ending of the basal stem and the distant origin of the axon strongly support that those cells were VENs (see fig. 2A in Evrard et al. 2012).

The second Golgi study on VENs in non‐human specimens was done by Butti et al. (2014). In this study, VENs were identified in the cortex of the pygmy hippopotamus (Butti et al. 2014). The morphology of two cells demonstrated as VENs in this study (figs 13B and 15C in Butti et al. 2014) differs from the VENs in the human ACC. On the other hand, several other cells, which were not defined as VENs, had a dendritic morphology similar to that of VENs in humans (figs 13C,F, 14D,E in Butti et al. 2014). However, the low power of magnification at which the neurons in this study were shown as well as the lack of a clear indication of the axon origin prevents us from making a stronger conclusion.

It is interesting to note that in the studies using Golgi staining no further elaboration was done on how to distinguish VENs from other types of modified pyramidal neurons (Watson et al. 2006; Evrard et al. 2012; Butti et al. 2014). A comparison of all Golgi studies analyzing VENs is shown in Table 3.

Table 3.

Comparison of different Golgi studies on VENs

Study	Specimen used for Golgi staining	Region analyzed	Golgi staining method	VEN morphological features
Cajal (1899)	1‐month‐old human girl	FI	Classic Golgi staining	Spindle‐like soma, prominent apical and basal dendrites, axon shown to arise from basal dendrite
Watson et al. (2006)	23‐year‐old human male	ACC and FI	Modified Golgi technique	Large elongated soma, prominent basal dendrite, no additional dendrites or branching for a half‐soma's distance along the length of the proximal dendrites, axon not shown, clear demarcation between soma and dendrites is visible in some neurons
Evrard et al. (2012)	One rhesus macaque	Anterior insular cortex	Rapid Golgi‐Cox	Spindle‐like soma, apical dendrite branches distally into several thinner spiny dendrites, basal dendrite branches into thinner spiny dendrites in layer VI, axon shown to arise from basal dendrite
Butti et al. (2014)	One female pygmy hippopotamus	Lateral gyrus, frontal magnocellular cortex, ACC	Modified rapid Golgi	Stout cell body, some with slender cell body, apical and basal dendrites almost as thick as the cell body, hint of basilar skirt, axon not shown

ACC, anterior cingulate cortex; FI, fronto‐insular cortex; VEN, von Economo neuron.

Identification of VENs based on their soma shape

It is surprising that a precise definition of VENs based on their somato‐dendritic morphology has not been established before, and that studies on VEN dendritic morphology are extremely rare. On the other hand, studies analyzing the distribution of VENs using methods that do not visualize the entire dendritic tree are numerous. We found that it is difficult to reliably identify VENs only based on the size and shape of their soma. For example, comparative Golgi analysis of VENs and other types of modified pyramidal neurons showed that many VENs with a smaller soma could not be distinguished from bipolar neurons with spines on Nissl or NeuN sections. The different shapes of VENs on cryoprotected and paraffin‐embedded Nissl sections show that the tissue shrinkage can significantly affect the final appearance of cells on such specimens (Braitenberg & Schüz, 1991). It is therefore possible that large bipolar neurons and pyramidal neurons with a long slender soma could greatly resemble VENs on Nissl sections. It is also possible that in non‐primate mammals the somatic morphology of some other cell types greatly resembles that of VENs on Nissl sections. Their dendritic topology, however, could be completely different from the VENs we described and might be more similar to that of other modified pyramidal neurons.

Over the years, different definitions of VENs were used for their identification on Nissl and immunohistochemical sections. Some authors defined VENs as oval or elongated cells found in layer V of the ACC and FI (Nimchinsky et al. 1999; Watson et al. 2006). When such definitions were used, VENs were distinguished from bipolar fusiform cells based on their laminar distribution, i.e. spindle‐shaped cells in certain layers (usually layer V) were defined as VENs, while similar spindle‐shaped cells in other layers were not counted as VENs (Nimchinsky et al. 1999; Butti et al. 2009). Such a definition automatically excludes the possibility of identifying VENs in other cortical regions or layers. Other authors defined VENs merely as elongated spindle‐shaped cells with thick apical and basal dendrites (Raghanti et al. 2015). Here the distinction between VENs and bipolar fusiform neurons is made based on their size, where large spindle‐shaped cells were defined as VENs and smaller spindle‐shaped cells were not counted as VENs (Butti et al. 2009; Raghanti et al. 2015). Such a definition of VENs automatically excludes the VENs with a smaller soma we described on Golgi sections, and possibly includes larger bipolar fusiform cells and other modified pyramidal cells of a similar soma shape. This might explain why in these studies VENs were present in multiple cortical regions and layers as well as in such a broad spectrum of different species.

The presence of VENs in upper cortical layers (Raghanti et al. 2015) is particularly surprising as none of the principal neuron types typically found in these layers resembles VENs (Braak & Braak, 1985). Layer II has a granular appearance due to the small principal neurons that are the dominant neuronal population of this layer. These neurons are characterized by triangular somas and short apical dendrites. Layer III is dominantly populated by typical pyramidal neurons (Braak & Braak, 1985). The upper cortical layers are, however, characterized by a considerable number of GABAergic interneurons, of which some subpopulations, for example, parvalbumin interneurons, are perpendicularly oriented and oval‐shaped (Condé et al. 1994). Such neurons could, in some mammalian species, resemble VENs on Nissl sections.

The unique somato‐dendritic morphology of VENs results in specific functional properties

The specific morphology of the VEN dendritic tree and the position of the axon origin likely contribute to specific electrophysiological properties of VENs, probably the triggering of action potential as well as burst or tonal firing (van Elburg & van Ooyen, 2010). We speculate that the presence of the basal dendritic tuft in the relative proximity of the axon origin may influence the firing rate of VENs and thereby their functional properties as a whole.

The intrinsic firing pattern of pyramidal neurons is determined by their dendritic topology (length, branching pattern and orientation of dendrites) and the distribution of ion channels (Cossart et al. 2006; van Elburg & van Ooyen, 2010). A high variability in somato‐dendritic morphology suggests that different types of principal neurons carry out specialized functions. In that sense, it is likely that VENs undergo extraordinary functional specialization when compared with other subpopulations of pyramidal and modified pyramidal neurons. VENs have an extensive amount of horizontally oriented thin side branches, which do not bifurcate frequently. Long and thin dendrites with a low level of branching have high impedance (Komendantov & Ascoli, 2009; Ferrante et al. 2013). Therefore, it is likely that they contribute to the triggering of action potential only if they contain a high amount of excitatory synapses and if they receive synchronous excitation. The side branches of VENs are densely populated by spines, showing that they contain a high amount of excitatory input (Megías et al. 2001). The high thickness of the core part suggests a low impedance of this domain that may be important for a rapid conduction of excitatory inputs to the axon hillock and initial segment. Furthermore, the specific origin of the axon in VENs implies an unusual position of the axon hillock and initial segment. The basal dendritic tree, especially the dendrite arising from a common trunk with the axon, may have significant influence on the triggering of action potential (van Elburg & van Ooyen, 2010).

Some large VENs have a dendritic tree that likely extends up to 1 mm in the horizontal plane. This suggests that large VENs receive a strong input from two columns. This is particularly interesting in the context of the columnar organization of the ACC. This region receives interdigitating cortical inputs from the dorsolateral prefrontal and associative parietal region (Goldman‐Rakic, 1988). Therefore, large VENs may be activated by a synchronous activity of afferents coming from these two cortical sources. On the other hand, the horizontal extension of small VENs suggests that they may be activated by inputs from a single column.

The fact that VENs are highly immunoreactive to peptides related to pain (ATF3, IL4Ra and NMB) suggests that they are involved in the interoception of one's homeostatic condition (Stimpson et al. 2011). Research also revealed that regions of the cortex with an abundance of VENs are connected with the frontal cortex, insular cortex and the amygdala (Allman et al. 2011). Furthermore, areas containing VENs have a frontoparietal connectivity profile, suggesting that VENs are probably involved in saliency detection and self‐regulation networks. This enforced the possible role of VENs in homeostatic functions (Cauda et al. 2013). The fact that VENs express both MAP2 and SMI‐32 strongly suggests that VENs are a type of pyramidal and projection neurons (Nimchinsky et al. 1995; Fajardo et al. 2008). The slightly less intense staining of VENs for MAP2 and SMI‐32 compared with surrounding pyramidal neurons, observed on immunohistofluorescence in our study, may indicate that their axons do not establish connections with multiple areas or do not have such an extensive local arborization, though further clarification is needed.

Evolution of VENs

Literature shows that VENs are abundant in the anterior insular cortex of the macaque monkey, but their presence in the ACC of this species is still not completely resolved (Nimchinsky et al. 1999; Evrard et al. 2012). During hominid evolution, VENs increase in size and density, and are found in both the anterior insular cortex and ACC (Nimchinsky et al. 1999). Moreover, the VENs described in the macaque insular cortex are smaller and their morphology is not as complex as in humans (Evrard et al. 2012). This suggests that VENs appeared in the ACC during hominid evolution and not before the division of Old and New World monkeys (Nimchinsky et al. 1999). If we put the increase in complexity and density of VENs into the generally accepted propositions of human‐specific aspects of brain development (Clowry et al. 2010), it is possible that large VENs represent a novel human‐specific neuronal type.

Regarding the numerous reports stating that VENs are also present in other species outside of the primate line, it is important to note that most of this research was done on Nissl‐stained tissue. In these studies, the definition of VENs was not consistent, and descriptions of their dendritic morphology were either absent or lacking. This does not exclude the possibility that VENs could be sporadically found in non‐primate species as well as in other regions of the cerebral cortex of apes and humans. However, we found that present evidence strongly suggests that VENs are abundant only in specific regions of primates. Their increase in density and size follows the general increase in complexity of the cerebral cortex in hominids (Nimchinsky et al. 1999; Evrard et al. 2012).

Conclusion

Von Economo neurons are a distinct population of neurons abundant in the ACC and FI of the human brain. VENs have a specific somato‐dendritic morphology, and the most notable morphological feature for their reliable identification is the extremely distant position of the axon origin. Based on their total dendritic length, VENs can be categorized into two groups – small and large, where the latter might represent a specific cell type that evolved in the hominid line. To clearly establish the function and evolution of VENs, further comprehensive studies, comparing their morphology on Golgi sections between different species, are needed. For such studies to be comparable, a clear and unambiguous definition of VENs must be used. We suggest that until specific markers for VENs are established, the results of quantitative studies of VENs using Nissl or NeuN staining should be taken with caution as they might not be reliable enough to draw more generalized conclusions. A combination of multiple staining methods, including Golgi staining, is needed to reliably identify VENs. The detailed morphological descriptions of VENs presented in this paper give ground for future advances in this field.

Author contributions

I.B., D.S. and Z.P. designed the study, analyzed the data and drafted the manuscript. I.B., D.S., D.DŽ., D.J. and M.R.R. contributed to the acquisition of data. N.J.M., M.R.R. and Z.P. performed a critical revision of the manuscript.

Conflict of interest

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.

Supporting information

Fig. S1 Microphotograph of a Nissl slide of Brodmann area 24.

Fig. S2 Three‐dimensional reconstruction of a VEN and two typical pyramidal neurons.

Fig. S3 Three‐dimensional reconstructions of two VENs and two pyramidal neurons.

Fig. S4 Three‐dimensional reconstructions of two VENs and one modified pyramidal neuron.

Fig. S5 Three‐dimensional reconstructions of two large VENs and two modified pyramidal neurons.

Fig. S6 Three‐dimensional reconstructions of three small VENs, one large VEN and two bipolar modified pyramidal neurons.

Fig. S7 Three‐dimensional reconstructions of one small VEN, two large VENs and two fork‐shaped modified pyramidal neurons.

Appendix S1. Detailed descriptions of supplementary figures.