Alterations in synaptic functions are reflected also in morphological differences between the two disorders. Reduced neuronal cell size and numbers of dendrites have been observed in patients with RTT alongside short dendrites and a decrease in dendritic spines (Armstrong, 2005; Belichenko et al., 2008). In RTT, there is an overall decrease in brain volume differences and altered brain structures that are also present in mice studies (Chen et al., 2001; Ballas et al., 2009; Xu et al., 2014).

In contrast to RTT and MeCP2 deficiency, a lack of FRMP is associated with increases spine density in human patients and mice models (Levenga et al., 2011; Hodges et al., 2017). Both in Fmr1 KO mice, and in brain from patients, it has been reported an increase in the number of elongated spines (filopodia) in absence of functional FMRP (Comery et al., 1997), suggesting a reduction in the pruning of spines (Irwin et al., 2000). However, recent studies on spine functionality (Thomazeau et al., 2020) and spine compartmentalization over time (Wijetunge et al., 2014), find that the altered morphology does not correlate with abnormal function between Fmr1 KO mice and controls. These findings challenge the view of immature status of connections in FXS (He and Portera-Cailliau, 2013), and even the turnover of the spines seems to be invariant to that of matched controls. However, all these measures can be dependent on the brain area and on the stage of development (He and Portera-Cailliau, 2013).

In fact, the presence of the filopodia is dependent on the stage of development and on the preparation. Nimchinsky et al. (2001) analyzed mutant and control mice at several developmental stages and showed that differences in spine morphology and density decreased between 1 and 4 weeks of age. The closing of these phenotypic gaps originally suggested that FMRP played a role in coordinating synaptogenesis in a time-dependant manner. Different cell populations of the hippocampus also show that long and irregular spines are also present in juvenile and adult Fmr1 KO mice, while in vitro and in vivo studies show discrepancies for spine densities as with LTP and LTD function (as reviewed by Bostrom et al., 2016). A summary of the studies investigating morphological correlates of synaptic functions in RTT and FXS is reported in Table 2.

Both RTT and FXS display endophenotypes that signal aberrant Long-term potentiation (LTP) and long-term depression (LTD), which are Hebbian Forms of plasticity.

In RTT, deficits have been reported in excitatory synapses and LTP. Hippocampal slices show reduced LTP in Mecp2 KO (Asaka et al., 2006) and Mecp2^308/Y mice (Moretti et al., 2006). Reduced potentiation is present around onset of symptoms while pre-symptomatic mice maintain the same level of activity observed in matching controls, suggesting that synaptic dysfunction and decline in plasticity is an early event in RTT. Pyramidal neuron synapses in the hippocampus are indeed potentiated in Mecp2 KO mice, however there is a failure to regulate AMPA receptors post-activation and an overall deficit in LTP (Li et al., 2016). Mecp2 KO mice display a lack of structural plasticity, and enlarged spines are observed regardless of stimulation or sham (Li et al., 2016). AMPA receptor-related transmission is enhanced at hippocampal synapses and over time, the decrease in internalization fails to counterbalance excessive accumulation. In addition, the lack of receptor trafficking prevents activated synapses from becoming plastic, affecting both LTP and LTD.

Long-term depression is also altered in RTT, however to a lesser extent than LTP. Moretti et al. (2006) induced LTD with two different stimulus paradygms in the hippocampus of Mecp2^308/Y mice and show that when LTD was induced by administration of the mGluR agonist 3,5-dihydroxyphenylglycine (DHPG), the response was comparable in mutant and wildtype mice. However, when paired pulse stimulation was applied, LTD was observed in wild type mice, but not in Mecp2^308/Y mice. Since both stimulation paradigms affect the post-synaptic sites, but only the paired pulse stimulation protocol is dependent on the presynaptic site, the authors conclude that the impairments in plasticity are due to altered presynaptic terminals, while there is some preservation of LTD. It has also been shown that hippocampal slices of symptomatic mice with an Mecp2 KO show no NMDAR-dependent LTD by low frequency stimulation (Asaka et al., 2006). More recently, in younger P15 Mecp2 KO mice, LTD has been induced by DHPG at two different stages: early and late (Lozovaya et al., 2019). Only early induced LTD displays a significant decrease in the amplitude of the response, implicating even earlier synaptic impairment onset in RTT mice.

In FXS LTP was reported to be reduced in several brain regions (Desai et al., 2006; Lauterborn et al., 2007; Suvrathan and Chattarji, 2011; Seese et al., 2012), however, the main form of plasticity studied in FXS is LTD.

The first to show alterations in synaptic plasticity in the absence of FMRP were Huber and colleagues (Huber et al., 2002), who showed an increase in mGluR-dependent LTD in the hippocampus of Fmr1 KO mice. The dysregulation of mGluR signaling was further confirmed by other authors (Weiler et al., 1997; Gross et al., 2012; Tian et al., 2017), and all the data support the theory that FMRP controls the translation of specific proteins involved in synaptic function, including glutamatergic receptors.

Similarly to what reported in RTT, AMPA aberrant receptors trafficking is driven by a lack of FMRP and consequently associated with cognitive deficits (Nakamoto et al., 2007). PSD-95 synthesis increases in response to DHPG-activation of mGluR, and is co-translated with FMRP (Todd et al., 2003). Further studies in other brain areas showed that LTD in enhanced in multiple cell populations, including cerebellum, where it controls abnormal motor behavior and development of synaptic circuitry in the somatosensory cortex (Greenough et al., 2001; Huber et al., 2002; Koekkoek et al., 2005).

Overall, LTP deficits have been shown in RTT and to a lesser extent in FXS. On the other hand, FXS tends to display enhanced LTD, while RTT shows some LTD preservation.

A summary of the experiments exploring Hebbian forms of plasticity in RTT and FXS is reported in Table 3. Beside LTP and LTD, other forms of plasticity are affected in RTT and FXS.

Homeostatic plasticity controls the changes in synaptic strength that individual neurons operate in response to prolonged changes of neuronal stimulation, and it also mediates the balance of excitation and inhibition. Both RTT and FXS have shown alterations in mechanisms of homeostatic plasticity, and these alterations may explain the general changes in neuronal activity in patients and animal models with RTT and FXS (Table 4).

One of the main mechanisms controlling homeostatic changes is synaptic scaling, and it relies on the synthesis, trafficking, and function of AMPA receptors.

Defects in functional levels of Mecp2 alter homeostatic plasticity. Hence, in RTT, where there is a loss of function of MeCP2, there is impaired ability to respond to homeostatic changes (Na et al., 2013). One of the first studies of homeostatic plasticity in RTT was performed in neuronal cultures. In this study, stimulation with bicuculline produced an increase in neuronal activity, with a sequential decrease in the expression of the AMPA receptors subunit GluR2 and a reduction of mEPSC amplitude. However, the expression of GluR2 is under the control of the transcriptional repressor MeCP2, whose levels also increase during bicuculline stimulation. As a result, the bicuculline treatment in Mecp2 KO neurons, does not lead to the expected decrease of the GluR2 receptor subunit at the synapse. These experiments were among the first to show the importance of Mecp2 in the control of synaptic scaling (Qiu et al., 2012), and they were confirmed by Blackman et al. (2012), who showed that reduced neuronal drive in Mecp2 KO preparations does not produce the expected increase in synaptic scaling as observed in WT controls. One of the factors controlling homeostatic plasticity in RTT is the Early Endosome Antigen 1 (EEA1), which regulates AMPA receptor endocytosis. EEA1 expression is reduced in Mecp2 KO mice, where the synaptic scaling is reduced. However, the increased expression of EEA1 in Mecp2 KO cultures, reinstates the ability to scale the synapses in response to changes in activity (Xu and Pozzo-Miller, 2017). It is important to remember that not only Mecp2 expression, but also its activation, is important for the proper functioning of the protein. Indeed, post translational modifications of MeCP2 have also been found to be implicated in synaptic function and homeostasis (Bellini et al., 2014) and the phosphorylated MeCP2 modulates synaptic scaling -down through mGluR5 (Zhong et al., 2012).

Fragile X mental retardation protein has been involved in synaptic scaling in response to both increased and decreased activity- as observed in Fmr1 KO mice and in neuronal cultures derived from patients with FXS. The increased activity of AMPA receptors consequent to a decrease in activity, is mediated by the Retinoic Acid (RA), which is produced in response to the altered activity, and promotes the synthesis of new AMPA receptors. However, in FXS, while the transcription of the AMPA receptor remains unchanged, the RA- mediated translation of AMPA is reduced, and only restoration of the proper full length functional FMRP re-establishes synaptic scaling in primary cultures derived from Fmr1KO mice (Soden and Chen, 2010). These results suggest that FMRP is essential for the postsynaptic response in RA-mediated synaptic scaling. However, RA is required for the combined action of TTX and NMDA blockade- as there are no changes in scaling with TTX alone. In this regard it is interesting to note that the RA does not affect spine morphology and number. Zhang and colleagues confirmed the impaired synaptic scaling in neuronal cultures derived from patients with FXS (Zhang et al., 2018) and the role of RA as mediator of synaptic scaling. The RA action is effective both on excitatory and inhibitory synapses. Interestingly, RA is one of the mediators of the post- to presynaptic communication and could be involved in linking the post-synaptic events to pre-synaptic adjustments. In FXS, the imbalance in protein synthesis determines a dysregulation in RA which is important at postsynaptic level, but also at the presynaptic level with the regulation of EPSC frequencies (Wang et al., 2011; McCarthy et al., 2012).

Synaptic downscaling – the scaling down in response to increased synaptic activity- is also altered in Fmr1 KO mice. One of the mechanisms of downscaling is the degradation of AMPA receptor through ubiquitination (Lee et al., 2018). The ubiquitination of AMPA receptors is mediated by a complex cascade of molecules, which includes cell-cycle molecules and phosphatases, and FMRP interferes with the mechanism that leads to homeostatic-dependent ubiquitination.

It is worth of note that the Neuroligin-Neurexin complexes, involved in circuitry development and function is also essential for presynaptic homeostatic plasticity (Sons et al., 2006). Several molecular studies in RTT and FXS report that these molecules are targets of FMRP and MeCP2 (Darnell et al., 2011; Gulmez Karaca et al., 2018; Maurin et al., 2018; Raman et al., 2018), suggesting that impaired homeostatic plasticity in these syndromes may be mediated by deficits in Neuroligin or Neurexin.

Another factor influencing the intracellular pathways controlling synaptic strength is IGF1, which is a protein involved in growth, maturation, and neuronal development. Although CSF levels of Insulin-Like Growth Factor 1 (IGF1) have been found to be unchanged in RTT (Riikonen, 2003), both the full IGF1 molecule, and its functionally active cleavage product (1-3)IGF1 have been shown to ameliorate symptoms of the RTT pathophysiology (Tropea et al., 2009; Castro et al., 2014; Khwaja et al., 2014; O’Leary et al., 2018). IGF1 and (1-3)IGF1 administration in a RTT mouse model increases PSD95, dendritic arborization and excitatory current. Interestingly, these treatments also appear to re-establish cortical plasticity in Mecp2 KO mice to levels observed in controls (Tropea et al., 2009; Castro et al., 2014). IGF1 signaling occurs primarily through PI3K-AKT and MAPK pathways (Fernandez and Torres-Alemán, 2012). IGF1R activates PI3K-AKT functions to increase neuronal survival (Dudek et al., 1997), while prolonged administration of IGF1 with growth hormone, determines pro-inflammatory responses via activation of MAPK (Wolters et al., 2017). IGF signaling to a lesser extent has been implicated in FXS, where correcting the low level of IGF1R was found to reduce macro-orchidism (enlarged testes) a phenotype characterizing all the adult male patients with FXS (Wise, 2017). More promising evidence comes from the use of the (1-3)IGF1 analog, NNZ-2566 (Trofinetide) as a treatment for FXS; in fact mouse models of FXS treated with NNZ-2566 showed improvements in cognitive function and hyperactivity. Also patients with FXS treated with NNZ-2566 improved in a number of clinical scoring tools (Deacon et al., 2015; Berry-Kravis et al., 2020), but further clinical tests are required to confirm the benefits of the treatment.

Both BDNF and IGF1 signals involve the activation of intracellular pathways involving the PI3K and MAPK cascades, which are related to both RTT and FXS and are involved in activity-dependent plasticity.

Cyclic adenosine monophosphate (cAMP) response element binding protein is involved in transcriptional changes induced by synaptic plasticity, including increasing neuronal excitability and synapse strengthening associated to LTP induction (Caracciolo et al., 2018). Transcriptional genes expressed by CREB activation include c-Fos, whose protein is linked to memory and learning (Gallo et al., 2018), and the co-localization of CREB and c-Fos is associated with long term synaptic plasticity (Miyashita et al., 2018). This molecule has been found to be decreased in Mecp2 KO mice and rectifying CREB levels can correct some of the RTT symptoms (Bu et al., 2017). The Fmr1 gene is thought to be bound and therefore regulated by CREB (Kanellopoulos et al., 2012; Rani and Prasad, 2015), however the relationship of CREB signaling to FXS synaptic plasticity requires further investigation, especially considering the role that FMRP has in the modulation of cAMP and cGMP levels, two molecules upstream the CREB expression (Delhaye and Bardoni, 2021). It was shown that In the hippocampus, Cilostazol (an inhibitor of Phosphodiesterase 3) increases the levels of c-fos and of insulin-like growth factor 1 (IGF-1) (Zhao et al., 2010) and activates CREB in PC12 cells (Zheng and Quirion, 2006). This result suggests a link between the levels of cAMP and cGMP – both targets of PDE3 – and the levels of IGF-1. Remarkably, the inhibition of PDEs both in FXS (PDE2, PDE4, PDE4D) and in RETT (PDE4) has been shown to improve socio-cognitive deficits in animal models and in patients.

Phosphatidylinositol-3-kinases (PI3K) are a family of intracellular signaling molecules functioning downstream of G protein coupled receptors and tyrosine kinases.

Once activated, PI3K subsequently phosphorylates AKT which regulates cell cycle and apoptosis (Chalhoub and Baker, 2009). Further downstream of AKT phosphorylation, is the mammalian Target of Rapamycin (mTOR) whose activation regulates nutrition, energy sensing and growth (Zoncu et al., 2011). In Mecp2 KO mice the activation of the PI3K/AKT/mTOR pathway is reduced, while in FXS it is increased. The alterations in PI3K/AKT/mTOR level in Mecp2 KO mice can be restored using IGF1, and treatment with PI3K antagonists rescue FXS defects (Gross et al., 2010; Ricciardi et al., 2011; Yuan et al., 2020).

Alterations of FMRP levels affect the mGluR-PI3K-AKT-mTOR cascade. In normal conditions mGluR-dependent LTD requires rapid translation of dendritic mRNA, but in Fmr1 KO mice LTD it is enhanced and insensitive to inhibition of protein synthesis (Huber et al., 2002; Nosyreva and Huber, 2006). This insensitivity is due to the de-repression of FMRP, which causes increased basal levels of mGluR-stimulated checkpoint proteins (Thomazeau et al., 2020). Hence in FXS, mGluR-LTD is decoupled from protein synthesis/mTOR activation. Interestingly in Fmr1 KO mice, mTOR phosphorylation is increased in embryonic neocortex and in postnatal hippocampus samples (Sharma et al., 2010; Casingal et al., 2020).

The MAPK signaling pathway includes extracellular signal-Regulated Kinase 1 and 2 (ERK1 and ERK2), which are essential for neuronal transcriptional events, including synaptic plasticity, learning and memory (Thomas and Huganir, 2004). When Mecp2 KO mice are treated with recombinant human IGF1, both AKT, ERK1, and ERK2 levels are increased in conjunction with the increase of the post-synaptic marker PSD95 (Castro et al., 2014). The MAPK pathway is upregulated in Fmr1 KO mice, and is modulated by NNZ-2566 (Deacon et al., 2015), suggesting that mechanisms controlling plasticity may be potential targets of therapeutics in FXS and RTT, although the mechanisms of action of NNZ-2566 requires further investigation.

Also at the synaptic level, MeCP2 and FMRP control the expression and localization of PSD95, which is strongly related to synaptic strength in excitatory synapses. In the hippocampus, MeCP2 controls the number of glutamatergic synapses. VGLUT1 and PSD95, which are respectively pre- and postsynaptic markers, and they are both downregulated upon loss of MeCP2 in mutant mice. Conversely, a two-fold increase of MeCP2 expression, determines an increase in density and colocalization of these two markers (Chao et al., 2007). PSD95 levels are rescued by (1-3)IGF1 to levels comparable in wild type animals (Tropea et al., 2009) achieving the same phenotype as double mutants for loss and doubling of MeCP2 (Collins et al., 2004; Chao et al., 2007). PSD-95 expression is also deregulated in Fmr1 KO mice. FMRP binds to PSD95 mRNA in vivo (Todd et al., 2003; Zalfa et al., 2007), suggesting that FMRP stabilizes the PSD95 transcript, leading to adequate expression levels of PSD95 (Zalfa et al., 2007).

All energy-demanding activities, including synaptic function and plasticity, are dependent on bioenergetics, which appears to be dysfunctional in both RTT and FXS. Recent works in both FXS (Mithal and Chandel, 2020) and RTT (Neul et al., 2020) pointed at dysfunctions in mitochondria: the organelles devoted to produce the energy necessary for neuronal function. Alteration in bioenergetics has been proposed for RTT (Jin et al., 2015; Crivellari et al., 2021) and FXS (D’Antoni et al., 2019), and reflect the evolutionary need to match the cognitive function with the capacity of producing the adequate supply of energy to fulfill the requests. The link between energy metabolism and cognition has been well reviewed, and it appears clear that several brain disorders are now depending on the capacity of the organism to provide the fuel requested by the brain and to control the number of oxidative species, which are the natural side products of aerobic metabolism. Therefore, the ability to produce energy goes hand in hand with the capacity to control the reactive oxygen species and the possible damage created by an excess of these radicals. Considering the oxidative stress in FXS, it is interesting to underline that El Bekay et al. (2007) found that Fmr1 KO mouse brains have higher levels of reactive oxygen species, nicotinamide adenine dinucleotide phosphate (NADPH)-oxidase activation, lipid peroxidation and protein oxidation compared to wild type mice. In the cortex of Fmr1 KO it was also reported an increased level of metabolites that result from the attack of unsaturated lipids by the superoxide anion (Maurin et al., 2014). In normal conditions, the superoxide anion is detoxified by Sod1, the level of which is reduced in the absence of FMRP (Bechara et al., 2009), thus providing a source of oxidative stress (Maurin et al., 2014).

It is also worth mentioning that there is a two-way interaction between the systems controlling the production of energy in the cell, and the ion homeostasis (Castaldo et al., 2009), and that such interaction controls the onset and progression of neurodegeneration. These additional mechanisms should be taken into account for uncovering the underlying mechanisms in brain disorders and for designing to routes of treatment.