Construction of a polarized neuron


  • This review was presented at The Journal of Physiology Symposium entitled Size Matters: Formation and Function of GIANT Synapses, which took place at the Annual Meeting of the Society for Neuroscience, New Orleans, LA, USA, on 12 October 2012. It was commissioned by the Editorial Board and reflects the views of the authors.

G. A. Spirou: Center for Neuroscience, West Virginia University School of Medicine, One Medical Center Drive, Morgantown, WV 26506-9303, USA.  Email:


Abstract  Aside from rare counterexamples (e.g. the starburst amacrine cell in retina), neurons are polarized into two compartments, dendrites and axon, which are linked at the cell body. This structural polarization carries an underlying molecular definition and maps into a general functional polarization whereby inputs are collected by the dendrites and cell body, and output is distributed via the axon. Explanations of how the polarized structure arises invariably coalesce around somatic polarity, defined by the roving location of the microtubule organizing centre, or centrosome, the Golgi apparatus, associated endosomes and the nucleus during early development. In some neurons, proper positioning of these structures can determine the sites for axon and dendrite elongation, and support processes that underlie cell migration. We briefly review these events as a basis to propose a new role for polarized arrangement of somatic organelles as a potential determinant for patterned innervation of the cell body membrane. We cite an example from preliminary studies of synaptogenesis at the calyx of Held, a large nerve terminal that selectively innervates the cell body of its postsynaptic partner, and suggest other neural systems in which polarity mechanisms may guide initial synapse formation onto the somatic surface.

  • image

[ Paul Holcomb is a doctoral student in neuroscience at West Virginia University, USA, under Dr George Spirou. Paul received a BE in biomedical engineering in 2005 from Vanderbilt University. His research focuses on mechanisms that influence the outcome of synaptic competition within the developing CNS. He resides in Morgantown, WV, USA, with his wife, Micky, and two children, Colin and Isabel. George A. Spirou, PhD, is Director of the Center for Neuroscience at West Virginia University, USA. He was trained in whole-animal and brain slice neurophysiology of the auditory system, completing a PhD in Neuroscience at the University of Florida and a postdoctoral fellowship in Biomedical Engineering at Johns Hopkins University. More recently his interests have included nano-scale, large-volume investigation of nerve terminals and neural circuits in both the mature and developing brain.]


calyx of Held


linkers of nucleoskeleton to cytoskeleton


medial nucleus of the trapezoid body


serial block-face scanning electron microscopy

Organelles that define somatic polarity during neural development

As neurons develop, the positions of their centrosomes, Golgi complex and nucleus define intra-somatic organization. The centrosome is an ancient organelle thought to have evolved in metazoans from the ciliary basal body (Bornens & Azimzadeh, 2007). The centrosome is comprised of so-called mother and daughter centrioles which, at their core, are 9 × 3 circular arrangements of microtubules and contain associated proteins that comprise the pericentriolar material (Paintrand et al. 1992; Bornens, 2008; Azimzadeh et al. 2009). The centriole is a highly stable arrangement of tubulin protein that can anchor the negative ends of microtubules and support their extension to form important elements of the cytoskeleton (Bornens, 2002). Centrosome position defines cell polarity during mitosis, whereby centrioles move to opposite poles of the cell and fission occurs along a perpendicular axis through the middle of the cell (Fig. 1A; Higginbotham & Gleeson, 2007). In mammals, neural precursors are positioned at the ventricle and undergo asymmetric mitosis that yields a postmitotic neuron. This process continues until the appropriate complement of neurons is born. Following mitosis, the formation of a nuclear membrane and growth of cytoplasmic organelles result in a neuronal cell body that is internally polarized by the alignment of the centrosome, Golgi apparatus and nucleus. This organization can be established cell autonomously, as evidenced in cell culture (Solomon, 1981), although in vivo extrinsic factors play important roles (reviewed in Barnes et al. 2008). Both intrinsic cell programs and extrinsic cues work in concert to influence the polarization of the neuron. The breadth of factors involved in this signalling and crosstalk is outside of the scope of this review, and has been surveyed recently in detail by other authors (Barnes & Polleux, 2009; de la Torre-Ubieta & Bonni, 2011). In the next sections, we explore further events in neural development that occur in preferred orientation to a polarized cell body, describable by intra-somatic location of any combination of these cytoplasmic organelles.

Figure 1.

Polarity of somatic organelles at multiple stages of neuron development 
A, during mitosis, intra-somatic polarity is defined by centrosome (light blue) position at opposite poles of cells aligned with reforming Golgi apparatus (green) and nucleus (red). B, prior to radial migration, neurons enter a bipolar phase with extension of an axon from the region of the soma (red) nearest the centrosome and Golgi apparatus. Microtubules (black) in the nascent axon are extended from nucleation sites in either the centrosome or Golgi apparatus. The second process typically emits at the opposite pole of the cell (blue). Most neurons initially migrate from their site of origin prior to entering a bipolar state for radial migration. C, intra-somatic polarity in radially migrating neurons is typified by a leading centrosome, often an intervening Golgi apparatus and a following nucleus. In the cortex the leading process is a dendrite (blue bar), and an axon (red) extends behind the translocating nucleus. Radial migration occurs along a glial substrate (orange bar).

Somatic polarity that correlates with location of axonal extension

Soon after cell division, neurons begin to build their axons and dendrites. In cultured Drosophila melanogaster neurons, just after mitosis the axon extends from the end of the cell adjacent to the centromere (de Anda et al. 2005). The situation may be less consistent in vertebrates; early studies on dissociated hippocampal neurons established phases for multiple neurite extension and concluded that the selection of one neurite to be the axon was a seemingly random process (Dotti et al. 1988). During the past 15 years, high-resolution time-lapse imaging combined with fluorescence labelling has permitted a more refined view of these developmental dynamics. This body of work supports the dogma that the axon forms before dendrites and probes the earliest events that determine which of the neurites are destined to form an axon (Barnes et al. 2008; Barnes & Polleux, 2009; Hoogenraad & Bradke, 2009; Cheng & Poo, 2012).

Multiple model systems applied to the study of axon formation follow a pattern of growth associated with intra-somatic polarity. For example, in hippocampal neurons cultured as early as embryonic day 16, centrosomes, Golgi apparatus and endosomes cluster to one region of the cell periphery (Fig. 1B). The first two neurites extend at opposite ends of the cell in alignment with this intra-somatic polarity, and the neurite nearest the centromere becomes the axon in 71% of neurons (de Anda et al. 2005; Calderon de Anda et al. 2008). The cerebellar granule neuron is another popular model to study acquisition of neuronal polarity because it generates, from opposite ends of the cell body, tangential axons and an ascending axon that together form a ‘T’ shape and one–three descending dendrites that collect mossy fibre inputs (Ramon y Cajal, 1995). Time-lapse imaging of cultured granule cells reveals that each branch of the ‘T’ is generated sequentially, that each branch originates from a centrosome/Golgi cluster, and that this cluster moves from one end of the cell body to the other to initiate each process (Zmuda & Rivas, 1998). These processes label for axonal markers (Kawaji et al. 2004). In a third model system, the cerebral cortex, most neurons are born in the ventricular zone, so named because it lines the lateral ventricles, via asymmetric mitosis of radial glial progenitor cells (Noctor et al. 2004). Neurons then migrate a short distance into the adjacent subventricular zone and extend multiple neurites in a multipolar morphology (LoTurco & Bai, 2006). Neurons next move closer to the ventricle and extend an axon to contact it (Noctor et al. 2004). In organotypic slice culture, centrosomes migrate to the ventricular zone side of the cell body prior to axon extension. A mixed picture is presented by zebrafish Rohon-Beard neurons, which extend ascending and descending central axons along the spinal cord longitudinal axis and a perpendicular peripheral axon that exits the spinal cord. Only the location of this peripherally directed axon is predicted by centrosome location (Andersen & Halloran, 2012). Additionally, in zebrafish rhombic lip and retinal ganglion cells axonogenesis occurs apparently along the somatic polarity axis, but from the opposite end of the soma from the centriole (Zolessi et al. 2006; Distel et al. 2010). This process appears to respect intra-somatic polarity but indicates differential roles for the centrosome in process extension (Zolessi et al. 2006; Distel et al. 2010).

Somatic polarity, cell migration and dendritogenesis

Migration of neurons is another developmental process where intra-somatic polarity is evident. Popular models to study neuron migration in the mammalian brain include the cerebral cortex and granule cells of cerebellar cortex due to the final product of a patterned, laminar array of neuronal cell bodies. In the cerebral cortex, excitatory neurons enter a bipolar phase where the process extended toward the ventricle is axonal and the oppositely directed process leads migration into the cortical plate and becomes the apical dendrite (Fig. 1C; Noctor et al. 2004). The nucleus moves through the apical dendrite by translocation or the entire neuron locomotes, typically along radial glial processes (Nadarajah et al. 2001). Within 2 h after localizing near the site for axon extension, the centrosome has relocated to the opposite pole of the cell body at the base of the process that becomes the apical dendrite (de Anda et al. 2010). In tangentially migrating cortical interneurons, which move via nuclear translocation, the centrosome proceeds through the apical dendrite to distances up to 30 μm from the soma, then the nucleus is squeezed forward by an actin/myosin II network (Bellion et al. 2005; Higginbotham & Gleeson, 2007). This saltatory process repeats until the soma reaches its appropriate position. A similar arrangement of centrosome and Golgi leading the nucleus occurs in cultured cerebellar granule neurons as they migrate along leading axonal processes (Gregory & Edmondson, 1988; Zmuda & Rivas, 1998; Solecki et al. 2004; Tanaka et al. 2004). However, during radial migration in organotypic culture, the polarized position of these organelles can be upset as the nucleus and Golgi can overtake the centriole during phases of rapid movement through the leading process (Umeshima et al. 2007). A similar observation was made for neurons born in the rhombic lip monitored in vivo in zebrafish (Distel et al. 2010). These results suggest that nucleus and centromere are not always tethered and can move along independent arrangements of microtubules (Umeshima et al. 2007). However, during resting phases of somatic migration, the centrosome regains its leading position, reestablishing intra-somatic polarity. The resting stage may be necessary for establishment of microtubules for the next phase of rapid movement (Umeshima et al. 2007; Distel et al. 2010).

Positioning the nucleus

Nuclear positioning in the postmitotic and postmigratory neuron defines the location of the cell body and establishes a basis for somatic polarity. The microtubule cytoskeleton is central to nuclear positioning, and accomplishes this feat across phyla and tissues by force generation through extension from a microtubule-organizing centre or movement of the nucleus along microtubule tracks (Reinsch & Gönczy, 1998). The anchoring of the nucleus to a specific location within the cell is mediated by multimolecular assemblies called linkers of the nucleoskeleton to the cytoskeleton, or LINC, complexes composed of multiple transmembrane proteins that cross the nuclear membrane and anchor the nucleus to the cytoskeleton. Selective deletion of LINC constituents in mice causes significant disruption in the localization of synaptic nuclei within skeletal muscle; double knockout of two of these proteins, Sun1 and Sun2, is embryonically lethal in mice unless rescued by CNS-specific expression of Sun1, indicative of an essential role for this protein in CNS development (Razafsky & Hodzic, 2009). Work in both D. melanogaster and zebrafish indicates that maintenance of nuclear position in highly polarized photoreceptor cells requires the dynactin complex, potentially through interaction between this complex and the LINC complex (Whited et al. 2004; Tsujikawa et al. 2007). Location of the nucleus can be important for cellular function, as disruption of nuclear positioning in zebrafish photoreceptors leads to deficits in synaptic differentiation and cell death (Tsujikawa et al. 2007).

A new view of somatic innervation

Specification of neuronal polarity sets up a more refined stage of neural development, whereby extending axons select and compete for postsynaptic targets restricted to regions of the dendritic tree (proximal versus distal) or particular dendritic structures (spine head versus dendritic shaft). Interestingly, several proteins involved in cell polarity have also been implicated in synaptogenesis (Wiggin et al. 2005). Although many neurons have particular inputs targeted more to the soma than dendrites, a few cell types receive inputs that are selectively and specifically targeted to the soma at a particular stage of development or in the mature state. These instances are recognizable and quantifiable because only a small number of inputs, and in many cases only one input, make this connection. Might the intra-somatic polarity of the neuron play a role in defining innervation territories for these cell types?

One system that exhibits somatic mono-innervation is the calyx of Held (CH) innervation of the principal cell of the medial nucleus of the trapezoid body (MNTB; Held, 1893). Early in development, MNTB principal cells are innervated by multiple excitatory inputs (Hoffpauir et al. 2006, 2010; Rodríguez-Contreras et al. 2008; Marrs & Spirou, 2012). However, the mature principal cell possesses only a single calyceal input located on the somatic surface (Spirou et al. 1998; Rodríguez-Contreras et al. 2006). To date, the development of the CH has yet to be studied with regard to somatic polarity. By combining serial block-face scanning electron microscopy (SBEM) and manual reconstruction of pre- and postsynaptic structures within the MNTB, the three-dimensional relationship among developing inputs, intracellular structures and cellular polarization can be determined. In order to examine key aspects of calyceal development, samples were obtained from mice during the first postnatal week encompassing the period of explosive growth of the calyx (Hoffpauir et al. 2006, 2010).

Initial observations from SBEM reveal nuclear eccentricity within MNTB principal cells, with the nucleus located at one pole of the cell, and the Golgi apparatus and other cytoplasmic structures located at the opposite pole. In order to explore this observation further, the centre of geometry (centroid) was calculated for both the nucleus and cell body of several principal cells innervated by newly-formed calyces. Comparison of nucleus and cell body centroids showed significant asymmetrical placement of the nucleus within the cell body (Fig. 2). The location of calyx growth on the somatic surface was mapped as their apposed surface area. We then noted that the apposed surface area was almost exclusively on the non-nuclear pole of the cell and was penetrated by a polarity vector drawn between the two centroids.

Figure 2.

3D reconstruction of MNTB principal cell 
The nucleus (red) is eccentrically located within the cell body (tan contours) and lies opposite from the newly formed calyceal input (blue). The centroids of the nucleus and cell body (green spheres) are maximally separated given their cellular geometry. The polarization axis is indicated by a vector passing through the centroids of the nucleus and cell body. Scale bar: 2 μm.

These initial observations suggest that there exists a somatic polarization within MNTB principal cells defined by the location of the nucleus within the cell. The segregation of large CH inputs onto the non-nuclear pole of the cell soma provides evidence for control of input location and potential input selection by a mechanism linked to somatic polarity. Interestingly, reconstruction of a P9 MNTB principal cell by Sätzler et al. (2002) showed the nucleus juxtaposed with the growing calyceal input. This observation suggests that somatic polarity is transient, and may be relaxed once innervating axons have found their target and grown large over the cell body.

Several other systems exhibit exclusive somatic innervation and may share features with the CH:MNTB. During development of climbing fibre mono-innervation of Purkinje cells, multiple inputs converge onto the cell body, but the ‘winning’ input is the first to cover the somatic surface where the primary dendrite will emerge (Hashimoto et al. 2009; Watanabe & Kano, 2011). This transitional state is suggestive of a polarized somatic surface, and raises the possibility that polarized arrangement of the same somatic organelles and associated mechanisms that determine axonal and dendritic initiation sites underlie this later stage of innervation. To our knowledge, the association of somatic innervation with Purkinje cell somatic polarity has not been investigated. Chick ciliary ganglion neurons are transiently mono-innervated during development by a calyx-type terminal that partially occupies the somatic surface in the region of axonal extension (Martin & Pilar, 1964; Landmesser & Pilar, 1972). However, the relationship between this innervation site and underlying organelle arrangement has not been probed. Innervation of rat submandibular ganglion transitions through a competitive phase of multiple somatic inputs to yield a single, winning somatic input (Lichtman, 1977, 1980). These inputs are located on small somatic processes. Although some illustrations suggest a polarized distribution of these processes on the somatic surface (McCann et al. 2008), systematic analysis has not been performed. In none of these systems has the position of the nucleus, Golgi and centrosome been mapped. In summary, the established developmental stages associated with polarized alignment of somatic organelles and the rapidity with which these organelles can reorient and reorganize suggest that intra-somatic polarization is a flexible, adaptive mechanism with as yet uncharacterized roles in neural circuit formation and function.



The authors thank Brian Pope for preparation of tissue samples. Eric Tucker and Pete Mathers provided helpful discussions and critique of the manuscript. This work was supported by NIH grants R01 DC007695 to G.S., CoBRE Grant P20 GM103503 to the WVU Center for Neuroscience (G.S.), and P41 GM103412 to the National Center for Microscopy and Imaging Research (M.E.). None of the authors of this manuscript has any potential or perceived conflicts of interest.