Regulation of the actin cytoskeleton in dendritic spines

Introduction

More than a century ago, after observing dendritic spines on Purkinje cell dendrites, Santiago Ramon y Cajal proposed that “such spines could be the points where electrical charge or current is received”. This hypothesis has proved to be correct, and it has been established that most excitatory synapses are formed between axon terminals and these ‘dendritic spines’.

These dendritic spines are heterogeneous in their shape and size. However, their morphology can be segregated into long-thin, stubby, and mushroom shaped. Their shape often reflects their stability and the strength of the synapse, the latter presumably due to AMPA receptor levels. Often, the destabilization of spines – their shrinkage from mushroom-shaped to long and thin leads to their disappearance [1]. Conversely, spines often appear as long thin filipodia and increase in area in an activity dependent manner [2]. The change in spine morphology, spine morphogenesis, is mainly dependent on the remodeling of β- and γ- actin, the main isoforms of actin present in neurons [3]. In this chapter we will discuss molecules involved in dendritic spine plasticity beginning with actin and moving upstream towards neuromodulators and trophic factors that initiate signaling involved in these plasticity events. We will place special emphasis on small GTPase pathways, as they have an established importance in dendritic spine plasticity and pathology.

Dynamic remodeling of the actin cytoskeleton is thought to be the driving force behind structural alterations of spines. As in other cell types, in neurons, actin is found as soluble monomeric G-actin and polymerized F-actin filaments, the latter likely conferring the characteristic spine morphology. The polymerization of free G-actin is subject to regulation by numerous pathways activated by various surface receptors [4]. Most notably, activations of N-methyl-D-aspartic acid (NMDA) receptors, lead to the aforementioned changes. The pathways that act as transducers of these changes are subject to modulation by converging pathways giving rise to a complex molecular network.

Actin Binding Proteins

While actin remodeling drives spine morphogenesis, this is in its turn regulated by a complex network of actin regulatory proteins. Closest to actin are the actin binding proteins (ABPs). The conversion of soluble G-actin into F-actin is a highly dynamic and reversible process that is regulated through interactions with ABPs. The differential effect of ABPs on actin (some favor polymerization while others depolymerization), confers intricate regulation of the cytoskeletal remodeling at the synapse. The actin-related proteins 2 and 3 (Arp2/3) complex is a major component of actin remodeling that is localized to dendritic spines of hippocampal neurons [5]. Upon activation, Arp2/3 binds existing acting filaments, nucleating them into a branched network of actin filaments [6]. Recent knockdown studies of Arp2/3 in hippocampal neurons have revealed its importance for dendritic spine formation [7]. An interesting consideration is that the Arp2/3 complex is the target of many converging pathways involved in dendritic spine morphogenesis. For example, the F-actin binding protein cortactin binds Arp2/3, activating and localizing it to dendritic spines [8, 9]. Another crucial Arp2/3 activator is WAVE-1 (Wiskott-Aldrich syndrome protein family member 1). WAVE-1 serves as a signal transducer between the Rho GTPase Rac1 and Arp2/3. Knockdown studies of WAVE-1 have revealed its importance in spine morphology. Depletion of other Arp2/3 activators including Abi2, N-WASP, and Abp1 alter the morphology and number of spines [7, 10, 11].

Several other ABPs regulate actin dynamics in spines and spien morphology, including profiling, drebrin, gelsolin, spinophilin and cofilin. Profilin, another key player in actin polymerization , targets to dendritic spines upon chemical or electrical stimulation of hippocampal neurons [12, 13]. Experiments utilizing a peptide competitor of profilin prevented profiling targeting and destabilized dendritic spines [14]. Concordantly, it has been observed that profilin translocates from the dendritic shaft into the dendritic spine in the amygdala after fear conditioning [15] Another important promoter of actin polymerization is drebrin. Drebrin is an F-actin binding protein that is highly concentrated in dendritic spines, where it associates with actin filaments [16, 17]. Studies have shown that drebrin accumulates in dendritic spines prior to PSD-95 during spine formation. Knock down of drebrin with siRNA disrupts accumulation of PSD-95 in spines. These studies suggest that drebrin’s role is to promote actin assembly and the clustering of PSD-95 in synaptic spines [18].

Gelsolin is another actin binding proteins whose actin binding activity is Ca²⁺ dependent. In the presence of high Ca² concentration, gelsolin binds to the ends of actin filaments and prevents further elongation. This action also serves to stabilize the actin filaments during synaptic plasticity [19].

Spinophilin, named after its prominent localization to dendritic spines, targets protein phosphatase 1 (PP1) to dendritic spines and stimulates its phosphatase activity [20, 21]. Spinophilin’s actin binding is modulated by protein kinase A (PKA) and Ca²⁺/calmodulin-dependent kinase II (CamKII), allowing for its activity dependent regulation [22]. Additionally, spinophilin has been shown to serve as a Rac1 regulator through its interaction with the Rac1 guanine exchange factor (GEF) Tiam1 [23].

The balance between G- and F-actin is also controlled by the actin depolymerizing (ADF) factor related protein cofilin. Depending on phosphorylation state, cofilin can either disassemble filaments or sever them providing a barbed end of actin assembly [24]. Knocking down cofilin in neuronal culture results in a reduction of F-actin turnover and a loss of dendritic spine density [25]. Phosphorylation of cofilin on serine 3 by LIM kinases inhibits its function. LIM kinase I (LIMK-1) is a ADF/cofilin specific kinase enriched in dendritic spines. Hippocampal neurons cultured from LIMK-1 mice show reduced cofilin phosphorylation and aberrant F-actin accumulation in spines [26]. Although the roles of the above molecules have predominantly been identified as ABPs, many of these have other roles within the synaptic plasticity network.

Small GTPases

Major regulators of actin cytoskeletal remodeling in many cell types, as well as in neurons, are the small G-proteins (or small GTPases) related to Ras. Since GTPases can exist in two states, an active GTP-bound and an inactive GDP-bound, they serve as ‘ON’ or ‘OFF’ binary molecular switches. These molecules are regulated by guanine exchange factors (GEFs) which catalyze the exchange of GDP for GTP, resulting in activation of the GTPase. Conversely, GTPase activating proteins (GAPs) enhance the hydrolysis of GTP into GDP, inactivating the GTPase. There are a few different families of GTPases involved in spine morphogenesis. The most widely studied is the Rho family-which includes RhoA, Rac1, Cdc42, and others which are not as well characterized. Tight regulation of these molecules is necessary for proper spine formation and function. It is generally accepted that Rac1 activation stimulates F-actin polymerization and stabilizes dendritic spines through the activation of downstream effectors p21-activated kinase (PAK), LIM kinase I (LIMKI), and the actin binding protein cofilin [27-29]. Luo and colleagues showed that transgenic mice expressing a mutant form of Rac1, lacking the ability to hydrolyze GTP thus remaining constitutively active, resulted in increased spine density at the expense of spine size in cerebellar Purkinje cells [30]. Experiments where constitutively active Rac1 was overexpressed in hippocampal cultured neurons and slices documented the formation of irregularly shaped protrusions and impairment of synapse formation which contrasts with in vivo data [30-32]. Overexpression of a dominant negative form of Rac1, incapable of interacting with GEFs, drastically decreases the number of spines and synapses in cultured hippocampal slices and dissociated neurons [32-34]. These data suggest that an optimal level of Rac1 activation is required for proper maintenance of dendritic spines, and only small fluctuations in Rac1 activity are responsible for spine morphogenesis.

Although less well studied in spines that Rac1, another small GTPase, Cdc42, plays a role in spine morphogenesis as demonstrated in experiments where a dominant negative version prevented morphology changes in cultured hippocampal neurons [35]. In contrast to Rac1 and Cdc42, RhoA has been implicated in spine destabilization, shrinkage and reduction in density. For example, overexpression of constitutively active RhoA in hippocampal slice cultures fosters spine retraction and elimination [31].

Other small GTPases from families different than Rho have also been implicated in spine morphogenesis. Ras and Rap are a pair of closely related GTPases in the Ras subfamily that share many common regulators and effectors but exert contrasting actions on dendritic spines. Whereas Ras has been shown to stabilize synapses and traffic AMPA receptors into spines in a phosphoinositide 3-kinase (PI3K)–Akt dependent manor [36], Rap has been shown to destabilize spines through B-Raf signaling.

In conclusion, small GTPases are potent regulators of spine remodeling. While only a few members of this large protein family have been studied in spines, the current consensus seems to be that Rac1, Cdc42, and Ras promote spine formation/stability, while RhoA and Rap promote spine destabilization, shrinkage and elimination.

Guanine Exchange Factors and GTPase Activating Proteins

Direct upstream regulators of small GTPases are the members of two large families of proteins, guanine exchange factors (GEFs) and GTPase activating proteins (GAPs). As small GTPases have only two functional states, and thus exhibit by themselves limited regulatory complexity, much of the fine tuning and specificity of these signaling networks is achieved at the level of GEFs and GAPs. These are often large proteins, with a modular mix of protein-protein interaction, targeting, enzymatic, lipid binding, and other domains, through which multiple signaling inputs and outputs are created within each protein. As more than 60 GEF and GAP-encoding genes are present in the human genome, cells are provided with a large repertoire of such proteins, with different domain combinations, that fit very specialized functions. Moreover, different GEFs and GAPs have distinct cellular and tissue distributions as well as subcellular localizations, and they interact with a distinct set of partners.

Through catalyzing the exchange of the GTPase bound GDP to GTP, GEFs serve to activate GTPases. Kalirin-7 is one such GEF, regulating the activity of Rac1. Kalirin-7 is especially unique due to the fact that it is the only known Rac1 GEF expressed in the cortex of adult mice [37]. Overexpression of this GEF in cortical cultures leads to an increase in spine head area and density. Concomitantly, knockdown of kalirin-7 through an RNAi approach reduces the spine area and density [38]. In the hippocampus, the role of kalirin-7 is obscured due to the presence of two other Rac1 GTPases, Tiam1 and β-PIX [28, 39]. Interestingly, mice in which the kalirin gene has been deleted, exhibit many phenotypes reminiscent of schizophrenia including deficits in working memory as well as reduced dendrtitic spine density in the cortex [40]. In the hippocampus, Tiam1 is regulated by NMDA receptor activation and has also been implicated in EphB receptor-dependent dendritic spine development [39, 41]. Likewise, the Rac1GEF β-pix, a downstream target of NMDA receptors, has been shown to be regulated by Cam kinase kinase and Cam kinase I [28].

As RhoA is associated with spine shrinkage and destabilization, GEFs that activate this GTPase have similar effects on dendritic spine morphology. For example, the recently identified GEF-H1 has been shown to co-localize with the AMPA receptor complex and negatively regulate spine density and length through a RhoA signaling cascade [42]. Another GEF involved in the destabilization and shrinkage of spines is Epac2. This Rap1 GEF, is activated by cAMP and leads to reduced spine AMPA receptor content, depressed excitatory transmission, as well as spine destabilization as demonstrated by live imaging studies. Conversely, inhibition of Epac2 leads to spine enlargement and stabilization [43]. Recent studies have associated Epac2 with autism. Thus, further studies centered around this GEF may shed light on the pathology of this complex disorder [44].

Despite their name, GTPase activating proteins (GAPs), serve to inhibit GTPase activity by increasing the rate at which the GTPase bound GTP is hydrolyzed to GDP. For example, SPAR1 has been found to be enriched in dendritic spines of cultured hippocampal neurons. Here it interacts with the PSD-95 and the NMDA receptor complex to dampen Rap activity and enlargement of dendritic spines [45]. In medium spiny neurons, Rap1GAP1 serves an analogous role to SPAR1 in pyramidal neurons. McAvoy et al. showed that overexpression of this GAP leads to increased spine area [46]. Surprisingly, p250GAP is a RhoA GAP associated with the NMDA receptors. Studies where p250GAP was knocked down in primary hippocampal neurons show an increase in dendritic spine width as well as elevated RhoA activity [47].

GAPs are not limited to the regulation of GTPases associated with spine destabilization and shrinkage. SynGAP is a Ras/Rap GAP associated with trafficking of glutamate receptors to the synapse. Heterozygous deletion of SynGAP was sufficient to result in an elevated number of mushroom-shaped spines. In addition, both Ras and Rac activation was decreased in the forebrain of these heterozygous animals. Activation of NMDA receptors in neurons cultured from SynGAP-knockout animals resulted in aberrant cofilin function. Finally, normal EPSPs were also disrupted in hippocampal slices cultured from heterozygous animals [48].

Together, GEFs and GAPs add another layer of GTPase control. GEFs and GAPs are often associated with transmembrane receptors, and receive direct signaling input from at least one, but often multiple such receptors. Next we will focus on some of the better characterized receptor systems that have been shown to modulate spine remodeling, namely those for trophic factors and neurotransmitters.

Trophic Factors

A few trophic factor signaling pathways have been identified to feed into and modulate the above mentioned pathways involved in dendritic spine morphogenesis. The trophic factor Neuregulin-1 (NRG1), polymorphisms in which are associated with schizophrenia, binds to the postsynaptic receptor tyrosine kinase ErbB4. Long term treatment of hippocampal pyramidal neurons with NRG1 has been shown to increase spine density as well as increasing the proportion of spines with a mature phenotype [50]. Furthermore, mouse models in which ErbB4 (along with ErbB2) was knocked-out in the CNS, show a deficit in dendritic spine density in both, the hippocampus and cortex [50]. Another study has demonstrated that the overexpression of ErbB4 leads to an increase in dendritic spine size [51]. Interestingly, this signaling interaction has recently been associated with schizophrenia [52]. There are indications that NRG1/ErbB4 regulate spines through small GTPase signaling.

A lot more is known about the regulation of spines by BDNF/TrkB signaling. The trophic factor, brain-derived neurotrophic factor (BDNF), and its high-affinity receptor, tropomycin-related kinase B (TrkB), have long been associated with synaptic formation and plasticity [53, 54] Numerous studies have reported BDNF-induced changes in dendritic spine density and morphology in a variety of neuron populations [53, 54] . Consistently, TrkB-deficient mice have significantly fewer dendritic spines on CA1 hippocampal neurons [55]. The TrkB receptor has also been show to interact with the Rac1 GEF Tiam1 [56]. This finding further implicates BDNF/TrkB signaling in dendritic spine plasticity. Of all trophic factors, the effects of NRG1 and BDNF on dendritic spine morphogenesis have been best described. However, future studies will identify undoubtedly other neurotrophic factors involved in regulating dendritic spine plasticity.

Neurotransmitter Signaling Regulating Dendritic Spine Plasticity

Other intercellular signaling systems also modulate dendritic spine morphology. Among these, most prominent are neurotransmitters, neuromodulators, and their receptors. However, aside from glutamate, relatively little is known to date about the role of other neurotransmitter systems in spine plasticity. The predominant receptor in regulating dendritic spine plasticity is the NMDA receptor. Numerous studies have shown that many of the signaling molecules mentioned earlier either interact with, or are downstream of NMDA receptors. Upon activation of this receptor, the dendritic spine undergoes a transient increase in calcium concentration [57]. This rise in calcium activates the calcium-sensing calmodulin (CaM). Calcium-bound CaM activates the CamK family of serine/threonine kinases including CamKI, CamKII and CamKIV [58]. These kinases go on to phosphorylate a variety of targets involved in spine structural plasticity including Kalirin-7, as well as other signaling and scaffolding proteins involved in plasticity [38, 59].

Aside from glutamate, other neurotransmitters have been shown to modulate dendritic spine plasticity. Activation of 5-HT2A receptors in pyramidal neurons increased spine size through a kalirin-7-Rac1-PAK-dependent mechanism [60]. This study is of particular importance as it provides a direct link between serotonergic signaling and dendritic spine morphogenesis, both implicated in schizophrenia.

Another important neurotransmitter implicated in dendritic spines is dopamine. A study examining dendritic spine density in the pre-limbic cortex of rats treated with 6-hydroxydopamine, a neurotoxin that selectively ablates dopaminergic and noradrenergic neurons, found a decrease in spine density in this region 3-weaks after toxin administration [61]. Similar findings were also reported in another study [62]. At the molecular level, activation of the D1/D5 receptors with the selective agonist SKF-38393 leads to spine shrinkage through activation of the Rap GEF Epac2 [43].

Data also support a role for the cholinergic system in regulating spines. Ablation of the cholinergic system using 192 IgG-saporin has been shown to decrease dendritic spine density in layer V cortical pyramidal neurons [63]. Furthermore, muscarinic acetylcholine receptors have recently been localized to the extrasynaptic membrane of pyramidal neurons [64], however their exact role in spine morphogenesis has not been determined. In vivo evidence has demonstrated that deletion of the beta2 subunit of the acetylcholine nicotine receptor leads to reduction of spines in the higher order association areas [65].

Less conventional neuromodulators have also been implicated in the regulation of spines. Classically defined as a hormone, estrogen has recently come into the spotlight as an important modulator of dendritic spine plasticity. A study in 2008 by Srivastava et al, demonstrated the non-linearity of signaling pathways. In this study treatment of cortical cultures with estradiol increased spine density while decreasing the AMPA receptor content of spines. These “silent synapses” were potentiated by activation of NMDA receptors, reminiscent of activity dependent maturation of silent synapses during development [66]. These effects were mediated by the Rap/AF-6/ERK1/2 signaling pathways [66]. Additionally, recent studies have demonstrated that treatment of rat cortical cultures leads to phosphorylation of WAVE1 and its targeting to spines, leading to the polymerization of actin [67]. Similar findings have been reported in hippocampal cultured neurons. Here, treatment of hippocampal cultures resulted in increased an increased number of synapses in addition to an increase kalirin-7 localization in dendritic spines [68]. These actions of estradiol seem to be mediated through the ER-β receptor as ER-β but not ER-α agonists are able to recapitulate these effects [68, 69].

Taken together these findings demonstrate the importance of modulatory neurotransmitters signaling in dendritic spine plasticity.

Epigenetic Mechanisms in Dendritic Spine Plasticity

In addition to the classical protein signaling networks, recent studies have implicated novel molecular mechanisms in the regulation of spine plasticity. Among these, epigenetic mechanisms merit special attention. Epigenetics is the study of inherited changes in phenotype caused by mechanisms other than changes in the underlying DNA sequence. A field still in its infancy, epigenetic research has only begun to amalgamate with neuroscience in general and synaptic and spine plasticity specifically. Nevertheless, a handful of studies have begun to elucidate the role of epigenetics in dendritic spine plasticity. The acetylation and deacetylation of histone proteins has been associated with regulation of gene transcription through the loosening of heterochromatin. A study by Guan et al., has shown the role of histone deacetylase 2 (HDAC2) in synaptic plasticity. In this study, the authors showed that neuron-specific overexpression of HDAC2 decreased spine density, synapse number, and enhanced learning. On the other hand, HDAC2 deficits resulted in an increased synapse and spine number. The same was observed in mice treated with HDAC2 inhibitors [70].

DNA methylation is another type of epigenetic modification. Emerging evidence is beginning to implicate this process in the formation of dendritic spines. Repeated exposure of rodents to cocaine increases spine density in the nucleus accumbens. However, this effect is mitigated in my lacking lysine dimethyltransferase G9a, the enzyme involved in transferring methyl groups to DNA cysteine residues [71].

The modulation of compact chromatin and nucleosomes, by deacetylases, kinases, phosphatases and other chromatin modifying enzymes is crucial for the binding of transcription factors (TFs) and initiation of transcription. The effort towards an understanding of epigenetic mechanisms has spilled over into synaptic and dendritic spine plasticity leading to an identification of transcription factors involved in these processes. As such, a recent study has identified two transcription factors, Cux1 and Cux2, in the control of dendritic spine morphology [72]. Cubelos et al., showed that mice deficient in either one of these factors are deficient in dendritic spine density in layer 2/3 but not layer 5 cortical pyramidal neurons [72].

Whereas the above mentioned transcription factors are crucial for dendritic spine maintenance, the TF MEF2 is needed for proper synapse elimination. Flavell et al., showed that MEF2 is involved in synapse elimination in hippocampal neuron cultures. Activity dependent dephosphorylation of MEF2 leads to the expression of activity-regulated cytoskeletal-associated protein (Arc) and SynGAP. Since Arc has been associated with AMPA receptor endocytosis [73], dephosphorylation of MEF2 induces destabilization of the synapse leading to spine elimination. These in vitro observations have recently been confirmed in vivo where Mef2c deficient mice showed an elevation in dendritic spine number compared to control [74, 75]. Additionally, MEF2 has also been shown to be involved in the spine elimination of striatal medium spiny neurons [76].

After TFs initiate transcription micro-RNAs (miRNAs) regulate protein expression by binding to mRNA and suppressing its translation. MiRNAs that are involved in dendritic spine formation and stabilization are just begging to emerge. Morgan Sheng’s group has recently discovered two miRNAs, miR-125b and miR-132 which associate with the fragile X mental retardation protein (FMRP) as knockdown of this protein ameliorates the effect of these miRNAs on spine morphology [77]. In this study they demonstrated that the NR2A subunit of the NMDA receptor is the target for both of these miRNA, however, their effect on spine morphology is conflicting. Whereas overexpression of miR-125b induced long narrow spines correlated with reduced synaptic transmission, overexpression of miR-132 lead to enlarged spine heads [77]. MiR-132 has also been associated with another mechanism. Through suppression of p250GAP translation, miR-132 induces spine formation. A similar effect is seen with knockdown of p250GAP. Inhibition of miR-132 results in smaller dendritic spines and reduced EPSCs [78].

Another cytoskeletal regulatory molecule that is regulated at the miRNA level is LIMK1. Schratt et al., found that miR-134 suppresses LIMK1, however, exposure of cultured cells to BDNF relieves this suppression and contributes to spine stabilization [79]. The ubiquitin ligase mind bomb-1 (MB1) is another miRNA target recently identified to be involved in regulating dendritic spines. Smart et al., have shown that suppression of MB1 by miR-137 results in a decrease in dendritic spine density in hippocampal cultured neurons [81].

A novel and interesting way by which miRNAs regulate dendritic spine plasticity is by inhibiting the translation of a protein involved in palmitoylation. MiR-138 has been shown to negatively control dendritic spine size in rat hippocampal neurons through the regulation of acyl protein thioesterase 1 (APT1) expression. Because APT1 is involved in the palmitoylation of proteins at the synapse, its modulation may effects dendritic spine morphogenesis through regulating the targeting of synaptic proteins to the synapse [80].

Although the field of epigenetic control of dendritic spine plasticity has only begun to emerge, the importance of control at the RNA level has already been demonstrated. Undoubtedly, more miRNA targets involved in the dendritic spine regulation will soon be elucidated.

Conclusions

Dendritic spine plasticity is a relatively young field. Recent technological advancement, particularly imaging technologies, have allowed for the study of the above molecules and their roles in these plasticity events. As fluorescent probes and delivery techniques continue to be developed, studies demonstrating the dynamic nature of signaling events will begin to phase out the currently used static approaches. For example, two seminal studies recently utilized time-lapse imaging to elucidate the molecular mechanisms in spine plasticity [43, 66]. Furthermore, development of 2-photon imaging has already revolutionized our understanding of the structural and functional dynamics of spines, through in vivo studies of the intact cortex [82-84]. As we move through the 21^st century, computational analysis and modeling of molecular pathway networks will provide a more comprehensive understanding of the non-linear molecular interactions regulating spine plasticity. This is particularly important as many proteins involved in spine plasticity have been implicated in various psychiatric disorders. Of particular importance are studies that show disease-related phenotypes in concord with dendritic spine aberrations such as the study by Cahill et al [40]. Additionally, there exists a dearth of studies in which an observed anatomical phenotype in correlation with a molecular abnormality is modeled in a laboratory setting. In conclusion, understanding the functional relationships between different signaling molecules associated with a particular disorder will undoubtedly shed light on the underpinnings of pathology as well as identify possible targets for treatment.

Figure.

Schematic representation of the molecules that regulate bidirectional dendritic spine plasticity. Pathways that promote spine enlargement and stabilization are on the right; pathways promoting spine shrinkage and destabilization are on the left.