Primary Cilia and Dendritic Spines: Different but Similar Signaling Compartments

Abstract

Primary non-motile cilia and dendritic spines are cellular compartments that are specialized to sense and transduce environmental cues and presynaptic signals, respectively. Despite their unique cellular roles, both compartments exhibit remarkable parallels in the general principles, as well as molecular mechanisms, by which their protein composition, membrane domain architecture, cellular interactions, and structural and functional plasticity are regulated. We compare and contrast the pathways required for the generation and function of cilia and dendritic spines, and suggest that insights from the study of one may inform investigations into the other of these critically important signaling structures.

INTRODUCTION

Cells contain multiple functionally specialized organelles and subcellular compartments. Among them are primary non-motile cilia (henceforth referred to as cilia) and dendritic spines, structures that protrude into the environment and convey external information to the cell. At first glance, cilia and dendritic spines may appear to be quite different, and certainly have been studied in very different contexts. Cilia are morphologically diverse microtubule-based structures that are present on nearly all cell types in metazoans including neurons, and that sense and transduce a myriad of external cues including chemicals, mechanical pressure, light and morphogens (Berbari et al., 2009; Drummond, 2012; Green and Mykytyn, 2010; Louvi and Grove, 2011) (Fig. 1A). In contrast, as their name implies, dendritic spines are found on the dendrites of only a subset of mammalian neurons. They are primarily actin-based postsynaptic structures also of diverse morphologies, and are responsible for neurotransmission of signals from presynaptic neurons (Calabrese et al., 2006; Lee et al., 2012; Rochefort and Konnerth, 2012) (Fig. 1B). Neurons can contain both multiple spines on their dendritic processes, as well as generally a single primary cilium emanating from their soma. As suggested previously, similar to dendritic spines, cilia can also essentially be thought of as postsynaptic structures specialized to respond to ‘presynaptic’ environmental stimuli (Shaham, 2010). However, in contrast to spines which respond only to signals sent by presynaptic neurons, cilia may integrate information more broadly from multiple sources. Nevertheless, the fundamental role of both organelles is ultimately similar - to sense and transduce extracellular cues.

Fig. 1.

The architecture of the cilium and mature dendritic spine. (A) The core of the cilium is comprised of the microtubule (MT)-based axoneme arising from the basal body. The transition zone (TZ) and transition fibers are indicated. The septin ring at the ciliary base is also shown. Anterograde and retrograde IFT are driven by kinesin/dynein motors. Other key components of the consensus ciliary membrane (e.g. channels and receptors) are also shown schematically. (B) An actin network is at the core of the mature dendritic spine. The spine depicted here displays canonical morphology - a pronounced spine head with the PSD at its tip is connected to the base via a slim neck. Lateral to the PSD is the endocytic zone. The septin cytoskeleton at the spine base is also shown. Major motors, signaling molecules, and cytoskeletal components of the spine are indicated. Parts of this figure are adapted from Saarikangas and Barral (2011).

Here, we compare and contrast the key mechanisms regulating the formation and maintenance of the cilium and dendritic spine. As we describe below, these structures often employ similar strategies and molecular mechanisms to execute their specialized signaling functions. Given the many commonalities in the underlying pathways, we suggest that investigations into mechanisms of ciliogenesis may greatly influence our understanding of the biogenesis and function of dendritic spines and vice versa.

SPECIALIZED MEMBRANE COMPOSITION AND MICRODOMAINS IN CILIA AND DENDRITIC SPINES

Dendritic spines and cilia share a remarkable degree of similarity at the level of their overall membrane composition. Both cellular compartments represent specialized membrane domains that concentrate signaling molecules mediating cellular responses to external stimuli. Cilia detect and transduce intercellular signals by concentrating the components of several developmental pathways including Hedgehog, canonical and non-canonical Wnt, and PDGFα cascades (Christensen et al., 2008; Corbit et al., 2005; Haycraft et al., 2005; Rohatgi et al., 2007; Schneider et al., 2005). Cilia also sense the environment by housing molecules such as olfactory, neuropeptide and monoamine receptors, phototransduction components, and channels in their membrane (Berbari et al., 2008; Domire et al., 2011; Gong et al., 2004; Handel et al., 1999; Kottgen and Walz, 2005; Liu et al., 2007; Mayer et al., 2008; Roepman and Wolfrum, 2007; Tobin et al., 2002) (Fig. 1A). Likewise, dendritic spine membranes are enriched in postsynaptic signaling complexes that include ionotropic and metabotropic glutamate receptors, ion channels, and signaling enzymes among others (Okabe, 2007; Sheng and Kim, 2011; Walsh and Kuruc, 1992) (Fig. 1B). In both structures, signaling molecules are not uniformly or randomly distributed throughout their membranes but are instead spatially organized into distinct microdomains (Newpher and Ehlers, 2009). This spatial organization likely contributes to the well-documented presence of calcium micro-domains within dendritic spines (Chen and Sabatini, 2012; Higley and Sabatini, 2012), as well as in a subset of cilia (Castillo et al., 2010).

The best studied example of a signaling microdomain in dendritic spines is represented by the postsynaptic density (PSD) at their tips (Fig. 1B). The PSD is an electron-dense region that concentrates and localizes postsynaptic signaling molecules such as AMPA and NMDA receptors, although the exact composition depends on the level of neuronal activity and cell type (Inoue and Okabe, 2003; Newpher and Ehlers, 2009; Palay, 1956; Sheng and Kim, 2011; Siekevitz, 1985). Molecules are tethered at the PSD via interaction with PDZ domain-containing scaffolding proteins allowing functional coupling of membrane proteins to intracellular signaling molecules (Cho et al., 1992; Feng and Zhang, 2009; Kim and Sheng, 2004; Kim et al., 1997; Kornau et al., 1995; Naisbitt et al., 1999; Sheng and Kim, 2011). Additional experimental evidence shows further compartmentalization of specific AMPA and NMDA receptor subunits within the synaptic microdomain, with particular subunits localized to the center or periphery of the synapse (Chen et al., 2008; Gladding and Raymond, 2011; Kharazia and Weinberg, 1997; Newpher and Ehlers, 2009; Shinohara et al., 2008). The ciliary membrane of many cell types is also organized into microdomains characterized by the asymmetric localization of signaling proteins (Fig. 1B). For example, a number of transient receptor potential cyclic nucleotide-gated and chloride channels are enriched in different segments of sensory cilia, and T2R bitter taste receptors differentially localize to distinct ciliary segments of human epithelia in a receptor-specific manner (Flannery et al., 2006; French et al., 2010; Lee et al., 2010; Liang et al., 2011; Shah et al., 2009; Wojtyniak et al., 2013). Interestingly, there is evidence that PDZ domain-containing scaffolding proteins similar to those in spines may be present in cilia and modulate signaling and ciliogenesis (Dooley et al., 2009; Sfakianos et al., 2007; Shiraishi et al., 2004). This compartmentalization into signaling microdomains within the spine and cilia may enable cells to exhibit remarkably precise cellular responses to defined external cues (Flannery et al., 2006; Hardingham and Bading, 2010; Lehnert et al., 2013; Leveille et al., 2008; Newpher and Ehlers, 2009).

Aside from their unique protein composition, cellular membranes of cilia and dendritic spines are characterized by lipid content that is distinct from that of the adjacent plasmalemma. A number of studies showed that ciliary and flagellar membranes are rich in cholesterol and sphingolipids - major constituents of highly organized lipid microdomains or ‘lipid rafts’ which have a well-established role in organizing diverse signaling pathways (Janich and Corbeil, 2007; Kaneshiro et al., 1984; Kaya et al., 1984; Simons and Toomre, 2000; Souto-Padron and de Souza, 1983; Tetley, 1986). Recent investigations into the ciliary lipidome of different cell types uncovered high lipid-order microdomains in these cells’ ciliary membranes and at the ciliary base suggesting that lipid rafts may be present in cilia, and may contribute to the assembly and function of cilia and associated signaling pathways, although this requires further investigation (Boesze-Battaglia and Albert, 1990; Brady et al., 2004; Iomini et al., 2006; Mitchell et al., 1990; Nair et al., 2002; Senin et al., 2004; Seno et al., 2001; Travis et al., 2001; Tyler et al., 2009; Vieira et al., 2006; Wang et al., 2009a). Similar to cilia, lipid microdomains have been implicated in a number of cellular functions including regulation of synaptic size and number, neuronal excitability, and receptor trafficking in dendritic spines. For instance, many ionotropic and metabotropic neurotransmitter receptors localize to the lipid raft domains in dendritic spines together with the downstream signaling machinery (reviewed in Allen et al., 2007). In hippocampal neurons, lipid raft disruption induces loss of spines and synapses (both excitatory and inhibitory) and alters surface stability and size of AMPA receptor clusters, thus affecting neuronal excitability (Hering et al., 2003). Moreover, neurotransmitter binding and signaling potency appear to be sensitive to changes in lipid raft composition (Francesconi et al., 2009; Gaudreault et al., 2004). The obvious similarities in the organization of signaling microdomains in both cilia and spines raises the possibility that similar mechanisms and molecules may operate in both organelles to further compartmentalize their specialized membranes.

THE GATEKEEPERS AT THE BASE

The unique biochemical composition and compartmentalized nature of both cilia and dendritic spines require that both compartments segregate specific transmembrane and soluble factors from the rest of the cell. There is a growing body of evidence to support the existence of diffusion barriers that physically and functionally separate cilia and spine membranes from the plasma membrane. Although these barriers exhibit distinct morphological features and different degrees of selectivity in the two structures, they appear to utilize similar principles to ensure functional segregation.

The existence of a physical diffusion barrier at the base of the cilium was initially suggested by electron microscopy, which revealed a “ciliary necklace” of intramembranous particles encircling the ciliary base, and Y-link structures connecting the ciliary membrane at the level of the necklace to microtubules of the axoneme (Gibbons and Grimstone, 1960; Gilula and Satir, 1972) (Fig. 1A). Functional evidence in support of a membrane diffusion barrier came from studies that used fluorescence recovery after photobleaching (FRAP) to show that bleaching a small region within the ciliary or plasma membrane resulted in rapid recovery of fluorescence, whereas little to no recovery of fluorescence was observed when the entire cilium was photobleached (Chih et al., 2012; Hu et al., 2010). This observation implied the existence of a physical barrier blocking free exchange of ciliary and plasma membrane components. Existence of a lipid diffusion barrier separating the ciliary and plasma membranes has also been proposed since GPI-anchored proteins targeted specifically to the apical plasma membrane of MDCK cells fail to diffuse into the ciliary membrane (Vieira et al., 2006). This exclusion phenomenon was correlated with the presence of a highly condensed lipid zone at the ciliary base of these cells (Vieira et al., 2006). In addition to a barrier for transmembrane molecules, the selective movement of soluble proteins into and out of the cilium may be impeded by a diffusion barrier. Recent studies have reported that a size-exclusion barrier may regulate entry of soluble proteins into the cilium (Breslow et al., 2013; Lin et al., 2013); proteins related to those in nuclear pore complexes may be components of this barrier in some cases (Kee et al., 2012; Ounjai et al., 2013). Finally, membrane curvature at the ciliary base may also hinder diffusion of the ciliary transmembrane proteins into plasma membrane and vice versa.

The existence of a diffusion barrier has been also proposed for dendritic spines. The spine morphology and dimensions of the spine neck at least in part determine the degree of molecular and functional compartmentalization of dendritic spines. The spine neck is believed to impose physical constraints restricting the movement of signaling and structural molecules between the spine head and the parent dendrite. The evidence for the spine neck acting as a diffusion barrier comes from numerous imaging studies that demonstrated restricted diffusion of calcium, AMPARs, and other molecules between the spine head and the dendrite (Ashby et al., 2006; Bloodgood and Sabatini, 2005; Muller and Connor, 1992; Noguchi et al., 2005; Sabatini et al., 2002; Svoboda et al., 1996). Furthermore, analysis of bidirectional mobility of photoactivatable GFP across the spine neck showed that the neck limits chemical diffusion in both directions and constitutes a diffusion barrier that is regulated in an activity-dependent manner (Bloodgood and Sabatini, 2005). Certain signaling proteins also translocate between the cilium and the cytosol in response to signaling cues, thus it is likely that permeability of the ciliary diffusion barrier is also regulated by extrinsic stimuli (Domire et al., 2011; Wang et al., 2009d; Zeng et al., 2010).

Studies have recently implicated the septin cytoskeleton in formation of the diffusion barrier at the ciliary base (Hu et al., 2010; Kim et al., 2010b; Saarikangas and Barral, 2011) (Fig. 1A). Components of the septin complex also localize to the base of the dendritic spines and play important roles in regulating dendritic spine morphology (Cho et al., 2011; Saarikangas and Barral, 2011; Tada et al., 2007; Xie et al., 2007) (Fig. 1B). Septins are a highly evolutionarily conserved family of GTPases capable of assembling into higher-order structures (e.g. filaments, bundles, and rings) that play important roles in regulating membrane compartmentalization in diverse contexts (reviewed in Saarikangas and Barral, 2011 and Weirich et al., 2008). Although the septin complex has been shown to delimit diffusion of transmembrane proteins into and out of the cilium via FRAP analyses (Chih et al., 2012; Hu et al., 2010), a similar direct role for septins in dendritic spines has not been described. It is important to note, however, that in addition to septins, a large number of proteins that localize to the ciliary base and may constitute the diffusion barrier have been recently characterized (Czarnecki and Shah, 2012; Garcia-Gonzalo and Reiter, 2012; Reiter et al., 2012). Similar protein complexes at the spine base have not been reported.

Taken together, these findings highlight a number of similarities in functional organization of a diffusion barrier at the base of the cilia and dendritic spines. Transport of signaling and structural molecules into these cellular compartments appears to be regulated by the membrane architecture at the base of cilia and dendritic spines as well as a septin cytoskeleton. In both contexts, diffusion barrier permeability/selectivity is modulated in response to activity and signaling, although likely to different extents. However, there are also marked differences. The transition zone as well as the ultrastructure at the base of cilia are highly complex and organized; similar structures and proteins have not been observed in spines. Moreover, certain stimulation protocols induce translocation of entire organelles including recycling endosomes, polyribosomes, and occasionally mitochondria into the spines of postsynaptic neurons (Bourne et al., 2007; Li et al., 2004; Ostroff et al., 2002; Park et al., 2006). No organelles have ever been detected inside cilia possibly reflecting the more selective nature of the barrier, or perhaps the more limited set of conditions under which cilia have been examined. Nevertheless, it is likely that the basic mechanisms, as well as a subset of components, of the physical barriers that regulate compartmentalization are conserved between cilia and spines.

TRAFFICKING PROTEINS TO AND FROM CILIA AND DENDRITIC SPINES

Cilia and spine-mediated signaling is strictly dependent on their protein content. Thus, maintenance of the adequate morphology and composition of these specialized domains requires continuous bidirectional protein and membrane trafficking. In this section, we focus on emerging similarities, as well as highlight some key differences in membrane and protein trafficking mechanisms utilized by cilia and dendritic spines.

As mentioned above, cilia lack biosynthetic machinery; therefore, all their molecular components need to be trafficked from the soma. The leading model for delivery of transmembrane cargo into the cilium implicates polarized exocytosis of Golgi-derived vesicles at the ciliary base (Hsiao et al., 2012; Nachury et al., 2010; Pazour and Bloodgood, 2008) (Fig. 2A). In contrast, under some conditions, biosynthetic organelles including polyribosomes are present within spines under some conditions allowing for local protein and lipid synthesis (Bourne et al., 2007; Cooney et al., 2002; Ostroff et al., 2002; Spacek and Harris, 1997) (Fig. 2B). Therefore, in addition to the canonical secretory pathway mediating delivery of postsynaptic components via motor-driven transport along dendrites, noncanonical trafficking mechanisms may exist in spines (Gray and Guillery, 1963; Kennedy and Ehlers, 2011; Spacek and Harris, 1997). However, similar to the situation in cilia, exocytic events occur largely at the spine base in the dendritic shaft, although exocytosis within the spine head has also been reported (Kennedy et al., 2010; Lin et al., 2009; Makino and Malinow, 2009; Passafaro et al., 2001; Patterson et al., 2010; Tao-Cheng et al., 2011; Yudowski et al., 2007) (Fig. 2B). Intriguingly, intraciliary vesicles have been observed in several systems suggesting exo-/endocytosis may take place within the cilium as well; however, both the presence of these vesicles and their roles require confirmation (Poole et al., 1985; Reese, 1965).

Fig. 2.

Polarized vesicle trafficking and lateral diffusion in the cilium and dendritic spine. (A) Proteins are synthesized in the Golgi and delivered to the base of the cilium by polarized vesicle exocytosis. Following delivery to the cilium, the protein cargo of the vesicles is transported along the ciliary axoneme by IFT and/or membrane diffusion. Some Golgi-derived vesicles may instead be targeted to the plasma membrane, and their protein cargo can subsequently enter the cilium via lateral diffusion. Recycling endosomes at the base of the cilium may constitute another source of the cilium-bound vesicles. (B) Golgi-synthesized transmembrane components of the dendritic spine are delivered to the membrane via exocytosis either within the spine head or at the dendritic membrane. In the latter case, proteins subsequently enter the spine head by lateral diffusion. Recycling endosomes (RE) at the base or within the spine promote local recycling of transmembrane proteins. Polyribosomes are found at the spine base and may translocate into the spine head in response to activity.

The best understood regulators of vesicular targeting to the ciliary base include the GTPase Rab8, its guanine nucleotide exchange factor (GEF) Rabin8, and the GTPase Rab11, which regulates Rabin8 activation (Chiba et al., 2013; Deretic et al., 1995; Knodler et al., 2010; Moritz et al., 2001; Westlake et al., 2007; 2011) (Fig. 2A). Similarly, Rab8 is required for AMPAR trafficking to the PSD; although, in this context, it mediates postsynaptic AMPAR delivery from an intracellular membrane compartment within the spine (Gerges et al., 2004) (Fig. 2B). The exocyst - an octameric protein complex that was initially implicated in polarized exocytosis in budding yeast S. cerevisiae - localizes to the cilia in mammalian cells and may control tethering of cilium-bound vesicles (Das and Guo, 2011; Mazelova et al., 2009; Rogers et al., 2004; Zuo et al., 2009) (Fig. 2A). Similarly, the exocyst components have been implicated in coordinating AMPAR targeting toward and fusion with the postsynaptic membrane in the dendritic spines and in regulating spine density and synaptic plasticity (Gerges et al., 2006; Teodoro et al., 2013) (Fig. 2B).

The sorting step of protein transport into the cilium is facilitated by ciliary targeting sequences (CTSs) (Berbari et al., 2008; Geng et al., 2006; Jenkins et al., 2006; Pazour and Bloodgood, 2008). In some cases, these sequences mediate protein delivery to the ciliary base by directly binding small GTPases (e.g. Rab8 and Arf4) or the BBSome - a coat complex of Bardel-Biedl Syndrome proteins (Deretic et al., 2005; Follit et al., 2010; Jin et al., 2010). In addition to the CTSs, certain post-translational modifications (PTMs) (e.g. palmitoylation and myristoylation) appear to be important for targeting ciliary components (e.g. rhodopsin, cystin, and fibrocystin) to specific microdomains of the cilium (Follit et al., 2010; Tam et al., 2000; Tao et al., 2009). The same PTMs have been implicated in targeting the PDZ domain-containing scaffolding components and signaling enzymes to the PSD membrane, suggesting similar mechanisms may regulate protein targeting to the dendritic-spine and ciliary lipid microdomains (Craven et al., 1999; Keith et al., 2012; Konno et al., 2002; Zhu et al., 2013). A number of sequence motifs required for protein targeting to the dendritic membrane have been identified; however, no consensus dendritic spine-targeting sequence appears to exist (Craven et al., 1999; Francesconi and Duvoisin, 2002; Kameda et al., 2008; Konno et al., 2002; Mitsui et al., 2005; Noel et al., 1999; Rivera et al., 2003; West et al., 1997). How these motifs mediate dendritic or postsynaptic protein targeting also remains largely obscure.

In addition to vesicular targeting of ciliary and synaptic components, membrane transport via lateral diffusion contributes to the delivery of certain transmembrane ciliary and dendritic-spine molecules. Studies in Chlamydomonas showed that the plasma membrane constitutes the source for replenishing the flagellar agglutinin pool following gamete fusion during early stages of fertilization (Hunnicutt et al., 1990). Likewise, under certain conditions, mammalian Smoothened was proposed to translocate laterally from the plasma into the ciliary membrane via lateral diffusion (Milenkovic et al., 2009). It should be noted, however, that the contribution of endo-/exocytosis to these trafficking events has not been completely ruled out. Recent work has also demonstrated that receptor proteins move both by directed transport as well as via lateral diffusion within the ciliary membrane (Breslow et al., 2013; Ye et al., 2013) (Fig. 2A). Many proteins with essential roles in synaptic development and function also undergo lateral diffusion within the membrane and this movement is a critical mechanism underlying synaptic plasticity (reviewed in Choquet and Triller, 2003). The best characterized example of this process involves movement of AMPAR subunits between the perisynaptic, extrasynaptic, and synapse-associated membrane domains (Adesnik et al., 2005; Ashby et al., 2006; Heine et al., 2008; Makino and Malinow, 2009; Passafaro et al., 2001; Tardin et al., 2003; Yudowski et al., 2007) (Fig. 2B). In this context, lateral diffusion is believed to mediate constitutive recycling of spine-localized AMPARs as well as contribute to activity-dependent regulation of synaptic transmission.

To maintain appropriate ciliary and spine membrane volume and composition, membrane delivery via exocytosis is counterbalanced by endocytosis-mediated membrane retrieval. The importance of postsynaptic endocytosis for synaptic transmission and plasticity is well documented (Anggono and Huganir, 2012; Kessels and Malinow, 2009). Endosome network components are found within or in close proximity to the dendritic spines, where they mediate trafficking, local recycling, or degradation of synaptic proteins (e.g. AMPARs) (Blanpied et al., 2002; Cooney et al., 2002; Park et al., 2004; Petrini et al., 2009; Racz et al., 2004) (Fig. 2B). Several recent studies have directly implicated endocytic proteins in the control of ciliary morphology and membrane composition, highlighting the significance of endocytosis for cilia formation and function (Hu et al., 2007; Kaplan et al., 2012; Kim et al., 2010a). A specialized endocytic membrane domain (the ciliary pocket) characterized by the presence of clathrin-coated pits is also present at the ciliary base, where it is likewise thought to mediate local internalization of ciliary components (Ghossoub et al., 2011; Molla-Herman et al., 2010; Rattner et al., 2010) (Fig. 2A).

Finally, it has been proposed that both dendritic spines and cilia may not only communicate external information to the cell, but also signal to other cells. It has been shown that Chlamydomonas flagella secrete ectosomes that carry proteolytic enzymes necessary for release of the daughter cells from the mother cell following mitosis (Wood et al., 2013). There is also evidence for cilia-derived exosome-like vesicles in neural tube fluid and urine; in the latter case, these vesicles specifically interact with cilia in vitro (Dubreuil et al., 2007; Hogan et al., 2009). Similarly, spinules (evaginations of the postsynaptic membrane into the presynaptic cell) and the presence of multi-vesicular bodies (exosome producing organelles) in dendritic spines have suggested the possibility that dendritic spines may also signal via exosomes (Cooney et al., 2002; Spacek and Harris, 2004; Tao-Cheng et al., 2009; Tarrant and Routtenberg, 1977). The hypothesis that both spines and cilia are secretory organelles is intriguing and constitutes a rich subject for future investigations into possibly analogous signaling mechanisms employed by these structures.

DYNAMIC REGULATION OF SHAPE AND STRUCTURE

Both cilia and dendritic spines are morphologically diverse. Spines have been morphologically classified into stubby, thin or mushroom based on the length and volume of the spine neck and head, respectively (Bourne and Harris, 2008). Ciliary shapes range from the relatively simple rod-like structures present on most cell types in mammals to the highly elaborate photoreceptor outer segments (Silverman and Leroux, 2009). Notably, dendritic spines, and as is becoming increasingly evident, cilia, are highly dynamic structures whose shapes are continuously modified by external and internal inputs. This structural remodeling is linked to plasticity in cellular responses. Although the cytoskeletal makeup of the spine and cilium is different, there are intriguing similarities in the basic principles suggesting that similar mechanisms may mediate activity-dependent morphological plasticity in both compartments.

Both cilia and spines can rapidly disassemble and assemble under certain conditions. In addition to shape, neuronal activity alters spine density by altering spine number. Thus, stimulus paradigms eliciting long-term depression or potentiation can lead to retraction or generation of new spines, respectively (Engert and Bonhoeffer, 1999; Maletic-Savatic et al., 1999; Matsuzaki et al., 2004; Nagerl et al., 2004; Zhou et al., 2004). Although the presence or absence of cilia does not appear to be directly regulated by external inputs, cilia assembly and resorption are tightly and reciprocally coordinated with cell cycle progression; cilia are generally present in G0/G1 cells, resorbed prior to mitosis, and reassemble post-cytokinesis (Avasthi and Marshall, 2012; Basten and Giles, 2013; Plotnikova et al., 2009). However, unlike spines, cilia present on differentiated cell types such as neurons do not appear to undergo dynamic resorption and re-assembly under non-pathological conditions.

Changes in shape of both cilia and spines are mediated by changes in the underlying cytoskeletal architecture. The core of the cilium is comprised of a highly organized and dynamic microtubule structure termed the axoneme, which undergoes continuous tubulin turnover at its distal tip (Fig. 1A). The dynamic behavior of the axonemal microtubules underlies ciliary homeostasis and is regulated by the same molecular mechanisms as those of other microtubule networks including different classes of microtubule associated proteins (e.g. motors, severing and plus-end binding proteins) and tubulin PTMs (Gaertig and Wloga, 2008; Janke and Bulinski, 2011; Verhey et al., 2011). Cilia (and the axoneme) are assembled and maintained by intraflagellar transport (IFT) - the bi-directional, motor-driven movement of macromolecular complexes along the axoneme -that, among other proteins, delivers tubulin to the ciliary tip (Rosenbaum and Witman, 2002; Scholey, 2008) (Fig. 1A).

Dendritic spines on the other hand are largely actin-rich structures (Dent et al., 2011; Shirao and Gonzalez-Billault, 2013). Actin organization within spines is not uniform: the mature spine head contains mostly branched actin filaments in a network, whereas a mixture of linear and branched filaments is found in the spine neck (Korobova and Svitkina, 2010) (Fig. 1B). Like the axoneme in cilia, the actin cytoskeleton in spines is also dynamic. Analysis of actin dynamics in the spine head uncovered several actin pools with distinct profiles - the most dynamic population is found at the tip of the spine head perhaps analogous to the cilia tip, while a stable pool resides at the lateral endocytic domain and the base (Frost et al., 2010; Honkura et al., 2008). As in the case of cilia, molecular motors such as myosins play a role in the delivery of important molecules to the PSD and regulation of spine morphology (Rubio et al., 2011; Ryu et al., 2006; Yoshii et al., 2013) (Fig. 1B).

In spines, activity induces rapid morphological changes that correlate with synaptic strength (Matsuzaki et al., 2004; Nagerl et al., 2004; Okamoto et al., 2004; Zhou et al., 2004). Activity-dependent spine remodeling has been studied extensively, and we refer the reader to several excellent reviews on this topic for details (Dent et al., 2011; Pontrello and Ethell, 2009; Shirao and Gonzalez-Billault, 2013). In brief, consistent with their cytoskeletal composition, spine morphogenesis and remodeling are largely regulated by signaling pathways converging on the actin cytoskeleton (although also see below) (Fischer et al., 1998; Hotulainen and Hoogenraad, 2010; Star et al., 2002; Zito et al., 2004). Numerous studies showed that pharmacological perturbation of actin polymerization or stability interferes with activity-dependent structural remodeling of spines, as does the genetic disruption of diverse actin-interacting proteins (Allison et al., 1998; Ethell and Pasquale, 2005; Fischer et al., 1998; Halpain et al., 1998; Hotulainen and Hoogenraad, 2010; Matsuzaki et al., 2004; Schubert and Dotti, 2007). In addition to modulating spine head morphology, the actin cytoskeleton plays an important role in regulating the organization of the PSD and in trafficking postsynaptic neurotransmitter receptors (Kim and Lisman, 1999; Krucker et al., 2000; Kuriu et al., 2006), important contributors to activity-dependent neuronal plasticity.

Similarly, ciliary morphology is also dynamically controlled by diverse extracellular stimuli. Studies in Chlamydomonas uncovered several important mechanisms regulating flagella dynamics. In Chlamydomonas, flagella length can be modulated by a number of pharmacological agents including lithium, cytoskeletal drugs, diverse anesthetics, environmental changes in pH and osmolarity, mating, as well as intracellular cAMP and calcium levels (Lefebvre et al., 1978; Mesland et al., 1980; Pasquale and Goodenough, 1987; Quader et al., 1978; Wilson et al., 2008). cAMP and calcium, as well as lithium, were also implicated in cilia length control in mammalian cells suggesting that some regulatory mechanisms of ciliary plasticity may be conserved (Besschetnova et al., 2010; Low et al., 1998; Miyoshi et al., 2009; Ou et al., 2009). In C. elegans, communication via small molecule pheromones has been suggested to modulate cilia dynamics, in part via modulation of IFT motors (Burghoorn et al., 2010). More recently, cilia were implicated in guiding migration of neurons and radial progenitors during cortical lamination by undergoing continuous changes in length, branching complexity, and orientation (Higginbotham et al., 2012; 2013). In this context, ciliary dynamic behavior was at least partly regulated by the small GTPase Arl13b (Higginbotham et al., 2012; 2013).

There is now some evidence that microtubules and the actin network also control spine and ciliary morphology, respectively. Stable actin filaments are associated with the base of the ciliary endocytic domain (the ciliary pocket), while the dynamic actin pool is present in the distal pocket regions (Molla-Herman et al., 2010). Interestingly, changes in actin dynamics at the ciliary pocket correlate with altered axoneme dynamics (e.g. bending), suggesting that actin may contribute to ciliary plasticity (Molla-Herman et al., 2010). However, while actin polymerization and branching are generally associated with spine enlargement, the relationship between actin dynamics and cilia length is more complex and context-specific (Fukazawa et al., 2003; Hotulainen et al., 2009; Kasai et al., 2010; Okamoto et al., 2004; Wegner et al., 2008). For example, pharmacological disruption of F-actin or mutations in the Arp2/3 complex promotes cilium elongation in mammalian cells; in contrast, in zebrafish Kupffer’s vesicle and quail oviduct, impaired actin assembly results in short cilia (Bershteyn et al., 2010; Boisvieux-Ulrich et al., 1990; Hotulainen et al., 2009; Kim et al., 2010a; Oishi et al., 2006; Sharma et al., 2011; Wegner et al., 2008). Microtubules have been observed within mushroom spines and implicated in activity-dependent spine remodeling (Gu et al., 2008; Hu et al., 2008; Jaworski et al., 2009; Kapitein et al., 2011). Microtubules within the spine are highly dynamic and likely contribute to spine morphogenesis by modulating actin assembly and turnover (Gu et al., 2008; Hu et al., 2008; Jaworski et al., 2009). It will be interesting to determine whether kinesin-mediated transport mechanisms such as those in cilia also operate in spines to deliver scaffolding and signaling molecules (Hu et al., 2011).

Although actin-dependent mechanisms are the major regulators of spine remodeling, the balance between membrane delivery and retrieval via endo-/exocytosis is also an important contributor to this process. Indeed, LTP-inducing stimuli promote AMPAR exocytosis from the recycling endosomes concomitantly with spine enlargement, and impaired endosomal recycling abolishes LTP-dependent spine expansion (Park et al., 2004; 2006; Petrini et al., 2009). Membrane trafficking may also play a role in remodeling ciliary morphology. Work from our lab has demonstrated that a subset of cilia in C. elegans sensory neurons undergo dramatic remodeling in response to altered sensory input; these morphological changes are at least partly mediated by the exocytic pathway (Mukhopadhyay et al., 2008). Outer segment length in photoreceptors is also dynamic and modulated by light exposure (Abramoff et al., 2013; Penn and Williams, 1986), further highlighting the importance of balanced membrane delivery for ciliary plasticity.

In summary, cilia and spines are highly dynamic cytoskeleton-based compartments with diverse morphologies that are highly tuned to changes in their extracellular environment. However, while much is now known about the mechanisms by which activity impinges on the actin and microtubule network in spines to modulate morphology, far less is understood regarding effect of environmental signals on cilia structural remodeling. We suggest that lessons learned from spines may perhaps be applied to cilia to further investigate the pathways contributing to their structural plasticity.

A LITTLE HELP FROM SUPPORT CELLS

The role of glial support cells in modulating synaptic function, both in the central and peripheral nervous systems, is now well described (Clarke and Barres, 2013; Fiacco and McCarthy, 2006; Halassa and Haydon, 2010; Volterra and Meldolesi, 2005). Indeed, the close association among pre- and post-synaptic sites and the surrounding glia has led to the concept of the ‘tripartite synapse’ (Araque et al., 1999; Newman, 2003b; Pannasch and Rouach, 2013) (Fig. 3A). Relevant to our discussion, in the central nervous system, astrocytes are in close contact with dendritic spines (Lehre and Rusakov, 2002; Ventura and Harris, 1999), and are critical modulators of excitatory synaptic transmission, synaptogenesis and synaptic plasticity.

Fig. 3.

Glial interactions with cilia and dendritic spine. (A) The “tripartite” synapse consisting of presynaptic and postsynaptic cells and glia. (B) A schematic representation of the mammalian photo-receptor showing close apposition of its outer segment with an RPE cell. CC, connecting cilium; IS, inner segment. (C) In the C. elegans amphid sensory organ, the glial processes form a channel which houses ciliated dendritic endings of a subset of sensory neurons. Cilia of another subset of sensory neurons are embedded in the amphid sheath glia. Redrawn from Perkins et al. (1986).

Astrocytes respond to presynaptic stimulation by changes in calcium transients, and release neurotransmitters to modulate synaptic activity (Araque et al., 1998a; 1998b; Cornell-Bell et al., 1990; Newman, 2003a; Panatier et al., 2011; Parpura et al., 1994; Pasti et al., 1997; Perea and Araque, 2005; Porter and McCarthy, 1996; Wang et al., 2006) (Fig. 3A). Moreover, astrocytes are intimately involved in remodeling spine morphology in response to activity. Similar to spines, astrocytic processes are also dynamic and their structural changes are correlated with changes in formation and maintenance of spine morphology (Haber et al., 2006; Hirrlinger et al., 2004; Nishida and Okabe, 2007; Verbich et al., 2012). Astrocyte-derived soluble as well as membrane-associated factors including ephrin/EPH signaling components have been implicated in direct glial-mediated spine remodeling (Beattie et al., 2002; Carmona et al., 2009; Hama et al., 2004; Murai et al., 2003). However, it is now evident that glia also indirectly influence synaptogenesis via production of components of the extracellular space, which in turn alter synaptic function presynaptically and/or postsynaptically (Allen et al., 2012; Bhatt et al., 2009; Christopherson et al., 2005; Dansie and Ethell, 2011; Eroglu et al., 2009; Kucukdereli et al., 2011). It is also being increasingly recognized that excitatory synapses are modulated by interactions with microglia, intrinsic immune cells of the central nervous system (Kettenmann et al., 2013; Miyamoto et al., 2013; Wake et al., 2013) (Fig. 3A). Thus, spine dynamics and functions are critically dependent on interactions with physically apposed cells and molecules.

In contrast to the well-studied roles of glia in regulating synaptic and spine properties, the contribution of glia or other support cells to the structure and function of cilia is not well understood. However, there are several clear examples of the critical importance of these support cells in biogenesis and function of both motile and immotile cilia. In the mouse brain, activation of the proinflammatory response in astrocytes results in loss of ependymal motile cilia, which in turn alters cerebrospinal fluid transport leading to hydrocephalus (Lattke et al., 2012). It is also becoming evident that disrupting the appropriate interaction between cilia and the extracellular matrix in many tissue types alters ciliary signaling resulting in profound cellular and systemic disorders (reviewed recently in Seeger-Nukpezah and Golemis, 2012). These observations indicate that similar to the importance of neuron-glial interactions in regulating spine structure and function, signaling between cilia and surrounding cells and molecules also contributes to maintaining cilia function.

Support cells play a particularly important role in regulating the functions and morphology of the sensory cilia of photoreceptors and chemosensory neurons. Phototransduction molecules of rod and cone photoreceptors are housed in ciliary specializations called outer segments which are in close physical contact with a monolayer of retinal pigmented epithelial (RPE) cells (Fig. 3B). Photoreceptor outer segments are constantly renewed; this process is mediated by shedding of the tips of the photoreceptor outer segments and concomitant transport of new ciliary proteins to generate new disks (Young, 1967; 1971; 1976; Young and Droz, 1968). Shed disks are phagocytosed by RPE cells, which also recycle the retinal cofactor and other molecules for reuse by photoreceptors (Bibb and Young, 1974; Chen et al., 1992; Gordon et al., 1992; Strauss, 2005; Young and Bok, 1969). The critical importance of RPE cells in maintaining outer segment structure and function is underscored by the fact that defects in RPE-mediated phagocytosis or recycling result in retinal degeneration and cell death (Bok and Hall, 1971; Edwards and Szamier, 1977; Gibbs et al., 2003; Mullen and LaVail, 1976). In this context, RPE-mediated phagocytosis is reminiscent of the synaptic ‘stripping’ and phagocytic functions of microglia in the mammalian brain (Paolicelli et al., 2011; Schafer et al., 2012; Trapp et al., 2007; Tremblay et al., 2010). In addition to RPE cells, Müller glial cells are also important for biogenesis of photoreceptor outer segments and recycling of cone retinal cofactor (Wang et al., 2009b). Disruption of Müller cell function alters outer segment assembly, further supporting a role for these glial cells in photoreceptor cilia formation and function (Jablonski and Iannaccone, 2000; Wang et al., 2004; 2009c).

The importance of glia in maintaining sensory cilia has been particularly well-described in C. elegans. Cilia are present at the dendritic ends of a subset of sensory neurons in the head sensory organs of C. elegans, and contain all molecules required for sensory signal transduction (Perkins et al., 1986; Ward et al., 1975). The majority of these cilia are embedded or ensheathed in glial processes (Ward et al., 1975) (Fig. 3C), similar to the ensheathing of spines by astrocytes. Physical and genetic ablations, as well as other genetic manipulations have shown that these glia are essential for cilia structural maintenance and function (Bacaj et al., 2008; Perens and Shaham, 2005; Procko et al., 2011; 2012). More tellingly, analogous to the roles of glia in spine remodeling, glia are also required for experience or developmental stage-dependent ciliary morphological plasticity (Procko et al., 2011). Although the mechanisms by which these glia regulate ciliary structure and function have not yet been fully described, a subset of the required glial molecules may be conserved. For instance, fig-1 encodes a molecule with domains similar to those present in vertebrate thrombospondin, a glia-secreted component of the extracellular space that regulates synaptogenesis (Bacaj et al., 2008; Christopherson et al., 2005; Eroglu et al., 2009). FIG-1 similarly acts in glia to regulate sensory neuron function in C. elegans (Bacaj et al., 2008), although it is unclear whether these functions are mediated via cilia. These observations indicate that the structure and function of both sensory cilia and dendritic spines are modulated via interactions with physically associated support cells, as well as components of the extracellular space.

CONCLUSION

Spines and cilia share evident similarities in many of the pathways and mechanisms that allow these organelles to sense and convey external information. Despite these similarities, there are obvious differences. Among these are the highly organized nature of the diffusion barrier in cilia as compared to spines, the presence of vesicular trafficking and localized translation machinery within spines but not cilia, and the critical contribution of IFT to cilia biogenesis and maintenance. In some cases, these differences may simply reflect the different contexts in which these organelles have been largely studied. For instance, as described above, the picture of relatively static cilia as compared to the dynamic spine is being revised as cilia are examined under changing environmental conditions. Similarly, there is some evidence that activity-dependent localized translation may not just be a feature of spines but may also occur near or in cilia and contribute to cellular functional plasticity (Kaye et al., 2009). However, other differences may reflect the fact that spines are dedicated biochemical as well as electrical compartments in neurons, whereas cilia play broader signaling roles in multiple contexts in multiple cell types.

Indeed, the similarities in the major underlying mechanisms are sufficiently striking that techniques and concepts developed through the study of one organelle may perhaps be readily adopted for the study of the other (for instance, see Lin et al., 2013 for an excellent review on tools to study diffusion barriers). As an example, the use of single particle tracking which has been used so effectively in spines (Frost et al., 2010; Tardin et al., 2003) to follow the kinetics of single molecules has been applied to a relatively limited extent in cilia (Imanishi et al., 2006; Ye et al., 2013). The use of related technologies in cilia would provide a great deal of new information about trafficking and turnover of ciliary proteins. Information about the structure and function of the transition zone in cilia could inform similar studies in spines and help further characterize the role of the spine neck as a membrane diffusion barrier. Studies in spines could also greatly inform our understanding of the mechanisms by which cilia are dynamically remodeled by signaling. Ciliary or spine dysfunction leads to a plethora of disorders ranging from developmental anomalies to behavioral syndromes (Koleske, 2013; Lee and Gleeson, 2010; Penzes et al., 2011; Waters and Beales, 2011). We propose that a coming together of findings from these seemingly disparate fields will allow for a broader and deeper understanding of how these important signaling structures form and function.