Emerging roles of the centrosome in neuronal development

The role of the centrosome—a microtubule‐organizing center—in neuronal development has been under scrutiny and is controversial. The function and position of the centrosome have been shown to play an important role in selecting the position of axon outgrowth in cultured neurons and in situ. However, other studies have shown that axonal growth is independent of centrosomal functions. Recent discoveries define the centrosome as an F‐actin organizing organelle in various cell types; thus, giving a whole new perspective to the role of the centrosome in lymphocyte polarity, cell division, and neuronal development. These discoveries compel the need to revisit centrosomal functions by investigating the fundamental mechanisms that regulate centrosomal F‐actin remodeling during neuronal differentiation and polarization. In this review, we summarize the up‐to‐date knowledge regarding the function of the centrosome in neuronal differentiation. We put special emphasis on recent findings describing the centrosome as an F‐actin organizing center. Additionally, with the available data regarding centrosome, microtubules and F‐actin organization, we provide a model on how centrosomal F‐actin could be modulating neuronal differentiation and polarity.

pericentriolar material (PCM). The centrioles are cylindrical structures each made of nine triplets of microtubules situated perpendicular to each other. High-resolution microscopy showed that the structure of the centrioles is rather dynamic and remodels during the cell cycle with dramatic structural reorganization before mitosis (Bowler et al., 2019). The PCM is a dense proteinaceous scaffold containing a large number of proteins including γ-tubulin, centrin, pericentrin, and ninein. Some of these proteins present at the PCM act as microtubule nucleators-γ-tubulin and γ-TuRCs (gamma-tubulin ring complex) (Kollman, Polka, Zelter, Davis, & Agard, 2010;Oakley, 1992;Zheng, Wong, Alberts, & Mitchison, 1995), where proteins like ninein play a role in the anchoring of microtubules (Delgehyr, Sillibourne, & Bornens, 2005;Mogensen, Malik, Piel, Boukson-Castaing, & Bornens, 2000). A nonclassical centrosomal protein, PCM-1 (Pericentriolar Matrix protein-1), that is present in the electron-dense protein granules in the cytoplasm was shown to recruit centrin, pericentrin, and ninein to build the PCM. Loss of PCM-1 disrupts the radial organization of microtubules without affecting microtubule nucleation, due to impaired recruitment of microtubule anchoring proteins to the centrosome (Dammermann & Merdes, 2002). High-resolution microscopic imaging methods revealed that the PCM is organized into two major structural domains-an inner layer juxtaposed to the centriole wall, and the outer layer with proteins extending further away from the centriole organized in a matrix-that provide separate functionality (Mennella et al., 2012). For instance, the key centrosomal proteins, such as pericentrin-like protein, pericentrin etc., form radially elongated fibrils from the centriole wall that extend outwards to support the 3D organization of the PCM by recruiting outer PCM proteins like Cnn and γ-tubulin (Mennella et al., 2012).
In most of the proliferating and migrating animal cells, the centrosome nucleates and anchors microtubules radially with their minusends oriented towards the centrosome and growing plus-ends towards the cell periphery (reviewed in Bartolini & Gundersen, 2006).
Although cells of some lower organisms do not contain centrosomes, they possess equivalent structures: in higher fungi-Saccharomyces cerevisiae, and Amoebozoa-Dictyostelium discoideum, centrosome equivalents are known as spindle pole bodies and nuclear-associated body, respectively (Juliette Azimzadeh & Azimzadeh, 2014), which can still organize microtubules in a radial fashion (reviewed in Bartolini & Gundersen, 2006). However, in the case of most fungal, somatic plant cells, and differentiated animal cells, such as muscle, epithelial, and neuronal cells, microtubules are arranged in a centrosomeindependent nonradial array (reviewed in Bartolini & Gundersen, 2006). Interestingly, cells with noncentrosomal microtubule organization are typically polarized and nonmigratory. This makes developing neurons an interesting case, for which several in vitro and in situ studies suggest a key role of centrosome-dependent radial microtubule organization in migrating and differentiating neurons (Higginbotham & Gleeson, 2007;Kuijpers & Hoogenraad, 2011;Rivas & Hatten, 1995;Schaar & McConnell, 2005;Tsai & Gleeson, 2005). It is suggested that the centrosome gradually loses its ability to radially organize microtubules after axons specification in cultured primary neurons (Stiess et al., 2010). Additionally, recent studies show that the centrosome can organize the actin cytoskeleton  during lymphocyte polarity (Obino et al., 2016), cell division (Plessner, Knerr, & Grosse, 2019), and neuronal development (Meka et al., 2019). This review is intended to summarize the data regarding the role of the centrosome during axon specification and neuronal migration with special focus on recent discoveries, which suggest the centrosome as an F-actin cytoskeleton regulator during neuronal differentiation. Due to the scope of the review, we will not deeply discuss the work related to molecular pathways regulating the microtubule and actin cytoskeleton described during axon extension.

| Neuronal differentiation in vitro
The initial attempts to study neuronal differentiation were made in dissociated hippocampal pyramidal neurons (Dotti, Sullivan, & Banker, 1988) and cerebellar granular neurons (Powell, Rivas, Rodriguez-Boulan, & Hatten, 1997). Dotti et al. (1988) observed a stereotypical pattern of hippocampal pyramidal neuronal differentiation in vitrowhich they divided into five stages. Neurons appear round in the first few hours after plating (Stage 1) followed by lamellipodia and filopodia formation. In Stage 2, several hours after plating, neurons develop 4-6 short neurites with dynamic growth cones. After extensions and retractions, one of the neurites outgrows the others (18-24 hr in vitro) to become the axon (Stage 3)-this marks the first step of polarity ( Figure 1). In the following days the remaining neurites mature into dendrites (Stage 4, after 4 days in culture) with dendritic spines (Stage 5, >2 weeks in culture) to establish synapses.
Developing in vitro cerebellar granular neurons initially form a single neurite (unipolar morphology) followed by a second neurite growing on the opposite pole of the cell body (bipolar morphology). The future axon continues to develop into a T-shaped branched process, whereas the second neurite becomes one of multiple emerging dendritic processes around the cell body (Powell et al., 1997). These in vitro studies demonstrate that axon and dendrite formations can occur without the need for a complex environment, implicating cellintrinsic mechanisms to guide neuronal polarization.

| Axon elongation
One important question to understand the process of neuronal differentiation/polarization is by which means one neurite initiates elongation as an axon. On this line, it has been shown that local actin instability in the growth cones or microtubule stability in the neurite shafts of Stage 2 neurons sustains axon elongation (Bradke & Dotti, 1999;Witte, Neukirchen, & Bradke, 2008). The neurite with more dynamic F-actin in the growth cone and more microtubules that are eventually more stable, develops as an axon, whereas the remaining neurites become dendrites (Bradke & Dotti, 1999;Geraldo, Khanzada, Parsons, Chilton, & Gordon-Weeks, 2008;Neukirchen & Bradke, 2011;Witte et al., 2008;Yu & Baas, 1994;Zhao et al., 2017). In cultured rat hippocampal neurons, the future axon was suggested to have more stable microtubules in its shaft, due to the enrichment of acetylated microtubules (Witte et al., 2008). Accordingly, global application of the MT-stabilizing drug Taxol, induced the formation of multiple axons (Witte et al., 2008). Moreover, microtubules stabilization increased F-actin dynamics in growth cones (Zhao et al., 2017). Altogether, these observations support the hypothesis that the interplay between microtubule stabilization in the axonal shaft and actin filament dynamics in the axonal growth cone set up the conditions to break the Stage 2 symmetry.
On the other hand, various studies point out the presence of stable microtubules in axons-determined either by measuring acetylated (Witte et al., 2008) or detyrosinated microtubules (Arregui, Busciglio, Caceres, & Barra, 1991). Yet, it is not completely clear which of the post translational modifications (PTMs) stabilize microtubules. Selective translocation of the Kinesin-1 motor domain into the nascent axon was described as one of the earliest events during the axon elongation (Jacobson, Schnapp, & Banker, 2006). Tubulin acetylation and detyrosination were shown to be required for Kinesin-1 motor domain translocation into the axons (Konishi & Setou, 2009;Reed et al., 2006). A later study, however, showed that acetylation of microtubules itself is not enough for sorting of Kinesin-1 into axons (Hammond et al., 2010). Moreover, taxol-induced translocation of  It was well described that polyamination, catalyzed by transglutaminase, another form of PTM, confers stability to the microtubules (Song et al., 2013). Whereas acetylation and detyrosination accumulate on long-lived microtubules (Garnham & Roll-Mecak, 2012;Song & Brady, 2015). The polyaminated microtubules are the most stable pool that are presumably acetylated and detyrosinated as well.
While acetylated and detyrosinated microtubules are stable, they are still dynamic but not as dynamic as the unstable or labile pool of microtubules (reviewed in Baas, Rao, Matamoros, & Leo, 2017). Overall, it seems that axons contain a higher percentage of stable microtubules compared to their dendritic counterparts (reviewed in Baas et al., 2017). However, more research is needed to unveil the exact mechanisms of how differential microtubule stability is achieved in axons versus other minor neurites (future dendrites). In addition to the centrosome, the Golgi apparatus can nucleate microtubules, too (Chabin-brion, Perez, Drechou, & Pou, 2001;Efimov et al., 2007). Golgi protein GM130 recruits AKAP450 (a γ-TuRCinteracting protein) to the cis-side of Golgi membranes to promote microtubule nucleation. Thus, GM130 allows centrosome-associated nucleating activity to extend to the Golgi, to ensure specific functions within the Golgi or for sorting specific cargos to the cell periphery (Rivero, Cardenas, Bornens, & Rios, 2009). Given that Golgi and centrosome colocalize during initial axon formation/specification (de Anda et al., 2005;Zmuda & Rivas, 1998), it is plausible to envision one of them acting as a compensatory factor while the function of the other is disrupted. Clear mechanistic data is still lacking to dissect specific functions of the centrosome and the Golgi to better understand their independent contributions to axon specification. Although, these results suggest a cell-intrinsic mechanism behind axon selection, the possibility of external cues playing a role cannot be discarded.

| Centrosomal and noncentrosomal factors contributing to axon specification
In this regard, N-cadherin, a membrane-bound adhesion molecule, clusters at the site of first neurite formation, which then triggers the accumulation of F-actin, centrosome, and Golgi. Extracellular Ncadherin is capable of determining the site where the first neurite emerges. The centrosome and Golgi apparatus are then guided towards this site (Gärtner et al., 2012;Gärtner, Fornasiero, Valtorta, & Dotti, 2014). Similarly, laminin-a cell-associated extra cellular matrix protein-attracts the centrosome to the site of axon formation in cultured retinal ganglion cells from zebrafish (Randlett, Poggi, Zolessi, & Harris, 2011). Therefore, the site of axon formation can be predicted by the positioning of centrosome and Golgi apparatus and extracellular substrates could attract them to the site of axon specification.
Altogether, these studies suggest an interplay between external cues and intracellular organization of the cell during initial polarization.
In contrast, several studies suggest a stochastic axon selection process, in which one of the minor neurites of a Stage 2 cell elongates and becomes an axon under the stimulus of extrinsic factors (Andersen & Bi, 2000;Bradke & Dotti, 2000). The axon specification is positively regulated by extracellular substrates such as laminin and neuron-glia cell adhesion molecules (Esch, Lemmon, & Banker, 1999). Similarly, Netrin-1, a chemoattractant, induces axon outgrowth by modulating Factin in the growth cone via Pak1-mediated Shootin1 phosphorylation (Toriyama, Kozawa, Sakumura, & Inagaki, 2013). However, these studies did not clarify, whether or not the extracellular substrates relocate the centrosome and/or Golgi to the site of axon specification. Therefore, it is possible that the organelle (centrosome/Golgi) positioning is vital to sustain the trafficking of material required to promote initial axon specification/growth, but not during extension or regeneration.
Once the axon is growing, the organization of microtubule arrays shifts from centrosome-dependent to centrosome-independent in hippocampal neurons in culture. Centrosome ablation during axon elongation does not affect axon extension or regeneration (Stiess et al., 2010).
The same study reported decentralization of centrosomal proteins, such as γ-tubulin, which in turn can organize acentrosomal microtubules in older neurons (Stiess et al., 2010). Along these lines, Augmin and γ-TuRC are shown to be crucial for uniform plus-end out microtubule polarity in axons of matured neurons (Sánchez-Huertas et al., 2016).
CAMSAP2, a microtubule minus-end protein can establish and maintain noncentrosomal microtubule networks and is suggested to be crucial for axon specification and polarity of neurons in vivo and in vitro (Yau et al., 2014). However, recently it has been described that CAMSAP3, but not CAMSAP2, affects axon formation (Pongrakhananon et al., 2018). CAMPSAP3 regulates microtubules stability and its absence promotes microtubules acetylation leading to the formation of multiple axons. In contrast, the lack of CAMSAP2 did not affect neuronal polarity (Pongrakhananon et al., 2018) as suggested earlier (Yau et al., 2014).

| Centrosome and neuronal development in situ: From genesis to differentiation
Multiple studies describe how the centrosome's position needs to be precisely regulated during neurogenesis, migration, and differentiation.
Neurons either migrate radially (e.g., cortical pyramidal neurons) or tangentially (e.g., cortical interneurons) to find their final position in the brain . Centrosome positioning along with microtubule and actin cytoskeleton dynamics plays an important role during neuronal migration (Higginbotham & Gleeson, 2007;Kuijpers & Hoogenraad, 2011;Schaar & McConnell, 2005;Tsai & Gleeson, 2005). Neurons follow a two-stroke saltatory movement during migration. This movement is characterized by a repetitive pattern of rapid extension and retraction of the leading process, which stabilizes 10 of microns ahead of the soma, followed by forward displacement of the centrosome into the leading process. Subsequently, the nucleus and soma translocate forward with concurrent retraction of the trailing process. During this process, the microtubules emanate from the centrosome, anteriorly into the leading process and posteriorly to envelope the nucleus to form a perinuclear tubulin cage (Rivas & Hatten, 1995). Live in situ imaging of organotypic mouse cortical and cerebellar slices showed that the centrosome first translocates to the leading process and displaces the nucleus forward along the microtubule network (Solecki, Model, Gaetz, Kapoor, & Hatten, 2004;Tsai, Bremner, & Vallee, 2007;Vallee, Seale, & Tsai, 2009).
Along these lines, mPar6 overexpression disrupts the perinuclear tubulin cage and inhibits centrosomal motion as well as neuronal migration in mouse cerebellar granule neurons (Solecki et al., 2004;Solecki, Govek, & Hatten, 2006). These data establish the centrosome as a functional link between the microtubule-based pulling forces generated in the leading process and the perinuclear microtubule network. Moreover, proper positioning of the centrosome in facial bronchio motor neurons of zebrafish seem to be important for their tangential migration (Grant & Moens, 2010).
However, studies from mouse cerebellar granule cells (Umeshima, Hirano, & Kengaku, 2007) and zebrafish neurons from tegmental hindbrain nuclei (Distel, Hocking, Volkmann, & Köster, 2010) differ with the centrosome-dependent pulling mechanisms of the perinuclear-cage during migration, suggesting a centrosomeindependent nuclear movement. Another study from cortical neurons in situ shows that Myosin II motors-dependent F-actin drives the coordinated movement of centrosome and soma during neuronal migration (Solecki et al., 2009). Although the presence of the actin cytoskeleton underlying the membrane in the leading and trailing process is already known in migrating neurons (Tsai & Gleeson, 2005), the data from Solecki et al. (2009) attributes an important function of the actin cytoskeleton in centrosome dynamics during neuronal migration.
In the developing cortex, neuronal migration and axon specification are intertwined, meaning that the axon specification happens while neurons migrate (Noctor, Martinez-Cerdeño, Ivic, & Kriegstein, 2004). Neurons originate in the ventricular zone/intermediate zone, where they generally have a multipolar shape. Eventually, they change to a bipolar morphology and migrate towards the cortical plate to stop migration ( Figure 2) and form different layers in the cortex. Early electron microscopic studies on mouse cortex, in situ, revealed that the centrosome is generally located at the origin of the extending axon in the multipolar neurons of the intermediate zone, which projects tangentially or towards the ventricular zone (Shoukimas & Hinds, 1978).
However, in more mature neurons already located in the cortical plate, the centrosome can be found in proximity to the apical/leading process or future apical dendrite, which is oriented towards the cortex surface (Shoukimas & Hinds, 1978). This initial observation suggests that the centrosome is dynamic during neuronal differentiation in the developing cortex. Indeed, live-imaging of multipolar cells in the intermediate zone demonstrated that the centrosome translocates transiently to the site of axon formation before or at the time of initial axon outgrowth (de Anda, Meletis, Ge, Rei, & Tsai, 2010). These observations were confirmed by another study (Sakakibara et al., 2014), that described centrosome dynamics during migration and axon specification ( They showed that laser-induced centrosome ablation (before axon formation) inhibits axon outgrowth in these neurons (Andersen & Halloran, 2012). In the same study, centrosome position and motility are shown to be regulated by LIM homeodomain transcription factor activity, which is specifically required for the development of peripheral axons in the Rohon-Beard sensory neurons. Furthermore, centrosome mis-localization is correlated with ectopic axon formation in laminin alpha-1 mutants (Andersen & Halloran, 2012).
In some bipolar neurons located in the cortical plate of the developing cortex, however, axon formation occurs at the opposite pole to the centrosome position. The centrosome of those bipolar cells is already at the base of a pia-directed leading process or future apical dendrite (Sakakibara et al., 2014), as described in zebra fish (Distel et al., 2010;Zolessi, Poggi, Wilkinson, Chien, & Harris, 2006). Altogether, the research from various biological systems confirms the fol- In contrast, Drosophila devoid of centrioles (mutants of Sas-4 and Sas-6) develop normally with typical axon trajectories (Basto et al., 2006;Peel, Stevens, Basto, & Raff, 2007). Interestingly, other studies show a PCM structure that is still present in those flies without centrioles, which is enough to form a polarized microtubule array (Baumbach, Novak, Raff, & Wainman, 2015;Pollarolo, Schulz, Munck, & Dotti, 2011). Moreover, acute ablation of centrioles does not affect the pericentriolar material (Cabral, Laos, Dumont, & Dammermann, 2019). Altogether the data from various in situ studies suggest an indispensable role of the centrosome, or at least a functional PCM, for axon specification.
More recently, an anterograde F-actin flow was described during neuronal migration (He, Zhang, Guan, Xia, & Yuan, 2010;Solecki et al., 2009) and at the base of the growth cones (Burnette et al., 2008). Importantly, using high-resolution microscopy techniques, it was demonstrated that mature axonal shafts contain actin "hotspots" and "trails," which contribute to axon anterograde actin transport (Chakrabarty et al., 2019;Ganguly et al., 2015). Altogether, these studies suggest a strong centrifugal component to the neuronal F-actin organization, in addition to the existing local assembly of filaments of F-actin in the growth cones.
Recent studies identified the centrosome as an F-actin organization center in vitro , with implications in lymphocyte polarity Obino et al., 2016) and cell division (Plessner et al., 2019). The centrosome nucleates actin filaments in vitro in an Arp2/3 complex, WASH nucleation promoting factor, and PCM-1 mediated-manner . On this regard, a following study showed that in resting lymphocytes, the centrosomal Arp2/3-dependent F-actin nucleation is required for centrosome tethering to the nucleus (Obino et al., 2016). In activated lymphocytes, however, the centrosomal F-actin nucleation is downregulated to facilitate the centrosome polarization to the immune synapse (Obino et al., 2016). More recently, centrosomal actin assembly was shown to be crucial for proper mitotic spindle formation and chromosome congression, which is required for maintenance of genomic integrity during mitosis (Plessner et al., 2019). These recent studies raise the necessity to investigate the function of centrosomal F-actin organization in neurons.
Accordingly, our recent work uncovered a somatic F-actin organization concentrated around the centrosome in developing cultured neurons and in situ (Meka et al., 2019). The F-actin in the soma of neurons appeared as dynamic aster-like structures, closely associated with PCM-1 satellites (Figure 3). Furthermore, with Live-STED (stimulated emission depletion) microscopy, we discovered constant extensions of F-actin fibers from these dynamic F-actin asters (Meka et al., 2019). Interestingly, the F-actin asters in the soma of developing neurons (Meka et al., 2019) and the so-called actin hotspots in axon shafts of more matured neurons (Chakrabarty et al., 2019;Ganguly et al., 2015) are Formin but not Arp2/3-dependent. Using a photoactivatable F-actin probe, PaGFP-UtrCH, (Burkel, Bement, & Labs, 2015), we were able to show an anterograde F-actin movement from the soma to the growth cones/neurite tips. Photoactivation of F I G U R E 3 Illustration of the radial somatic F-actin organization in a developing neuron and how it is affected upon on PCM-1 knockdown and its implications in neuronal polarization (a hypothetical idea). F-actin organized around the centrosome is a source of peripheral F-actin, which helps to modulate growth cone dynamics (Meka et al., 2019). PCM-1 knockdown (that affects PCM protein recruitment and disturbs radial MTOC organization-discussed in the introduction section) results in the reduction of somatic F-actin asters and promotes neurite elongation due to decreased F-actin delivery to the periphery (left side). We hypothesize that the neurite elongating more during symmetry breakage is receiving less F-actin compared with the neurites that are not growing ( Cofilin, DrebrinS142D and CofilinS3E, (Garvalov et al., 2007;Worth, Daly, Geraldo, Oozeer, & Gordon-Weeks, 2013)  Whereas, Semaphorin 3A acts as an inhibitory signal that suppresses axonal growth (and thus promotes dendritogenesis) in cultured hippocampal neurons (Shelly et al., 2011). Long-range inhibitory signaling mediated by Ca 2+ waves suppresses outgrowth of minor processes by activating RhoA, and thereby ensuring neuronal polarization (Takano et al., 2017). These data suggest that positive feedback signals are continuously activated in one of the minor neurites resulting in axon specification and elongation. In turn, negative feedback signals are propagated from a nascent axon terminal to all minor neurites to inhibit the outgrowth of all neurites at the same time, thereby leading to an asymmetric growth. All the downstream signaling mechanisms suggested for the positive and negative feedback signals seem to converge at two fundamental aspects related to cytoskeleton dynamics. Our recent experiments offer a novel point of view for this conundrum, given that our photoactivation experiments with PaGFP-UtrCH show that the intensity of the photoactivated somatic F-actin signal reaching the periphery is inversely proportional to the length of the neurites (Meka et al., 2019). Therefore, F-actin in growth cones could act as a growth-inhibiting factor. Along these lines, removal of F-actin from growth cones/neurite tips by pharmacological means promotes neurite outgrowth (Bradke & Dotti, 1999;Forscher, 1988).
Moreover, photoactivation of PaGFP-UtrCH in PCM-1 downregulated cells show that the intensity of the photoactivated F-actin signal moving from soma to the periphery is significantly reduced, compared to the control counterparts (Meka et al., 2019). These findings could explain the F-actin content depletion observed in growth cone/neurite tips of the PCM-1 downregulated neurons described above. Altogether, it is possible to envision that impaired somatic F-actin delivery in PCM-1 knockdown cells could be the reason for the excessive neurite extension observed in these neurons ( Figure 3; Meka et al., 2019).
Based on these observations, we hypothesize that the somatic organization of F-actin around the centrosome preferentially provides Factin to peripheral places, where the outgrowth should be limited (eventually the shorter neurites, Figure 3). Thus, acting as an inhibitory or a pulling force, which suppresses neurite growth. On the other hand, places receiving less F-actin, but more microtubules (Baas, 1999;Yu & Baas, 1994), should be primed for growth, leading for asymmetrical outgrowth/axon formation.

| CONCLUSION
An important question in neurobiology is how neurons achieve distinct functional domains, one axon and several dendrites, which allow them to connect properly? Axon specification is the key event deter- nisms are now the challenges ahead. Moreover, it will be interesting to know the underlying mechanism by which the centrosome plays an antagonizing role in delivering more microtubules and less F-actin to a growing neurite. The emerging microscopic technologies with high spatial and temporal resolution could unravel these unknown aspects of centrosome-cytoskeleton organization/dynamics for a better understanding of their role in neuronal polarization and development.