Small G-protein Signaling in Neuronal Plasticity and Memory Formation: the Specific Role of Ras Family Proteins

Neurofibromin

The first evidence suggesting a role of Ras in learning and memory comes from studies on neurofibromatosis type I (NF1), an autosomal dominant disease caused by loss-of-function (LOF) mutations in neurofibromatosis type 1 oncogene (nf1). 30–60% of children with NF1 have learning disabilities (North et al., 2002; North et al., 1997; Ozonoff, 1999; Riccardi, 1981). The nf1 gene encodes neurofibromin, a protein identified as a RasGAP (Figure 1B). LOF nf1 mutations result in decreased cycling from the active to inactive state of Ras, thus leading to Ras hyperactivity. This protein is universally expressed, but its expression level is highest in brain (Daston et al., 1992; Gutmann et al., 1991; Nordlund et al., 1995).

Mice with heterozygous nf1 knockout (Nf1^+/−) were established as an animal model for NF1 disease by Jacks et al. (1994). Subsequently, Silva et al. (1997) found that these mice showed impaired spatial memory in the Morris Water Maze, a hippocampal-dependent memory task, suggesting a role of neurofibromin in the hippocampus during memory processing. Consistent with this idea, Nf1^+/− mice were able to overcome their performance deficit with extended training, which bypasses the requirement of the hippocampus (Packard and McGaugh, 1996; Pouzet et al., 2002). The impairment in spatial memory in Nf1^+/− mice was exacerbated with a further heterozygous mutation in the NR1 subunit of NMDA receptors (NMDARs), supporting the findings that the severity of learning disabilities in NF1 patients co-varies with genetic backgrounds (Easton et al., 1993). A previous study has suggested that neurofibromin is expressed in many brain regions (Nordlund et al., 1995). However, Silva et al. (1997) did not observe any memory deficit of Nf1^+/− mice in cued fear conditioning, a form of amygdala-dependent and hippocampal-independent memory task, suggesting that neurofibromin does not affect amygdala processing of fear memory. It requires future studies whether this lack of effect is due to differential expression levels of neurofibromin in the amygdala compared to the hippocampus. The role of neurofibromin in memory and oncogenesis can be dissociated. Mice carrying a homozygous knockout of exon 23a of nf1 (NF1^23a−/−), which encodes part of the GAP-related domain, showed no predisposition for tumor formation and developed normally. However, they exhibited impaired spatial memory (Costa et al., 2001). This study also highlights the indispensable role of the GAP activity of neurofibromin in memory formation.

Follow-up studies were carried out to explore the cellular and molecular mechanisms underlying memory impairment in Nf1^+/− mice. Costa et al. (2002) found that the spatial memory deficit in Nf1^+/− mice was rescued by a further heterozygous knockout in K-Ras or N-Ras. Pre-training injection of an FTase inhibitor or lovastatin, both of which inhibit membrane anchoring of Ras, and thus disrupt Ras signaling, also rescued the memory impairment (Costa et al., 2002; Li et al., 2005). Consistent with behavioral findings, Nf1^+/− mice showed impaired LTP at hippocampal Schaffer collateral-CA1 synapses in response to theta-burst stimulation (TBS), which was rescued by inhibitors or LOF mutations of Ras. Furthermore, Guilding et al. (2007) found that basal ERK phosphorylation and downstream CREB phosphorylation was elevated in the hippocampus of Nf1^+/− mice, but there was no change in the PI3K cascade, suggesting that neurofibromin specifically suppresses the ability of Ras to activate the Raf-MEK-ERK-CREB cascade. Application of a subthreshold dose of a MEK inhibitor, which reversed the abnormal increase in ERK and CREB phosphorylation, reversed the LTP deficit in Nf1^+/− mice. Collectively, these findings suggest that the abnormal increase in Ras activity may account for memory impairment in Nf1^+/− mice.

Interestingly, Costa et al. (2002) found that in the presence of picrotoxin, a GABA antagonist, LTP was normally expressed in Nf1^+/− mice. Cui et al. (2008) went on to examine the role of neurofibromin in different cell types in the brain. Cell-specific heterozygous deletion of nf1 in GABAergic neurons, but not in excitatory neurons or astrocytes, resulted in deficits in spatial memory and hippocampal LTP, similar to those observed in Nf1^+/− mice. Consistent with these observations, GABA release was elevated in Nf1^+/− mice. Furthermore, a subthreshold dose of picrotoxin, which did not affect wild-type mice, rescued the memory deficit in Nf1^+/− mice. These data suggest that the engagement of neurofibromin in memory formation is cell-type specific: it restricts GABA release at inhibitory synapses and thus favors potentiation of synaptic strength.

Studies in Drosophila also suggest that neurofibromin plays important roles in learning and memory. In an aversive conditioning task, Nf1-deficient flies showed both a learning deficit immediately after one-cycle training and a long-term memory deficit 24hr after spaced repeated-trial training (Guo et al., 2000; Ho et al., 2007). Interestingly, these flies performed normally in memory tests after massed training, suggesting that the engagement of neurofibromin in memory formation is sensitive to the pattern of training trials. Neurofibromin has an N-terminal GAP-related domain, and a C-terminal region which regulates cAMP levels by mediating G-protein coupled receptor (GPCR) -dependent activation of adenylate cyclase. Ho et al. (2007) found that the C-terminal region of neurofibromin was important for immediate memory, whereas its GAP activity was specifically required for the formation of long-term memory following spaced training.

In summary, these data suggest that hyperactive Ras induced by LOF mutations in neurofibromin, a specific RasGAP, can result in memory impairment. Despite the ubiquitous distribution of neurofibromin, surprisingly, its engagement in memory formation is specific to (i) certain types of memory tasks, (ii) GABAergic neurons (but not excitatory neurons or astrocytes), and (iii) Ras-ERK cascade (but not Ras-PI3K cascade) (Figure 3A). Findings in Drosophila suggest that neurofibromin can also regulate learning by Ras-independent activation of adenylate cyclase. This regulation has also been found in rodents (Tong et al., 2002), but its contribution to learning in rodents has not been reported. Furthermore, despite the importance of neurofibromin in cognitive function, little is known about how the activity of neurofibromin is regulated, especially by stimuli that give rise to memory formation and synaptic plasticity. Finally, it should be noted that in addition to learning disabilities, NF1 patients have many other cognitive deficits (North et al., 2002; North et al., 1997; Ozonoff, 1999; Riccardi, 1981). For example, most NF1 children are hyperactive and have attention problems, which are also observed in Nf1^+/− mice and can be rescued by Ras inhibitors. The degree to which this attention deficit affects their performance in memory tasks needs to be carefully evaluated in the future.

SynGAP

SynGAP was identified through screening of PSD-95/Discs large/zO-1 (PDZ) binding proteins and sequencing of postsynaptic density (PSD) proteins (Chen et al., 1998; Kim et al., 1998). SynGAP is highly expressed in the brain, especially in CA1 and dentate gyrus of the hippocampus. It is enriched in the PSD fraction of the brain, and directly interacts with PSD-95 and NMDARs in spines. SynGAP contains a C2 domain which has a Ca²⁺-binding motif, and this domain is essential for turning on its GAP activity (Pena et al., 2008) (Figure 1B). SynGAP is dually specific for both Ras and Rap; however, an in vitro study showed that it has greater potency for Rap (Krapivinsky et al., 2004).

Electrophysiological recording in hippocampal slices from SynGAP heterozygous knockout mice (SynGAP^+/−) showed that Schaffer collateral-CA1 LTP induced by a variety of stimulation protocols was impaired, suggesting a crucial role of SynGAP in LTP (Kim et al., 2003; Komiyama et al., 2002). This deficit could not be rescued by a further knockout of H-Ras, suggesting H-Ras may not be involved in SynGAP signaling. However, this study does not rule out the possibility of other Ras isoforms and other members in Ras family as SynGAP targets. Compared to a more general role of SynGAP in LTP, the engagement of SynGAP in LTD is specific to certain LTD-inducing stimuli. It was reported that in SynGAP^+/− mice, LTD induced by 1Hz stimulation is intact; however, LTD induced by NMDA treatment is significantly impaired (Carlisle et al., 2008; Kim et al., 2003). Finally, Rumbaugh et al. (2006) found that in cultured cortical neurons, SynGAP negatively regulated ERK signaling and positively regulated p38MAPK signaling. It also suppressed surface expression of AMPA receptors (AMPARs), postsynaptic glutamate receptors are important for the expression of LTP and LTD. The effects of SynGAP were lost with LOF mutations in the GAP domain, suggesting that its GAP activity is crucial during the formation of different forms of synaptic plasticity, mostly likely acting through its regulation of the activity of Ras family proteins.

At the behavioral level, Komiyama et al. (2002) and Muhia et al. (2010) observed impairment in the formation of spatial memory of SynGAP^+/− mice in the Morris Water Maze. However, both contextual and cued fear conditioning as well as novel object recognition appear to be normal in these mice (Muhia et al., 2009; Muhia et al., 2010). In addition to spatial memory deficits, SynGAP^+/− mice exhibited a range of behavioral and cognitive impairments, such as hyperactivity, reduction in anxiety, working memory deficit, social isolation, as well as an impairment in controlling operant behaviors (Muhia et al., 2009; Muhia et al., 2010). The complexity of the behavioral phenotype of SynGAP^+/− mice makes it hard to interpret whether their impairment in memory tasks is due to deficits in learning and memory or to changes in performance variables such as arousal state.

Taken together, the current findings suggest an important role for SynGAP in hippocampal LTP and certain types of LTD, which may be mediated by Ras-ERK and Rap1-p38MAPK cascades (Figure 3B). Furthermore, SynGAP is critical for a variety of cognitive functions. However, to elucidate its specific function in learning and memory, conditional SynGAP knockouts in a spatially and temporally restricted fashion are required.

RasGRF

In contrast to neurofibromin and SynGAP, which suppress Ras activity, RasGRF1 and RasGRF2 are GEF proteins, promoting Ras activation. They are both abundantly expressed in brain (Chen et al., 1993; Shou et al., 1992). Activation of RasGRFs following increased intracellular Ca²⁺ concentration or GPCR activation leads to increased activity of both Ras and Rac, a small G-protein in the Rho family (Fan et al., 1998; Innocenti et al., 1999) (Figure 1B). Thus, RasGRFs may coordinate the activation of Ras and Rac to change synaptic strength.

Brambilla et al. (1997) have generated a line of RasGRF1 knockout mice (RasGRF1^−/−). Although RasGRF1 expression was clearly lost in both hippocampus and amygdala, these mice exhibited impaired memory only in amygdala-dependent tasks, such as inhibitory avoidance, and contextual and cued fear conditioning. However, spatial learning and memory, which critically rely on hippocampal function, but not amygdala, was normal. Consistent with their behavioral findings, amygdala LTP was impaired in these mice, whereas hippocampal LTP was unaffected. However, Giese et al. (2001) found opposite results with a different line of RasGRF1^−/− mice: they showed impairment in hippocampal-dependent memory tasks, but not amygdala-dependent tasks. It is not clear what factors may contribute to the discrepancy in these findings. It may be due the differences in the training protocols or the differences in the genetic backgrounds of these two lines of mice. Nevertheless, both studies suggest a role for RasGRF1 in memory formation. This general conclusion is further supported by a recent study by Fasano et al. (2009), who examined the same line of RasGRF1^−/− mice used by Brambilla et al. (1997), as well as a line of mice mildly over-expressing Ras-GRF1 (RasGRF1^OE). They found that hippocampus-dependent spatial memory in these mice was comparable to wild-type controls. However, passive avoidance learning requiring the amygdala was impaired in RasGRF1^−/− mice and facilitated in RasGRF1^OE mice. These findings are consistent with those from Brambilla et al. (1997). They also found that RasGRF1 was expressed in adult striatum and mediated ERK activation induced by dopamine or glutamate treatment. Given that striatum plasticity is significantly affected by drug addiction (Chang et al., 2007; Fasano and Brambilla, 2002), they explored the role of RasGR1 in cocaine addiction. They found that the increase in locomotion following chronic cocaine treatment was impaired in RasGRF1^−/− mice, and enhanced in RasGRF1^OE mice. In addition, RasGRF1^−/− mice exhibited impaired performance whereas RasGRF1^OE mice showed enhanced preference in a conditioned place preference task, suggesting that RasGRF1 can also modulate maladaptive memory in drug addiction.

A number of studies have also examined memory formation and synaptic plasticity in RasGRF1 and RasGRF2 double knockout mice (RasGRF1^−/−/RasGRF2^−/−). For example, Li et al. (2006a) found that post-pubescent RasGRF1^−/−/RasGRF2^−/− mice exhibited impaired LTP and LTD at hippocampal Schaffer collateral-CA1 synapses. Further examination using hippocampal slices from RasGRF single knockout mice revealed that RasGRF1 mainly regulates LTD by mediating the signaling between the NR2A subunit of NMDARs and p38MAPK activation. In contrast, RasGRF2 significantly contributes to LTP formation by linking NR2B activation to ERK activation. Interestingly, the deficits in LTP, but not LTD, in young RasGRF1^−/−/RasGRF2^−/− mice can be rescued by exposure of adolescent mice to an enriched environment (Li et al., 2006b). Early-age exposure to an enriched environment can also rescue contextual fear memory deficits in these mice (Arai et al., 2009). These findings suggest that the engagement of RasGRFs in memory formation is influenced by the previous experiences of animals during the adolescence.

RasGRFs also regulate electrophysiological properties of neurons other than LTP and LTD. Enhanced basal synaptic transmission in both amygdala and hippocampus has been reported in RasGRF1^−/− mice (Brambilla et al., 1997). In addition, Tonini et al. (2001) reported that the excitability of hippocampal neurons was significantly increased in RasGRF1^−/− mice. Finally, Li et al. (2006a) found that RasGRF1/2 double knockout or RasGRF2 knockout reduced paired-pulse facilitation (PPF) at hippocampal Schaffer collateral-CA1 synapses, suggesting that RasGRF2 may also regulate presynaptically-expressed short-term plasticity.

In summary, RasGRFs are important for regulating synaptic plasticity and intrinsic properties of neurons. Evidence suggests that RasGRFs regulate synaptic plasticity by transforming Ca²⁺ influx from NMDARs to activation of Ras family proteins and subsequent activation of ERK and p38MAPK signaling cascades (Figure 3C). The effects of RasGRFs on synaptic plasticity and neuronal function may likely underlie the observed role for these Ras-family regulators in memory.

Epac (cAMP-GEF)

It is well-known that cAMP can enhance neurotransmitter release (Chavez-Noriega and Stevens, 1994; Chen and Regehr, 1997; Zhong and Wu, 1991). Canonically, it is thought to exert its effect through protein kinase A (PKA) (Capogna et al., 1995; Trudeau et al., 1996). However, it was reported that cAMP-mediated synaptic facilitation was not completely eliminated by PKA inhibitors, suggesting the existence of other cAMP targets (Beaumont et al., 2002). Exchange Protein directly Activated by cAMP (Epac) was identified as a RapGEF that is activated by cAMP (de Rooij et al., 1998; Kawasaki et al., 1998) (Figure 1B). Thus, it was also called cAMP-GEF. cAMP binding to Epac leads to an increase in Rap1 and Rap2 activity, without affecting H-Ras or R-Ras, and the Rap activation was independent of PKA activity. Two isoforms of Epac have been identified. Epac1 is universally expressed, whereas Epac2 is predominantly expressed in the brain.

The investigation of the function of Epac has been greatly facilitated by the innovation of cAMP analogs that specifically activate either Epac or PKA. Presynaptic loading of an Epac-specific cAMP analog increases postsynaptic responses in the Calyx of Held in rats, suggesting a role for Epac in the regulation of neurotransmitter release (Kaneko and Takahashi, 2004). Similar results have been found in a number of other experimental systems (Cheung et al., 2006; Gekel and Neher, 2008; Zhong and Zucker, 2005). In mouse hippocampus, Epac activation enhances Schaffer collateral-CA1 LTP. A single train of 100Hz stimuli usually only induces translation-independent early-LTP which decays within 30min. However, in the presence of an Epac-specific cAMP analog, the induced LTP can last for at least 2hr and is dependent on translation and ERK activity (Gelinas et al., 2008). In this study, a change in basal transmission or PPF by Epac-specific cAMP analog was not observed, implying that, in this instance, Epac may not be involved in presynaptic plasticity. Ster et al. (2009) reported that application of a higher concentration of Epac-specific cAMP analog to hippocampal slices caused synaptic depression which was dependent on Rap1 and p38MAPK signaling but not on ERK. They also did not observe any change in paired-pulse facilitation. Similarly, Woolfrey et al. (2009) found that Epac activation decreased AMPAR-mediated synaptic transmission in cultured rat cortical pyramidal neurons.

At the behavioral level, Ma et al. (2009) reported that bilateral infusion of cAMP analogs specific for either PKA or Epac to mouse dorsal hippocampus immediately after a weak training protocol for contextual fear conditioning enhanced long-term memory tested 24hrs later. In addition, infusion of PKA inhibitors following a strong training protocol impaired memory, which could be rescued by Epac-specific cAMP analog. These data suggest a role for Epac in memory formation and that Epac and PKA may have overlapping downstream targets during this process. Furthermore, cAMP is known to play important roles in hippocampus-dependent memory retrieval (Isiegas et al., 2008; Izquierdo et al., 2000). Dbh^−/− mice, which are deficient in cAMP production, exhibit impairment in the retrieval of memory for contextual fear conditioning. Ouyang et al. (2008) found that infusion of general cAMP analogs, or a combination of PKA-specific and Epac-specific cAMP analogs to dorsal hippocampus before test, rescues the retrieval deficit in Dbh^−/− mice. Interestingly, either PKA- or Epac-specific cAMP analog alone had no effect, suggesting a synergistic interaction between PKA and Epac signaling cascades during memory retrieval. The role for Epac in memory retrieval is further supported by a study performed by Ostroveanu et al. (2009): infusion of Epac-specific cAMP analog to dorsal hippocampus before test enhanced memory retrieval in contextual fear conditioning and passive avoidance. Moreover, infusion of Epac2 siRNA to knockdown the major isoform of Epac in hippocampus blocked memory retrieval.

Collectively, Epac has been reported to be engaged in a number of forms of synaptic plasticity (Figure 3D), including enhancing neurotransmitter release, facilitating LTP, and inducing synaptic depression. However, there are inconsistencies in the reports regarding where and how Epac contributes to plasticity, which may be due to differences in the composition of molecular networks in specific model systems or in the concentration and application time of Epac-specific cAMP analogs. In addition, Epac can recruit divergent signaling cascades (ERK and p38MAPK, respectively) in the service of synaptic potentiation versus synaptic depression. It will now be of interest to examine the factors that control this differential routing of Epac in the service of specific forms of synaptic plasticity. Furthermore, current studies on Epac in memory consolidation and retrieval have only used fear memory tasks, which are known to trigger neuromodulator release to generate cAMP in hippocampal neurons. This may implicate Epac is a critical element bridging the interactions between neuromodulatory systems and the hippocampus. It is not clear whether Epac has similar roles in non-emotional memory. It is also important to examine how the behavioral effects of Epac link to its synaptic and molecular functions. Finally, cAMP has also been reported to activate another GEF protein, CNRasGEF, which is also highly expressed in a variety of brain regions (Pham et al., 2000). Interestingly, CNRasGEF can activate both Ras and Rap1; however, its activation of Ras requires cAMP binding, while its activation of Rap1 is cAMP-independent. The role of CNRasGEF in memory formation and neuronal plasticity has not yet been examined.

Other Regulators

In addition to GAPs and GEFs, a few other proteins have recently been found to directly regulate the activity of Ras family proteins during memory formation.

Suprachiasmatic nucleus (SCN) circadian oscillatory protein (SCOP) was originally identified in SCN, and its expression is regulated by circadian rhythms (Shimizu et al., 1999). Later, SCOP has been found to be expressed in hippocampus. It localizes in lipid rafts, membrane microdomains that are highly packed with protein receptors and flanked by glycosphingolipids. SCOP specifically binds to K-Ras, which is also present in lipid rafts. This interaction prevents K-Ras from binding to guanine nucleotides, thus holding K-Ras in the inactive state (Shimizu et al., 2003). Shimizu et al. (2007) reported that object recognition training in mice induced rapid degradation of SCOP in the hippocampus, in parallel with an increase in ERK activation. Pre-training over-expression of SCOP blocked the formation of long-term memory for object recognition without affecting short-term memory. These findings suggest that SCOP is an inhibitory constraint for long-term memory formation.

Sprouty-related protein with an EVH-1 domain (Spred1) is another inhibitor for Ras-Raf signaling, predominantly expressed in adult mouse brain (Engelhardt et al., 2004). It strengthens the binding between Ras and Raf, preventing Raf from being activated by Raf kinases (Wakioka et al., 2001). Denayer et al. (2008a) reported that Spred1 knockout mice (Spred1^−/−) exhibited impaired spatial learning and memory in the Morris Water Maze and the T-maze. In parallel with behavioral deficits, these mice showed reduced basal synaptic transmission which was especially pronounced at high stimulation strength. They also showed increased PPF at 10 and 20ms interpulse intervals compared to wild-type controls, suggesting a change in presynaptic function. In addition, LTP was decreased whereas LTD was increased in Spred1^−/− mice. At the molecular level, ERK activation in the hippocampus of Spred1^−/− mice during LTP induction was significantly enhanced, suggesting that hyperactive Ras-ERK signaling might contribute to the changes in synaptic plasticity and memory formation in these mice.

Finally, Corkscrew (csw; Drosophila homolog of SHP-2) is a tyrosine phosphatase that transduces signals from receptor tyrosine kinases to activate Ras-ERK signaling, the precise mechanism underlies which is not clear (Lu et al., 1993; Matozaki et al., 2009; Oishi et al., 2006; Perkins et al., 1996). Pagani et al. (2009) have examined the role of csw in memory formation induced by different patterns of training in Drosophila. Interestingly, they found that the magnitude of csw activity determined the optimal spacing between training trials in an olfactory aversive conditioning task. In flies over-expressing wild-type csw, ERK was activated faster with a normal post-trial deactivation rate; this allowed memory to be induced by training with shorter inter-trial intervals. In contrast, in flies expressing a gain-of-function (GOF) mutant csw, ERK was also activated faster but the deactivation was delayed. In these flies, long-term memory could only be induced with longer inter-trial intervals. This study suggests a role of csw-mediated Ras-ERK signaling in the spacing effect of memory formation.

Taken as whole, studies at the level of upstream regulators reveal an interesting point: either hyper-activation (by inhibiting negative regulators) or inhibition (by inhibiting positive regulators) of Ras family proteins can cause impairment in neuronal plasticity and memory formation. These data suggest that the signaling of Ras family proteins is required for neuronal and behavioral plasticity. However, elevation of these signaling cascades in inhibitory neurons (such as by neurofibromin knockout, Figure 3A), can result in memory impairment. Even in excitatory neurons, the signaling from Ras family proteins may require a particular dynamic range in order to integrate signals properly. Thus, if the signaling from Ras family proteins is elevated in the basal state, there may not be sufficient range available for it to be activated properly when responding to signals, thereby occluding memory formation. This appears to be common theme in the nervous system and may be involved in a wide range of brain disorders (Hoeffer and Klann, 2010; Kelleher and Bear, 2008).

Ras

As described earlier, membrane association is critical for Ras signaling. A number of studies have examined the role of Ras in memory formation by local infusion of FTase inhibitors (which disrupt the attachment of Ras to membranes) into specific brain regions during different stages of memory processing. Merino et al. (2006) found that pre-training infusion of an FTase inhibitor into basolateral amygdala (BLA), but not other amygdala sub-regions, disrupted long-term memory for contextual and cued fear conditioning. Intra-BLA infusion of the inhibitor post training or prior to testing had no effect on long-term fear conditioning, suggesting a requirement of localized Ras signaling in BLA during the induction of fear memory. Similar findings have been reported by Ou et al. (2006). At the celluar level, FTase inhibitors attenuate LTD in hippocampal dentate gyrus (Murray and O’Connor, 2004) and disrupt LTP in the hippocampal CA1 region (O’Kane et al., 2004). However, caution should be expressed in interpreting these results because (i) FTase inhibitors target a number of G-proteins (thus, their phenotypic effects may not be due to specific inhibition of Ras signaling), and (ii) FTase inhibitors target newly synthesized Ras without affecting pre-existing functional Ras. Thus, a lack of effect of the inhibitors cannot completely exclude a contribution of Ras signaling to the process.

In addition, genetic manipulations have been utilized to examine the function of specific Ras isoforms in memory formation. Ohno et al. (2001) have combined pharmacological and genetic manipulations to study the role of K-Ras in plasticity. They found that K-Ras heterozygous knockout mice (K-Ras^+/−) showed robust memory for contextual fear conditioning and normal hippocampal Schaffer collateral – CA1 LTP and ERK activation. However, a sub-threshold dose of MEK inhibitor, which did not affect wild-type mice, blocked memory formation, LTP and ERK activation in K-Ras^+/− mice, revealing an engagement of K-Ras-ERK signaling in memory processing. Another Ras isoform, H-Ras, is abundantly expressed in CNS and relatively enriched in synaptosomes. Manabe et al. (2000) found that H-Ras knockout (H-Ras^−/−) mice showed increased tyrosine phosphorylation of NMDARs in hippocampus. In addition, NMDAR responses and hippocampal LTP induced by high-frequency stimulation (HFS) were larger, suggesting that H-Ras can restrain LTP induction by inhibiting NMDAR function. Interestingly, Komiyama et al (2002) found that LTP induced by pairing low-frequency stimulation with postsynaptic depolarization was normal in H-Ras^−/− mice, suggesting that the engagement of H-Ras in LTP is sensitive to the stimulation patterns. Kushner et al. (2005) examined mice expressing a constitutively active form of H-Ras (H-Ras^G12V) in forebrain postnatal neurons. Electron microscopy identified that H-Ras^G12V was predominantly localized in axon terminals of CA1 pyramidal neurons, and that H-Ras^G12V-containing axon terminals had more vesicles docking at active zones than wild-type controls. H-Ras^G12V mice showed enhanced synaptic potentiation, even in the presence of NMDAR blockers. In addition, these mice exhibited enhanced acquisition of spatial memory in the Morris Water Maze, and enhanced short-term and long-term conditioned responses in contextual fear conditioning. The enhancement in synaptic potentiation and memory caused by H-Ras^G12V expression was reversed by knockout of Synapsin I, a protein involved in neurotransmitter release, suggesting that H-Ras can also contribute to memory formation by mediating presynaptic facilitation. Collectively, these studies suggest that the two Ras isoforms, K-Ras and H-Ras, regulate distinct aspects synaptic plasticity during memory formation.

Rap1

Rap1 shares ~50% sequence identity to Ras. The ability of Rap1 to activate b-Raf, which is predominantly expressed in the brain and leads to persistent ERK activation, has identified it as an attractive target for the studies of molecular mechanisms underlying learning and memory (Storm et al., 1990; York et al., 1998). Morozov et al. (2003) generated a line of transgenic mice carrying inducible dominant-negative Rap1 (iRap) in the forebrain. Induced expression of iRap resulted in decreased b-Raf but increased c-Raf activity in hippocampus. The imbalance in Raf activity specifically reduced the activity of a membrane-associated pool of ERK, and decreased ERK-mediated phosphorylation of the A-type potassium channel, Kv4.2. In parallel with these molecular changes, iRap expression reduced cAMP-dependent forms of LTP and the excitability of CA1 pyramidal neurons. Furthermore, pre-training expression of iRap impaired spatial learning and memory, and context discrimination. Collectively, these findings suggest a model in which Rap1 couples cAMP signaling to selectively activate a membrane-associated pool of ERK to control (i) the excitability of CA1 pyramidal neurons, (ii) the induction of LTP, and (iii) memory formation. In another study, Pan et al. (2008) examined mice with postnatal forebrain knockout of Rap1 (Rap1^−/−). These mice exhibited deficits in short-term and long-term memory in both contextual and cued fear conditioning induced by moderate training, but not by strong training. They also showed reduced short-term generalization of contextual fear conditioning. Both the forebrain and the thalamus project to the amygdala, an important locus for associative fear memory. Rap1^−/− mice showed specific LTP impairment at cortico-amygdala synapses, whereas LTP in the thalamic-amygdala pathway was spared. In addition, at cortico-amygdala synapses of Rap1^−/− mice, there was an increase in basal synaptic transmission. This study suggests that in the amygdala, Rap1 in presynaptic terminals of projections from the forebrain is critical for fear conditioning.

Rap2

Rap2 is another major member of the Ras family, sharing ~50% identity to Ras and ~60% identity to Rap1 (Pizon et al., 1988). It has been reported as an antagonist of Ras signaling. It can also activate JNK signaling cascades (Machida et al., 2004; Taira et al., 2004). Ryu et al. (2008) found that transgenic mice expressing constitutively active Rap2 (Rap2^V12) under the control of the CaMKII promoter show reduced ERK activity, enhanced Schaffer collateral-CA1 LTD and decreased spine density in hippocampal CA1 region. Behaviorally, these mice exhibit hyperactivity and impaired spatial learning. In addition, although contextual fear conditioning was normal, Rap2^V12 mice showed impaired extinction of contextual fear conditioning, which was accompanied by a reduction in ERK activation in hippocampus during early extinction trials. Collectively, these data suggest that Rap2 opposes ERK activity in Ras-ERK signaling in hippocampus, which may in turn regulate synaptic plasticity and cognitive function.

Overall, studies at the level of Ras family proteins provide direct and strong support for the hypothesis that the signaling of Ras family proteins is essential for regulating a variety of forms of neuronal plasticity in different brain regions, and that dynamic changes in their activity are crucial for the formation of many types of memories that are supported by plasticity within these same brain regions.

b-Raf

Ras family proteins signal through Raf to activate the MEK-ERK signaling cascade (Figure 2A). There are three isoforms of Raf: a-Raf, b-Raf and c-Raf (Raf-1). Among them, b-Raf has attracted special interest in neurobiology studies because it is predominantly expressed in the brain and is the major isoform of Raf in neurites (Storm et al., 1990). In addition, compared to other isoforms, b-Raf is a stronger MEK-ERK activator: it has higher affinity for MEK and mediates sustained ERK activation (Papin et al., 1996). To directly evaluate the role of b-Raf in memory processing, Chen et al. (2006) generated a line of conditional b-Raf knockout mice using the Cre-loxP system to restrictively knockout b-Raf expression in the postnatal forebrain neurons (b-Raf^−/−). These mice exhibited impaired spatial learning and contextual discrimination, while contextual fear conditioning and conditioned taste aversion were spared. They also showed impaired hippocampal Schaffer collateral – CA1 LTP. At the molecular level, training-induced ERK activation in the hippocampus was significantly reduced. Collectively, these findings suggest a critical role of b-Raf in regulating hippocampal ERK activation, LTP, and memory formation.

PI3K

In contrast to Raf, PI3K activated by Ras family proteins phosphorylates phosphatidylinositol lipid substrates, which in turn activate Akt (PKB) to regulate mTOR-dependent protein synthesis, a type of translation closely associated with cell growth. In addition, PI3K activation can lead to the activation of a number of GAPs and GEFs, thus PI3K acts both upstream and downstream of Ras family proteins (Rodriguez-Viciana et al., 1996a; Vanhaesebroeck et al., 2010) (Figure 2B).

Lin et al. (2001) reported that pre-training infusion of PI3K inhibitors to lateral or basolateral amygdala (LA/BLA) impaired the induction of long-term, but not short-term, cued fear conditioning, and blocked training-induced Akt and CREB activation, suggesting that PI3K signaling is specifically required for long-term memory. In hippocampus, administration of PI3K inhibitors to the CA1 region impaired acquisition, consolidation and retrieval of memory for inhibitory avoidance (Barros et al., 2001). It also impaired the retrieval and extinction, but not the acquisition or consolidation, of contextual fear conditioning (Chen et al., 2005). In addition, Mizuno et al. (2003) reported that chronic systemic injection of a PI3K inhibitor, wortmannin, impaired hippocampal-dependent spatial learning in a radial arm maze task. In contrast, Horwood et al. (2006) did not observe a requirement of PI3K in the same memory task using a different PI3K inhibitor, LY294002, whereas the inhibitor impaired long-term memory for object recognition. It is not clear what factor(s) caused these contradictory results. One possibility is that different types, concentrations and application time of the inhibitors used in these two studies might differentially affect Akt versus other downstream effectors of PI3K, which in turn could result in different learning effects. Furthermore, Sui et al. (2008) recently found that inhibition of PI3K in medial prefrontal cortex (mPFC) immediately after training blocked long-term memory for trace fear conditioning. Finally, Tohda et al. (2007) examined knockout mice lacking the p85α regulatory subunit of PI3K, which showed reduced Akt phosphorylation in cerebral cortex, striatum, hippocampal DG and CA3 regions, but not in CA1 region. These mice exhibited impaired spatial learning as well as motivation deficits and restlessness in the Morris water maze task. Thus, PI3K signaling is important in various forms of memory as well as other cognitive functions.

At the cellular level, PI3K inhibitors block amygdala LA-BLA LTP, as well as Akt and ERK activation induced by tetanic stimulation, suggesting a linkage of PI3K to ERK signaling in synaptic plasticity (Lin et al., 2001). HFS-induced LTP in mPFC and tetanus-induced LTP in hippocampal dentate gyrus were also impaired in the presence of PI3K inhibitors (Horwood et al., 2006; Kelly and Lynch, 2000; Sui et al., 2008). At hippocampal Schaffer collateral-CA1 synapses, Sanan et al. (2002) reported that the expression (but not the induction) of LTP induced by HFS was blocked by a PI3K inhibitor. However, Opazo et al. (2003) reported opposing findings in which PI3K inhibitors partially blocked LTP induction without affecting its expression. The contradiction might be due to differences in species, drug concentration or stimulation protocol. Opazo et al. (2003) also found that LTP induced by theta pulse stimulation (TPS) was suppressed by PI3K inhibitors and ERK inhibitors, whereas LTP induced by pairing presynaptic stimulation with postsynaptic depolarization only required PI3K activity, suggesting that PI3K and ERK signaling differentially contribute to LTP induced by different patterns of stimulation. Furthermore, PI3K activity is required for the persistence of late-phase LTP induced by multiple trains of HFS or TBS (Karpova et al., 2006; Raymond et al., 2002). Finally, PI3K-Akt-mTOR signaling was found to be engaged in metabotropic glutamate receptor (mGluR)-dependent LTD (Banko et al., 2006; Hou and Klann, 2004).

In addition to mammalian systems, PI3K also plays important roles in synaptic plasticity in invertebrates. Its activity is required for long-term synaptic facilitation in crayfish and Aplysia (Beaumont et al., 2001; Hu et al., 2006). Interestingly, a recent study in Drosophila found that Aβ42, which forms plaques in Alzheimer’s disease, induced PI3K hyperactivity, whereas Aβ42-induced memory loss was rescued by PI3K inhibition (Chiang et al., 2010). This study suggests that, similar to ERK, hyperactive PI3K can also lead to memory impairment.

Rin1

Rin1 is a GEF for Rab5, a small G-protein of the Rab family involved in endocytosis of surface receptors, including AMPARs (Brown et al., 2005; Tall et al., 2001). Active Ras can bind to and enhance the activity of Rin1 (Han and Colicelli, 1995) (Figure 2C). Rin1 also competes with c-Raf for Ras binding (Han and Colicelli, 1995). Thus, it is thought to antagonize the effects of the Ras-Raf-MEK-ERK cascade in memory processing. Rin1 is highly expressed in the brain, relatively enriched in hippocampus, amygdala, striatum and forebrain cortex, and is localized in the soma and dendrites of neurons (Dhaka et al., 2003). Dhaka et al. (2003) found that Rin1 knockout mice (Rin1^−/−) exhibited enhanced short-term and long-term memory for cued fear conditioning and conditioned taste aversion. Since the amygdala is essential for both tasks, they also examined amygdala LTP in these mice and found that the magnitude of LTP was increased. Knockout of Rin1 did not affect hippocampal LTP or hippocampus-dependent spatial learning and memory, suggesting a brain region (amygdala)-specific role of Rin1 in memory processing. A recent study examined fear extinction and latent inhibition in Rin1^−/− mice (Bliss et al., 2010). In this study, control and knockout mice were extensively trained to show comparable strong levels of memory for cued fear conditioning, and afterwards, extinction trials were delivered. 24hrs after extinction training, wild-type controls showed a significant reduction in conditioned fear responses. However, conditioned responses in Rin1^−/− mice remained the same as their pre-extinction level. Furthermore, latent inhibition, in which in wild-type animals pre-exposure of a cue impairs subsequent associative learning, was impaired in Rin1^−/− mice. Depotentiation has been proposed as mechanism for fear extinction (Kim et al., 2007). However, Rin1^−/− mice exhibited enhanced depotentiation in amygdala LA-BLA synapses compared to wild-type mice, arguing against depotentiation as the mechanism mediating extinction. It remains to be examined whether plasticity in other amygdala circuits might account for the behavioral impairments.

In summary, the studies examining the effectors of Ras family proteins further support the model that signaling of Ras family proteins plays essential roles in multiple forms of memory and synaptic plasticity in a variety of brain regions. However, the engagement of each downstream effector of these G-proteins in the induction of plasticity appears to depend on the brain regions involved, as well as the type of memory task and pattern of stimulation utilized. It should also be noted that although Ras family proteins play significant roles in activating these effectors, the effectors can also be activated independent of Ras family proteins (Rodriguez-Viciana et al., 1996a; Vanhaesebroeck et al., 2010; Wellbrock et al., 2004). Thus, further evaluation is required to determine whether the specific changes in both memory and synaptic plasticity caused by blocking these downstream effectors are uniquely due to disruption of the signaling from Ras family proteins.

AMPARs

Changes in AMPAR trafficking are important components of synaptic plasticity, and have been implicated in multiple forms of adaptive behavior, including learning and memory (Kessels and Malinow, 2009). It is commonly thought that insertion of AMPARs into the postsynaptic membrane enhances synaptic strength, which underlies the expression of LTP. Conversely, the removal of AMPARs from synapses weakens synaptic strength, leading to LTD or depotentiation.

To understand the role of Ras in AMPAR dynamics, Zhu et al. (2002) transfected hippocampal CA1 pyramidal neurons with wt-, ca- or dominant negative (dn)-Ras (H-Ras sequence was used in this study) and examined AMPAR current by whole-cell patch recording at −60mV. Overexpression of wt- or ca-Ras enhanced whole-cell response mediated by AMPAR at the basal state, whereas dn-Ras reduced the responses. Furthermore, they found that dn-Ras suppressed AMPAR responses by inhibiting synaptic insertion of AMPARs containing GluR2L, an AMPAR subunit with a long cytoplasmic tail, and ca-Ras enhanced AMPAR responses by promoting insertion of GluR1-containing AMPARs. To probe the role of Ras in long-term plasticity, Zhu et al. induced LTP by pairing presynaptic stimulation with postsynaptic depolarization, and induced LTD by low frequency stimulation. They found that Ras activation and subsequent ERK activation was required for LTP in AMPAR current, but not for LTD, supporting the model that Ras promotes synaptic insertion of AMPARs during LTP. To further elucidate the intermediate steps between Ras and AMPARs, Qin et al. (2005) transfected hippocampal slices with different ca-Ras mutants, targeting specifically c-Raf, b-Raf/RalGEF or PI3K. They found that c-Raf was not required in AMPAR insertion, whereas b-Raf-MEK-ERK signaling was both necessary and sufficient for synaptic insertion of GluR2L-containing AMPARs during spontaneous activity. However, insertion of GluR1-containing AMPARs required a synergistic interaction between neuromodulatory signaling and neuronal activity to activate both b-Raf-MEK-ERK and PI3K-Akt cascades. In addition, Hu et al. (2008) reported that the impairment in GluR1-dependent LTP associated with Fragile X Syndrome could be rescued by overexpression of wt-Ras or by reducing the threshold of PI3K-Akt activation by blocking an inhibitor in the cascade.

In contrast to Ras, Zhu et al. (2002) found that overexpression of wt-Rap1 or ca-Rap1 resulted in a decrease in AMPAR response whereas dn-Rap1 expression enhanced AMPAR responses. The decrease in AMPAR responses by ca-Rap1 was mediated by p38MAPK-dependent removal of GluR2-containing AMPARs from postsynaptic membranes. Furthermore, dn-Rap1 blocked LTD in AMPAR current, which involved removal of GluR2-containing AMPARs, whereas LTP was not affected. In support of the idea that Rap1 activation promotes AMPAR removal, Cingolani et al. (2008) reported that Arginine-Glycine-Aspartic Acid peptides (RGD-peptides, which disrupt integrin signaling) induced Rap1 activation, and in parallel, a reduction of mEPSC amplitude through internalization of GluR2-containing AMPARs. dn-Rap1 expression reversed the RGD-induced reduction in mEPSC amplitude.

Similar to Rap1, Zhu et al. (2005) found that Rap2 activation also led to removal of AMPARs from synapses. However, unlike Rap1, Rap2 activation activated JNK, but not p38MAPK or ERK. Inhibition of JNK or an upstream activator of JNK blocked Rap2-mediated depression of AMPAR response. Inhibition of Rap2 or JNK activation also blocked AMPAR-mediated depotentiation. Furthermore, ca-Rap2 expression caused removal of GluR1- and GluR2L-containing AMPARs from synapses. Finally, blocking Rap2 activation resulted in enhanced basal GluR2L phosphorylation, and increased GluR1 phosphorylation induced by ca-Ras expression, suggesting that Rap2 removal of AMPARs involves activation of a protein phosphatase.

Collectively, these findings suggest that Ras, Rap1 and Rap2 control different steps of AMPAR trafficking by recruiting distinct signaling cascades. During spontaneous activity, the recycling of GluR2L-containing AMPA receptors is bi-directionally controlled by Ras-ERK and Rap2-JNK cascades, and the insertion of GluR2-containing AMPA receptors is suppressed by Rap1-p38MAPK cascade. In response to LTP-inducing stimuli, Ras activates both ERK and PI3K cascades to drive GluR1-containing AMPA receptors into synapses. In contrast, during LTD, Rap1 removes GluR2-containing AMPA receptors by activating p38MAPK. At already-potentiated synapses, low frequency stimulation can activate Rap2-JNK cascade to remove GluR1-containing AMPA receptors, resulting in depotentiation.

However, these findings contradict a number of reports that we have discussed in the previous section: (i) Manabe et al. (2000) found that H-Ras knockout mice exhibited larger LTP by removing the inhibition on NMDARs, but not affecting AMPARs. Here, Zhu et al. (2005) found dn-H-Ras expression blocked AMPAR insertion and LTP. A caveat of the study by Zhu et al. (2005) is that dn-H-Ras was expressed with C-terminal attachment of green fluorescent protein (GFP), which may affect the specific subcellular localization of dn-H-Ras (Omerovic and Prior, 2009; Zhang and Casey, 1996). Thus, the blockade of AMPAR insertion may not be caused by inhibiting H-Ras, but rather by disrupting the signaling from other Ras isoforms, such as K-Ras, which has been found to be critical for LTP (Ohno et al., 2001). (ii) Morozov et al. (2003) found that dn-Rap1 expression reduced LTP induced by TBS, whereas Zhu et al. (2005) reported that dn-Rap1 expression did not affect LTP induced by a pairing protocol. A reason for the contradictory results may reside in the different LTP-induction protocols that are used in these studies. In fact, in the study by Morozov et al. (2003), Rap1 was recruited in LTP induction when cAMP signaling was triggered. They found that LTP induced by multiple HFS, which is cAMP-independent, was not blocked by dn-Rap1. (iii) Ryu et al. (2008) found that ca-Rap2 expression enhanced LTD induction and decreased ERK activity, without affecting depotentiation and JNK, which are opposite to the findings by Zhu et al. (2005). One possible explanation is that in the study by Ryu et al. (2008), Rap2 activity was chronically elevated, which may induce compensatory mechanisms that change the rules for the induction of synaptic plasticity.

NMDARs

NMDARs are dually regulated channels which allow influx of Ca²⁺ in response to a combination of glutamate binding and membrane depolarization. Ca²⁺ influx from NMDARs plays an essential role in the induction of a variety of forms of LTP and LTD. Although changes in the activity of Ras family proteins do not affect NMDARs during spontaneous activity (Zhu et al., 2002), several studies suggest that H-Ras signaling can regulate the function of NMDARs during the induction of long-lasting synaptic plasticity.

Thronton et al. (2003) treated brain slices with Tat-H-Ras (Tat is a cell-permeable peptide) and found a decrease in the tyrosine phosphorylation and membrane retention of the NR2A subunit of NMDARs. They also detected a decrease in the activity of a cytosolic tyrosine kinase, Src. The effect of Tat-H-Ras on NMDARs was blocked in slices from Src knockout mice. Furthermore, consistent with the notion that NR2A-containing NMDARs are critical for LTP induction, Tat-H-Ras significantly reduced hippocampal Schaffer-collateral CA1 LTP. In another study, Suvarna et al. (2005) found that internalization of NR2A-containing NMDARs induced by acute ethanol exposure was blocked in the presence of Tat-dn-H-Ras. These findings are consistent with the earlier report by Manabe et al. (2000), in which H-Ras knockout resulted in larger LTP, collectively suggesting that H-Ras activation can reduce surface retention of NR2A-containing NMDARs by inhibition of Src.

Ral

Ral is a small G-protein present at high levels at synaptic vesicle membranes. Both active Ras and Rap1 can activate Ral-GEFs by recruiting them to their target Ral in the membrane as well as by enhancing the catalytic activity of Ral-GEFs via PI3K (Feig, 2003) (Figure 2C). A number of studies suggest that Ral plays an important role in regulating neurotransmitter release. Polzin et al. (2002) reported that Sec6/8, which targets secretory vesicles to release sites in plasma membranes, bound specifically to active Ral. Furthermore, in synaptosomes prepared from mouse cortical neurons, dn-RalA expression blocked the enhanced glutamate release induced by a PKC activator, without affecting basal release. It also reduced the refilling of the readily releasable pool of synaptic vesicles after Ca²⁺-triggered depletion. In another study, Owe-Larsson et al. (2005) blocked Ral signaling in cultured hippocampal neurons by either pharmacological inhibitors or expression of dn-RalA, and observed a significant reduction in neurotransmitter release probability. Thus, Ras/Rap1 may regulate neurotransmitter release by activating Ral. However, it has not been elucidated whether Ral activity is required for memory formation. In addition, since Ras-independent mechanisms of Ral-GEF activation exist, it is not clear whether Ral-mediated regulation of neurotransmitter release is uniquely under the control of Ras/Rap1.

Synapsin I

Synapsin I is a protein that associated with the surface of synaptic vesicles and is critical for maintaining a reserve pool of synaptic vesicles as well as regulating the rate of vesicle release (Hilfiker et al., 1999). The C-terminus of Synapsin I contains consensus sites for phosphorylation by ERK, which is a major downstream effector of Ras family proteins. In synaptosomes prepared from rat or mouse cortex, the enhancement of neurotransmitter release induced by BDNF requires ERK phosphorylation of Synapsin I (Jovanovic et al., 2000). Another independent study found that AMPA stimulated ERK activation and subsequent Synapsin I phosphorylation in synaptosomes that occurred in parallel with a significant increase in synaptic vesicle recycling rate (Schenk et al., 2005). Recently, Vara et al. (2009) examined the role of the ERK-Synapsin I cascade in synaptic plasticity at hippocampal mossy fiber-CA3 synapses. They found that post-tetanic potentiation was enhanced by ERK inhibition. The enhancement was lost in mice lacking Synapsin I, suggesting an inhibitory role of ERK-Synapsin I signaling in this form of plasticity. Furthermore, in Aplysia, ERK has been reported to mediate the effect of transforming growth factor β1 (TGF-β1) on reducing synaptic depression through phosphorylation and re-distribution of Synapsin I (Chin et al., 2002). Collectively, these studies suggest that Ras family proteins may contribute to presynaptic plasticity through regulation of the ERK-Synapsin I cascade.