Shaping neurodevelopment: Distinct contributions of cytoskeletal proteins

Summary

Development of a neuron critically depends on the organization of its cytoskeleton. Cytoskeletal components, such as tubulins and actins, have the remarkable ability to organize themselves into filaments and networks to support specialized and compartmentalized functions. Alterations in cytoskeletal proteins have long been associated with a variety of neurodevelopmental disorders. This review focuses on recent findings, primarily from forward genetic screens in Caenorhabditis elegans that illustrate how different tubulin protein isotypes can play distinct roles in neuronal development and function. Additionally, we discuss studies revealing new regulators of the actin cytoskeleton, and highlight recent technological advances in in vivo imaging and functional dissection of the neuronal cytoskeleton.

Introduction

Neuronal development is a multi-step process from cell proliferation and migration, formation of synapses and establishment of proper cell morphology. A highly organized cytoskeletal network, composed of microtubules (MTs), actin and intermediate filaments (IFs), is required for establishment and maintenance of neuronal cell morphology. Cytoskeletal proteins are encoded by large gene families, individual members of which have distinct, overlapping, or redundant functions. While it has been proposed that isotypes of cytoskeletal proteins might play specific roles in neuronal trafficking and function [1], in vivo evidence remains sparse. Owing to the defined neuronal connectome and technical ease of in vivo single-cell imaging, the nematode Caenorhabditis elegans is an excellent model in which to dissect the function of specific cytoskeletal protein isotypes [2,3]. Here, we discuss recent studies that provide unique insights into how the neuronal MT and actin network are regulated in neurodevelopment. We will compare and contrast the findings from C. elegans to those from relevant human diseases. For more comprehensive discussion on the neuronal cytoskeleton and intracellular trafficking in growth cones and synapses, we refer the readers to several recent reviews [4–8].

in vivo evidence for a ‘tubulin code’ in regulating neuronal microtubule organization and dynamics

MTs are polarized filamental polymers formed by αβ-tubulin heterodimers [9]. Since the molecular cloning of tubulin genes 40 years ago [10], our understanding of tubulins at the cellular levels continues to reveal their immense complexity. All known animal genomes express multiple genes encoding different isotypes of α- and β-tubulin, which undergo a variety of post-translational modifications (PTMs). The diversity of tubulin isotypes and PTMs has led to the proposal of the ‘tubulin code’, in which the behavior of a MT filament depends on its composition of tubulin isotypes and their PTMs [11,12]. While the structure of MTs is resolved at atomic resolution [13], the endogenous composition of individual MT filaments within a cell remains to be challenging to address.

The C. elegans genome has nine genes encoding α-tubulins (TBA) and six encoding β-tubulins (TBB), which display broad or selective expression in different tissues. Protein sequence differences between tubulin isotypes are mostly in the Carboxy-Terminal Tail (CTT). This domain mediates interactions with MT-associated proteins (MAPs) and motor proteins [14–16]. A recent study used CRISPR-Cas9 genome editing to examine the contribution of tubulin isotypes to MT dynamics in C. elegans embryos. By genetically varying the ratio of each tubulin isotype, the authors identified distinct effects on MT growth rate, stability and dynamics in the one-cell embryos [17]. This observation provides in vivo evidence that diversity of tubulin protein isotypes contribute to MT functions, supporting the tubulin code hypothesis. Several recent studies have begun to address neuron-type and function-specific regulation of individual tubulin isotypes and of PTMs.

Tubulin isotype in mechanosensory neurons differentially affects neuronal development and mechanosensation

Neuronal morphogenesis is a dynamic process involving reversible or irreversible changes in the MT cytoskeleton that can occur in parallel or sequential manner. An early study using mammalian PC12 cells revealed distinct β-tubulin isotypes were selectively expressed, assembled and localized upon differentiation [18]. Subsequent studies in other types of cultured cells also show that distinct isotypes of α- and β-tubulin can be co-expressed in the same neurites and differentially incorporated into MT filaments [19]. Although these studies provide evidence that multiple tubulin isotypes are present in the neurons, the in vivo functions of distinct tubulin isotypes in neurodevelopment remain to be challenging to study. Several recent studies in C. elegans address the in vivo functions of distinct tubulin isotypes in neuronal MT organization and remodeling.

In C. elegans most neurons display unipolar or bipolar morphology. A long-standing experimental model for studies of MTs has been the MT-rich mechanosensory neurons (touch receptor neurons, TRN) (Figure 1A). TRNs extend long axons filled with uniformly organized MT bundles of unusual 15-protofilament (pf), made of MEC-12/α-tubulin and MEC-7/β-tubulin, both of which are necessary for sensing gentle touch [20,21]. TRNs also have classical 11-pf MTs, which do not form bundles and are overall less common than 15-pf MTs [22,23]. Genetic mutants lacking mec-7 function [mec7(0)] display a total loss of touch sensation, but not a total loss of MTs in TRNs. Instead, mec-7(0) mutants display fewer 15-pf MTs and more 11-pf MTs. The 11-pf MTs are likely composed of other tubulin proteins, including at least three TBAs and two TBBs [24]. However, none of the tba(0) or tbb(0) mutants show touch insensitivity resembling that of mec-7(0) [24]. While the exact function of 15-pf MTs remains an ongoing topic of research, these studies support a correlation between 15-pf MTs and touch sensitivity. It remains unaddressed how MEC-7 contributes to the formation of 15-pf MTs in TRNs. MEC-7 shares 90% amino acid sequence identity with TBB-1 and TBB-2, with major differences in the CTT exposed to the exterior surface of MTs, suggesting its differential incorporation might involve distinct interactions with MAPs or PTMs.

Figure 1. Cell biology and microtubule composition of cilia and mechanosensory neurons.

(A) Schematic of cephalic male (CEM) ciliary neurons, BDU interneurons and mechanosensory ALM and PLM neurons in wild type C. elegans.

(B) Top: Schematic of mechanosensory neurons in β-tubulin/mec-7 mutants. Bottom: Distinct mutations in tubulin affect microtubule (MT) function differently. Missense mutations predicted to interfere with tubulin polymerization (orange) resulted in reduced MT stability and defective PLM neurite outgrowth. Missense mutations affecting surface residues of the MTs (purple) may interfere with MT-associated protein or motor binding, leading to increased MT stability and ectopic ALM neurite outgrowth. The structure of the αβ-tubulin dimer (1jff.pdb) is shown on the right. Orange and purple stars mark the positions of representative mutations in the tubulin structure that affect neurite outgrowth.

(C) Schematic of MT doublet in C. elegans CEM neurons. The MT doublet consists of an A-tubule with 13-pf MTs and a B-tubule with 10-pf MTs.

(D) The tubulin glutamylase TTLL-11 and deglutamylase CCPP-1 regulate tubulin glutamylation in CEM neurons. Both hyperglutamylation in ccpp-1(0) mutants or hypoglutamylation in ttll-11(0) mutants result in reduced extracellular vesicle (EV) release.

Forward genetic screens following genome-wide random mutagenesis are powerful tools to discovering missense mutations that offer unique insights into the role of specific amino acid residues in a given protein. The initial screens recovering tubulin mutants such as mec-7 and mec-12 were based on behavioral assays or drug resistance to benzimidazole, and most of these mutants display normal TRN development [20,23,25–27]. Recently, two forward genetic screens for aberrant TRN morphology have revealed distinct roles of tubulin isotypes. Zheng et al isolated mutants with abnormal TRN axon morphology, and also re-examined previously isolated touch insensitive mutants [28]. This systematic effort analyzed the effects of >60 missense mutations in the two tubulin genes, MEC-12/α-tubulin and MEC-7/β-tubulin, on neurite outgrowth. Additionally, this study uncovered a role of another α-tubulin/TBA-7, in destabilizing TRN MTs, as tba-7(0) mutants show hyperstable MTs and ectopic neurite formation [28]. mec-7 missense mutations mimicking the null phenotype are located throughout the protein, potentially interfering with tubulin folding. However, missense mutations in the GTP-binding and intradimer-interacting regions are predicted to interfere with tubulin polymerization, resulted in reduced MT stability and defective neurite outgrowth [28]. On the other hand, missense mutations affecting surface residues of the MTs, which may interfere with MAP binding, increased MT stability and induced ectopic neurite outgrowth (Figure 1B) [28]. TRNs in these mutants have larger-diameter 15-pf MTs, which are bundled with closer spacing than the wild type, suggesting PTMs and MAP binding at the MT exterior surface may influence changes in MT organization.

Mutations in TRN tubulins were also discovered independently in a screen for ribosome localization defects using RIBOS (tissue specific Ribosome Imaging Based On Split GFP), a novel method to visualize ribosomes in specific cell types (Figure 2) [29]. Tagging of the last β strand of GFP (GFP11) to endogenous ribosomal proteins does not produce green fluorescence nor interfere with ribosomal functions. Co-expression of GFP11 with first ten β-strands of GFP (GFP1-10) in specific tissues allows reconstitution of the split GFP and fluorescence. In wild type larvae, axons of developing TRNs show fluorescently labeled ribosomes as small puncta with notable concentration at the synaptic branch of PLM neurons. Upon maturing to adulthood, axonal ribosome puncta decrease in number and are no longer detected at the axonal branch, indicating a change of ribosomal localization during neuronal maturation. By screening for mutants displaying altered ribosomal distribution in TRNs, Noma et al. identified several missense mutations in mec-12 and mec-7 that show misshapen soma with altered axonal and somatic localization of ribosomes [29]. Most of the mec-7 mutations disrupt the GDP/GTP binding sites, suggesting roles of microtubule polymerization in regulation of ribosome dynamics in the TRNs. Compartmentalized ribosomes have long been proposed to act in local protein synthesis [30]. Altered ribosomal localization in tubulin mutants may affect local mRNA translation, thereby affecting the development of neurons.

Figure 2. Recent advances in microtubule imaging and study of cell type specific gene functions.

(A) Expression of GFP::TBA-1/α-tubulin (MTs) and RFP::PTRN-1 (MT minus-end) proteins allows live imaging of MTs in vivo. MT length and spacing can be reconstructed following line scan of GFP::TBA-1 signals and visualized as a model to represent MT coverage and length [67].

(B) Schematic of ribosome imaging based on split GFP (RIBOS). A ribosomal protein is fused to the smaller part of split GFP (GFP11), whereas the larger part (GFP1-10) is expressed in specific cells (e.g. mechanosensory neurons). Upon coexpression of both transgenes, binding of GFP1-10 to GFP11 in the targeted cell allows visualization of ribosomes [29].

(C) Schematic of cell-specific degradation of GFP-tagged protein. A protein of interest is tagged with GFP, whereas the GFP nanobody-SOCS fusion gene (degron) is expressed in specific cells (e.g. mechanosensory neurons). Upon crossing of these strains, the degron binds to GFP-tagged proteins, promoting their ubiquitylation and subsequent protein degradation by the proteasome [71].

The findings from both studies underscore the importance of careful analyses of missense mutations in understanding tubulin functions, which would otherwise be difficult to uncover using null mutants. Many missense mutations in human tubulin genes are linked to neurodevelopmental diseases such as microcephaly and cortical malfunction [31,32]. As the impact of individual missense mutations in human neurons is difficult to analyze at the cellular level in vivo, C. elegans offers an excellent opportunity to address their effects. For example, introduction of disease-associated human β-tubulin mutations, located at the GDP/GTP-binding sites or at the outer surface of the tubulin, into the mec-7 gene using CRISPR/Cas9 genome editing affects TRN development in C. elegans [28]. This observation highlights the potential utility of C. elegans studies to understand neurological diseases.

Tubulin isotypes contribute to intraflagellar transport in ciliary sensory neurons

Many sensory neurons rely on cilia to sense and respond to the environmental cues. In C. elegans cephalic male (CEM) neurons, ciliary MTs form doublets consisting of complete 13-pf A-tubules attached to incomplete 10-pf B-tubules [33]. These doublet MTs splay in the middle to form A-tubule and B-tubule singlets and rejoin at the tubular terminal to form the A/B-tubules (Figure 1C). CEMs express at least two α-tubulins, TBA-6 and TBA-9, and one β-tubulin TBB-4 (Figure 1A) [34]. In tba-6(0), tba-9(0) or tbb-4(0) mutants, cilia form but CEM neurons show reduced sensory responses and mislocalized signaling molecules [34]. A recent study of tba-6(0) mutants found an altered ciliary MT ultrastructure in which the A/B-tubules fail to splay in the middle (Figure 1C) [35]. Further, loss of tba-6 also affects intraflagellar trnsport (IFT) and contents of the extracellular vesicle released. Effects of mutations in other tubulin genes, such as tba-9 and tbb-4, to the ciliary MT ultrastructure remain to be examined. TBA-6 has a longer CTT compared to other α-tubulin isotypes, hinting at possible roles of PTMs or MAP binding.

Tubulin glutamylation regulates ciliary MT organization

Tubulins undergo multiple PTMs including tyrosination, polyglutamylation and acetylation [36]. It is well accepted that PTMs can have profound effects on the stability and specific functions of MTs [14]. Among PTMs, polyglutamylation entails addition of glutamate side chain to the tubulin CTTs. Extensive biochemical studies have shown that tubulin tyrosine ligase-like (TTLL) family proteins are responsible for tubulin polyglutamylation [37], whereas the cytosolic carboxypeptidase (CCP) family members regulate deglutamylation [38]. Recent studies in C. elegans have used mutants lacking these enzymes to dissect how tubulin PTMs affect neuronal development.

In C. elegans CEM neurons, loss of the tubulin deglutamylase CCPP-1 results in excessive glutamylation of ciliary MTs, leading to disorganization of MTs (Figure 1D) and defective transport in cilia [39]. These phenotypes are suppressed by the loss of the polyglutamylase TTLL-11 [40], supporting the roles of CCPP-1 and TTLL-11 in fine-tuning glutamylation levels in cilia MTs. Importantly, both hypo- and hyper-glutamylation result in reduced extracellular vesicle release, suggesting tight control of glutamylation in the neuron (Figure 1D). The MT-severing enzyme Spastin is recruited to polyglutamylated MTs [41]. A recent study purified human MTs lacking PTMs and used TTLL7 glutamylase to modify CTTs in vitro showed that affinity of Spastin to MTs increases as the level of glutamylation increases [42]. Surprisingly, the MT-severing activity of Spastin also decreases in hyper-glutamylated MTs, implying that Spastin can have MT-severing or MT-stabilizing activity depending on the extent of glutamylation.

Intermediate filaments contribute to MT dynamics regulation

In C. elegans another striking example of circuit plasticity concerns six GABAergic ventral cord Dorsal D (DD) motor neurons, which in adults form neuromuscular junctions (NMJs) to dorsal body wall muscles, with their ventral processes receiving inputs from the cholinergic motor neurons [43]. However, in L1 animals DD neurons form NMJs to ventral body wall muscles, receiving cholinergic inputs in their dorsal processes [44]. This dramatic re-organization of DD synapse connectivity, known as “DD remodeling”, has provided a vibrant model to examine cytoskeleton in neural plasticity. Imaging of MT dynamics using GFP-tagged plus end-binding protein (EBP::GFP) revealed that in DD neurons the number of dynamic MTs increases significantly at the onset of DD remodeling, and requires temporal activation of the conserved MAPKKK DLK-1 in conjunction with changes in MT cytoskeleton. The increased MT dynamics during DD remodeling facilitates synaptic vesicle transport mediated by plus-end directed kinesin motors and minus-end directed dynein [45,46]. DD neurons express TBA-1 α-tubulin and TBB-2 β-tubulin. In the absence of dlk-1, a gain-of-function mutation in tba-1 blocks the up-regulation of MT dynamics [45]. By analyzing genetic suppressors in the background of tba-1 dlk-1 double mutants, a new finding reports an unexpected role for IFs in regulating MT dynamics. IFs are comprised of different subunits and are generally thought to provide structural support for the cell; for example, in neurons, neurofilaments (NFs) maintain axon caliber in large diameter myelinated axons. In DD remodeling, accumulation of IFs is observed concurrent with hyperstablization of MTs, which in turn prevents synaptic vesicle transport during rewiring (Kurup et al., 2018, in press). In some mouse models, overexpression of NFs impairs axonal transport, which is thought to be a precursor to neuron degeneration [47]. Additionally, changes in NF levels have also been associated with neuropsychiatric disorders such as bipolar disorder and schizophrenia [48]. Thus, these observations underscore furthering understanding of the complex relationship between various cytoskeletal elements in vivo.

New regulators of actin cytoskeleton during neurodevelopment

An axon navigates and migrates before forming synapses with its targets during development. These processes involve formation of motile growth cones and neurite branches, in which local assembly of actin cytoskeleton pushes the membrane forward. The assembly of filamentous actin (F-actin) from actin monomers is facilitated by the Arp2/3 complex, whereas the filament-severing protein cofilin and the F-actin binding protein coronin regulate F-actin disassembly [49]. Disruptions in actin or its associated proteins lead to neurological diseases such as periventricular heterotopia [50]. During neuronal development, reorganization of the actin cytoskeleton depends on extracellular cues, such as Wnt, Ephrin, Bone morphogenic protein (BMP) and Sonic hedgehog (Shh) signaling pathways [51–54]. Several recent studies in C. elegans have brought new insights into how Wnt signaling pathways regulate actin organization during neurodevelopment.

A systematic analysis of TRN morphology in Wnt mutants in C. elegans led to the finding that the secreted Wnt CWN-1 and EGL-20 inhibit F-actin assembly through their receptor MIG-1/Frizzled and the planar cell polarity protein (PCP) VANG-1 [55]. The Wnt-PCP signaling pathway restricts F-actin assembly at future neurite branching sites through inhibition of the small GTPases CED-10/Rac and MIG-2/Rho. Rac and Rho are recruited to the membrane to promote formation of membrane-associated actin cytoskeleton and drive axon growth and branching [56,57]. How these small GTPases are recruited to the membrane remains poorly understood. The C. elegans PLM neurons form gap junctions with a pair of interneurons called BDU (Figure 1A), a process that also involves Wnt signaling [58]. A genetic screen for defective BDU-PLM connection identified a calponin homology (CH) domain containing protein CHDP-1 that promotes cell protrusions [59]. The CH domain is generally thought to bind F-actin [60]. Surprisingly, CHDP-1 does not bind to actin directly, but interacts with CED-10/Rac through its CH domain, thereby enhancing CED-10 localization to the cell cortex to regulate actin dynamics [59].

Upon receiving extracellular stimuli, such as Wnts, neurons activate the interaction of Wiskott-Aldrich Syndrome protein (WASP) and WASP-family verpolin-homologous protein (WAVE), which then interact with transmembrane proteins and transduce the extracellular stimuli to activate Arp2/3-mediated actin nucleation [61]. Many transmembrane proteins, such as MIG-13/Lrp12, regulate actin dynamics through the WASP/WAVE pathway. Loss of Lrp12 in mouse primary neuron culture causes impaired neurite branching [62]. However, the mechanistic connections between MIG-13/Lrp2 and actin cytoskeleton remain elusive. Recently, Zhu et al reported a mechanism in which the adaptor proteins ABL kinase/ABL-1 and GRB2/SEM-5 mediate interaction between MIG-13 and WASP/WAVE proteins in neuroblast migration in C. elegans [63]. MIG-13 interacts with and recruits ABL-1 and SEM-5 to the leading edge of the neuroblast, thereby promoting Arp2/3- mediated actin nucleation through WVE-1/WAVE and WSP-1/WASP, respectively. These two pathways function in parallel and semi-redundantly to promote neuroblast migration.

Advances in technology for microtubule imaging and cell-specific gene and organelle functions

Classical studies of the TRN cytoskeleton using chemical fixation and serial section electron microscopy (EM) provided crucial insights into MT organization [64]. Live imaging using EBP::GFP has been used to examine MT dynamics in C. elegans [65,66]. However, determining MT spacing, length and shrinkage in vivo remains problematic. Recently, an in vivo analytic method was developed, using optimized expression of GFP-TBA-1 to measure MT length, in addition to MT dynamics (Figure 2A) [67]. Further, by imaging cargo transport and computationally reconstructing MT organization, the authors found that axonal transport pauses at ends of a MT filament before switching to a new polymer.

With fluorescent protein tagging technology, in vivo imaging of axonal transport is now routine. However, as multiple motors can transport a cargo, an interesting question is how to determine the exact contribution of specific motors in vivo. A new method took advantage of the blue light-inducible heteromerization of LOV peptide and engineered PDZ (ePDZ) [68]. Transgenic animals are generated to express cargoes tagged with the LOV peptide and motor proteins tagged with ePDZ. Upon blue light exposure, binding of LOV to ePDZ allowed studying of specific contributions of the motor proteins in axonal transport [68]. As CRISPR/Cas9 technology enables editing of endogenous genes in C. elegans [69,70], a strong desire in the future is to label endogenous proteins to avoid altering ratio of cytoskeletal or motor proteins in vivo.

As many cytoskeletal proteins involved in neurodevelopment also play essential functions during earlier developmental processes or in multiple cell types, it is often difficult to discern their specific roles during neurodevelopment. Wang et al developed an in vivo knock-down system to degrade GFP-tagged proteins in a cell-specific manner [71], by fusing GFP nanobody with a SOCS-box protein that interacts with the E3 ubiquitin ligase (Figure 2C). Expression of GFP nanobody::SOCS in specific cell promotes ubiquitylation of GFP-tagged proteins and subsequent protein degradation by proteasome. Application of this technique in germline- and epidermal-specific tissues has identified the roles of NOCA-1/ninein and γ-tubulin in regulation of non-centrosomal microtubule arrays [72]. We anticipate that application of this technique in neuron tissues will allow systematic dissection of cytoskeletal functions in neurodevelopment.

Conclusion

Recent advances provide exciting opportunities to dissect how subtle alterations in the composition of the cytoskeleton can affect neuronal development, maintenance and function. Moreover, the ease of genome editing further empowers C. elegans as an experimental model for neurodevelopmental diseases since many are associated with missense mutations in conserved residues. By combining live imaging, genetic screening and genomic editing in C. elegans, we foresee new regulators of cytoskeletal dynamics will be described and our understanding of neurodevelopment will be expanded.