Fragile X mental retardation protein and synaptic plasticity

FMRP is an RNA-binding protein and a repressor of translation which is well-conserved from mouse to human. FMRP associates with mRNAs through one of three RNA-binding domains [20, 21], in some cases in conjunction with adaptor proteins [22, 23]. There is evidence that FMRP can repress translation both by blocking initiation and elongation [15, 24, 25]. A point mutation in one FMRP/mRNA binding domain is sufficient to recapitulate plasticity phenotypes seen in the Fmr1 KO mouse [26] and in at least one case FX in a human patient [27]. Thus it is likely that FMRP regulates plasticity mainly in its role as a repressor of translation.

FMRP is regulated by posttranslational modifications. Phosphorylated FMRP stalls ribosomal translocation and inhibits translation, whereas dephosphorylation of FMRP upregulates translation [28‐30]. Bidirectional regulation of FMRP phosphorylation by the S6 kinase and protein phosphatase 2A (PP2A) in response to activity provide a potential link between synaptic stimulation and local translation [24].

FMRP is well-positioned to be a key regulator of synaptic plasticity for three main reasons. First, the protein is found in dendritic spines [31‐34], important postsynaptic sites of plasticity induction and maintenance. Secondly, FMRP regulates dendritic mRNA translation [16, 17], which is required for multiple forms of plasticity [35]. Finally, FMRP itself is dynamically regulated by activity: experience and synaptic activation can trigger its local translation and rapid degradation, in addition to the posttranslational regulation mentioned above. Multiple experimental manipulations associated with synaptic plasticity have been shown to increase FMRP levels, including exposure to an enriched environment, a complex learning task, and pharmacological activation of group 1 metabotropic glutatmate receptors (mGluRs) [31, 36‐38]. Importantly, FMRP is synthesized rapidly, on the same time scale (10–30 minutes) as induction of stable synaptic plasticity [37]. In hippocampal cultures, activity- and mGluR-dependent increases in dendritic FMRP may result from increased trafficking of existing FMRP, rather than de novo FMRP synthesis [33, 39, 40]. Either way, FMRP is an ideal candidate to be involved in regulating synaptic plasticity because of its rapid, transient rise in dendrites following well-characterized plasticity induction paradigms, as well as its role as an inhibitor of translation.

Long-term potentiation (LTP) and long-term depression (LTD) are well-characterized forms of synaptic plasticity associated with learning and memory. These persistent changes in synaptic strength can be induced by a variety of manipulations and their expression mechanisms are diverse. Different induction protocols rely on different mechanisms for maintenance, including the requirement for protein synthesis. A particularly compelling example of a form of plasticity requiring local translation is metabotropic glutamate receptor-dependent LTD (mGluR-LTD) in the CA1 region of the hippocampus. Activation of group 1 mGluRs (mGluR1 and 5), either with paired-pulse low-frequency synaptic stimulation (PP-LFS) [4] or with the selective agonist (S)-3,5-dihydroxyphenylglycine (DHPG) [41‐43], results in a persistent decrease in synaptic strength that is mechanistically distinct from classical NMDA receptor (NMDAR)-dependent LTD [41, 44]. It is important to note that there are several mechanisms downstream of mGluR activation that can depress synaptic transmission, and these can be differentially expressed depending on the induction protocol, age, rearing history, and species (e.g., [44‐48]). However, under appropriate experimental conditions the maintenance of mGluR-LTD requires rapid protein synthesis within minutes of induction [4, 49]. This protein synthesis is likely to be synaptic, as mGluR-LTD can still be induced if the dendritic layer is physically severed from the cell body layer [4]. mGluR-LTD is expressed, in part, by the removal of AMPA receptors from synapses, which also requires rapid de novo translation [50]. The new protein synthesis may be instructive rather than merely permissive for synaptic plasticity since activation of group 1 mGluRs rapidly stimulates protein synthesis in hippocampal slices [12], dendrites and synaptoneurosomes [51, 52].

Fmr1 knockout mice show enhanced hippocampal mGluR-LTD [8, 14, 49, 53] (Table 1). A subsequent study found a similar enhancement in cerebellar mGluR-LTD, which shares many of the same expression mechanisms [54]. Consistent with the electophysiological data, loss of FMRP leads to excessive mGluR-mediated AMPAR internalization [55]. In addition, mGluR-LTD no longer requires new protein synthesis in the Fmr1 KO mice [49, 56]. These results, combined with what is known about FMRP function, suggest that FMRP acts to inhibit the synthesis of proteins required for mGluR-LTD. In the absence of FMRP, these “LTD proteins” are already available or over-expressed in dendrites resulting in enhanced magnitude and protein synthesis-independent persistence of this form of plasticity (Figure 1A) [13]. Conversely, postnatal overexpression of FMRP reduces the magnitude of mGluR-LTD in both wildtype and Fmr1 KO neurons [49] and restores its protein synthesis dependence [57]. Moreover, reducing mGluR5 signaling in Fmr1 KO mice restores both protein synthesis rates and LTD magnitude in the hippocampus to wildtype levels [53, 58], suggesting that mGluR5 and FMRP act in functional opposition to maintain an optimal level of synaptic protein synthesis throughout development and into adulthood (Figure 1A).

While the effects of protein synthesis inhibition on mGluR-LTD can be seen within minutes, most forms of synaptic plasticity do not require de novo synthesis until several hours after induction. This is best characterized by late phase LTP (L-LTP), a persistent form of potentiation lasting at least 4 hours. The late maintenance phase of L-LTP requires protein synthesis but initial induction does not [59, 60]. Due to FMRP’s conjectured role in translation regulation, L-LTP was one of the first forms of plasticity studied in the Fmr1 KO mouse [61]. Interestingly, no difference has been found in the magnitude of L-LTP in the Fmr1 KO [61, 62]. The fact that removal of FMRP affects protein synthesis-dependent LTD but not LTP suggests that FMRP may specifically regulate the translation of proteins required for the expression of LTD (Figure 1B). However, while the magnitude of L-LTP is unchanged, it is possible that L-LTP is qualitatively different in its requirement for new protein synthesis when FMRP is absent, as is the case for mGluR-LTD (and LTP priming, see below). Therefore, it will be important to test the protein synthesis-dependency of L-LTP in Fmr1 KO mice to show that FMRP truly does not play a role in regulating the persistence of LTP.

Alternatively, FMRP may be required for the regulation of local but not somatic translation in the context of L-LTP (Figure 1C). L-LTP is traditionally induced by multiple trains of high frequency tetanus or theta burst stimulation, protocols that rely on cell-wide transcription and translation [63‐65]. L-LTP was characterized in the Fmr1 KO mouse using these classical paradigms [61, 62]. However, using a less intense induction protocol results in L-LTP that is maintained specifically by local dendritic translation [5, 66]. This form of L-LTP, similar to mGluR-LTD, is sensitive to inhibitors of translation but not transcription, and can be maintained in isolated dendrites. It will be interesting to determine if this locally expressed form of L-LTP is regulated by FMRP.

While the role of FMRP in L-LTP is unclear, FMRP is known to be involved in LTP in other contexts. In particular, FMRP is involved in regulation of an mGluR-dependent form of metaplasticity that sets the threshold for LTP. Originally described in rats [67], weak activation of group 1 mGluRs, in itself insufficient for LTD induction, facilitates the subsequent induction of LTP (“LTP priming”). As with mGluR-LTD, this facilitation requires translation but not transcription [68]. This prompted the examination of the role of FMRP in LTP priming [69]. mGluR-dependent priming of LTP is of comparable magnitude in WT and Fmr1 KO mice; however, while LTP priming requires acute stimulation of protein synthesis in WT mice, it is no longer protein synthesis-dependent in the Fmr1 KO. Thus, while mGluR-LTD and LTP priming are qualitatively different functional consequences of Gp1 mGluR-stimulated protein synthesis in the hippocampus, both processes are altered by the removal of FMRP (Figure 1D). These results suggest that the mRNA under translational control of FMRP may code for proteins required for bidirectional changes in synaptic strength. Thus, the proteins regulated by FMRP should be conceptualized as plasticity gatekeepers rather than solely “LTD proteins.”

In Fmr1 KO hippocampal slices, LTP induction is deficient with a weak 5 theta burst protocol but is normal with a strong 10 theta burst protocol (Figure 2A) [70]. In addition, FMRP modulates the induction threshold for spike-timing dependent long-term potentiation (STD-LTP). This form of Hebbian plasticity is induced by temporally staggered presynaptic and postsynaptic activity within a very short window [71, 72]. In somatosensory and prefrontal cortices, STD-LTP is deficient in Fmr1 KO neurons [73, 74]. However, if the postsynaptic stimulus strength is increased from a single spike to a burst of five spikes, STD-LTP does occur in KO neurons (Figure 2A) [74]. Therefore FMRP is not required for expression of STD-LTP, but the threshold is raised in its absence. A possible mechanism for ongoing regulation of LTP thresholds by FMRP is discussed later in this review.

In addition to its role in translation-dependent forms of Hebbian plasticity, FMRP can also modulate some forms of homeostatic plasticity. Synaptic scaling is a form of homeostatic plasticity that acts to keep the strength of synapses within a functional range in response to extreme changes in activity. Broadly, a decrease in activity leads to a subsequent cell-wide increase in synaptic strength (“scaling up”) and an increase in activity leads to a decrement in synaptic strength (“scaling down”) [75]. Two types of scaling up have been described in hippocampal slice culture: one that requires transcription [76] and one that requires local translation [77]. Interestingly, only the translation-dependent form of synaptic scaling is deficient in neurons lacking FMRP. Postsynaptic viral expression of FMRP corrects deficient translation-dependent scaling up in Fmr1 KO neurons [78]. Scaling down of synapses in response to high levels of activity (following prolonged blockade of inhibition) has also been observed [79] and requires mGluR5 activation [80, 81]. However, the role of FMRP and local protein synthesis in scaling down has not been directly examined.

While the role of FMRP has been best characterized in mGluR-dependent forms of plasticity, it is not specific to these receptors. Removal of FMRP occludes TrkB-mediated increases in protein synthesis [12] and alters other forms of G protein-coupled receptor (GPCR)-dependent LTD and LTP [82, 83]. The common thread between these processes is their reliance on local dendritic translation. Indeed, evidence suggests that FMRP may specifically be important for the regulation of local rather than somatic translation (Figure 1C), as removal of FMRP affects translation but not transcription-dependent forms of Hebbian and homeostatic plasticity.

While many forms of translation-dependent synaptic plasticity are abnormal in Fmr1 KO mice, other forms of hippocampal plasticity, including NMDAR-dependent LTD and early-phase LTP, are normal [8, 61, 69, 84, 85]. These observations suggest that FMRP regulates plasticity mainly in its role as a regulator of translation. However, removal of FMRP has also been shown to affect some forms of synaptic plasticity that do not require de novo translation, such as early-phase LTP in other brain areas, including the cortex and amygdala [61, 85‐89]. Some of these effects could be explained by FMRP modulation of protein synthesis-dependent plasticity thresholds; however it seems likely that many represent end-stage consequences of altered synaptic development in the Fmr1 KO.

A case in point is altered LTP in the amygdala. A substantial deficit in basal transmission was reported at the same synapses that showed impaired LTP [88]. Reduced synaptic connectivity might have caused the defective LTP, and might have arisen as a consequence of increased FMRP-dependent protein synthesis during the development of amygdala circuitry.

In order to determine how FMRP regulates synaptic plasticity, we must identify the synaptic proteins whose translation is regulated by FMRP. FMRP has a wide variety of targets - it has been shown to selectively bind approximately 4% of the mRNA in the mammalian brain [90]. Recently, over 800 mRNA binding targets of FMRP were identified using a novel high throughput cross-linking immunoprecipitation (HITS-CLIP) assay [10]. These targets include genes coding for pre- and post-synaptically expressed proteins: 27% of pre-synaptic protein mRNAs (90 genes) and 23% of postsynaptic protein mRNAs (257 genes) are FMRP targets [10]. More specifically, the HITS-CLIP study found that 31% of mRNAs coding for proteins in the NMDAR complex (58 genes), 62% in the mGluR5 complex (32 genes), and 33% in the AMPAR complex (3 genes) are FMRP targets. These three receptor complexes are important for the induction and maintenance of synaptic plasticity, suggesting that FMRP likely acts broadly as a translational regulator rather than solely regulating one or two “plasticity proteins.”

The finding that many FMRP targets encode presynaptic proteins is interesting and illuminating. In the mature nervous system the evidence for local protein synthesis in axons or axon terminals is still sparse; however during early axon development and synapse formation local protein synthesis is believed to play an important role in pathway and target selection [91, 92]. Thus, the absence of FMRP regulation of protein synthesis during early development very likely alters synaptic connectivity well before the onset of experience-dependent postnatal plasticity. In addition, outside the CNS, local control of translation in sensory afferent terminals plays a role in nociceptive sensitization and neuropathic pain [93]. FMRP is localized to these terminals and Fmr1 KO mice show altered nociceptive sensitization [94]. These results suggest that in the spinal cord, presynaptic FMRP may inhibit local translation and can regulate pain plasticity even into adulthood.

Long-lasting synaptic potentiation and depression have long been considered potential neural correlates of learning and memory. In conjunction with FMRP’s role in synaptic plasticity in multiple brain areas, FMRP is also important for a wide range of behavioral learning tasks in mice. Fmr1 KO mice show deficient amygdalar trace fear memory [87], cerebellar learning [54], inhibitory avoidance learning [58], and have difficulties with a prefrontal cognitive learning task [128]. Drosophila mutants lacking FMRP also have impaired long-term memory [129]. Overall, learning and memory deficits in the Fmr1 KO mouse are a likely behavioral consequence of abnormal synaptic plasticity.