Beyond Widespread Mecp2 Deletions to Model Rett Syndrome: Conditional Spatio-Temporal Knockout, Single-Point Mutations and Transgenic Rescue Mice

Rett syndrome (RTT) is one of the leading causes of intellectual disabilities in women. In addition to a few autistic features, characteristic symptoms that distinguish from classical autism include stereotypic hand movements, motor coordination deficits, breathing abnormalities, seizures and loss of acquired speech as well as purposeful hand use. RTT is highly associated with MECP2, the gene encoding for the transcription factor that binds methylated Cytosine in C-p-G islands in DNA, controlling gene expression and chromatin remodeling. In this review, we will briefly discuss current perspectives on MeCP2 function, and then will describe in detail novel mouse models of RTT based on lossof-function of Mecp2 and their use for establishing rescue models, wherein we pay close attention to behavioral and morphological phenotypes. Widespread deletions Single point mutations Exon 1 Exon 2 Exon 3 Exon 4


Introduction
The autism spectrum disorder Rett syndrome (RTT; Online Mendelian Inheritance in Man #312750) is the leading cause of intellectual disabilities in girls with an incidence of 1:10,000 worldwide [1,2]. RTT cases occur sporadically with an extremely low familial rate [3,4]. Individuals with RTT develop typically until 6-18 months when a constellation of neurological hallmarks appears. Besides autistic features such as aberrant visual and aural contacts, characteristic symptoms that distinguish from classical autism include stereotypic hand movements, motor coordination deficits, breathing abnormalities, seizures and loss of acquired speech as well as purposeful hand use [5,6]. Underlying these traits are mutations of the gene MECP2 located in chromosome Xq28 that encodes MeCP2 (methyl-CpG-binding protein 2) in its two isoforms (e1 and e2) following alternative splicing of its four exons [7][8][9]. In this review, we will briefly discuss current perspectives on MeCP2 function, and then will describe in detail several mouse models of RTT based on loss-of-function of Mecp2 and their use for establishing rescue models ( Figure 1, Table 1), wherein we pay close attention to behavioural and morphological phenotypes. As to electrophysiological manifestations, we refer the readers to a recent review [10].

MeCP2 Function
MeCP2 is a transcription factor that binds methylated Cytosine in C-p-G islands in DNA, controlling gene expression and chromatin remodeling [11][12][13]. MeCP2 dysfunction as a result of pathogenic mutations (i.e. missense, nonsense, truncations due to premature STOP codons) accounts for most RTT cases. As an X chromosome-linked disorder, complete absence of functional MeCP2 causes early mortality in boys, who hardly survive over 2 years. Females live much longer owing to the presence of a considerable amount of functional MeCP2 produced by the allele in the non-inactivated X chromosome [14], albeit reduced by approximately 50% from typical levels (i.e. mosaicism of X chromosome inactivation).
Normal MeCP2 function requires two major domains, the methyl-CpG binding domain (MBD) and the transcriptional repression domain (TRD). MeCP2 was initially thought to primarily play an inhibitory role in gene expression by recruiting chromatin remodeling protein complexes and transcriptional co-repressors like mSin3a and Histone Deacetylases (HDACs) [15]. This initial view of MeCP2 function as a transcriptional repressor has been extended recently to include an activating function, as one of the significant findings from a microarray study demonstrated that the majority of affected genes are activated in MECP2 overexpressing mice and down-regulated in Mecp2 null mice [16]. However, only a small number of genes were selected in that study to confirm MeCP2 binding specificity, and the possibility that MeCP2 might indirectly promote gene expression via regulation of gene silencers such as REST/CoREST [17] or microRNAs [18,19] has not been formally excluded. Furthermore, the changes observed in the Chahrour et al. [16], microarray study and others [20,21] are usually small in magnitude, which suggests that even though directly bound to targeted gene, MeCP2 might not act as a classical transcriptional factor but rather as a fine-tuning gene regulator.
Research on RTT has been mostly focused on neuronal abnormalities, such as reduced cell body size, delayed dendritic maturation and synaptogenesis, reduced dendritic branching and axonal arborization, altered spine and presynaptic bouton density (see detailed discussion below for each mouse model). These findings  led to the assumption that RTT originates exclusively from neuronal dysfunction. However, this traditional view has been questioned by recent observations of impaired morphology in wildtype neurons when co-cultured with MeCP2-deficient glial cells [22]. Consistently, restoration of MeCP2 levels in glial cells in vivo significantly alleviated several RTT-like behaviors in Mecp2 null mice, as well as improving neuronal morphology [23]. Such non cell-autonomous effects on neurons are also displayed by MeCP2-deficient microglia, which release high levels of glutamate that are deleterious to co-cultured neurons [24]. Consistently, transplants of wildtype bone marrow to Mecp2 null mice arrested the progression of RTT-like features due to the generation of myeloid cells that differentiated into a microglial phenotype [25].
Another classical notion is that MeCP2 is indispensable for normal brain development, which infers that irreversible neurological dysfunction would occur in the absence of MeCP2 from an early age. However, recent studies found that re-activating the endogenous Mecp2 gene at normal levels in mature fully symptomatic Mecp2 null mice reverts several RTT-like phenotypes [26]. Such observations indicate that MeCP2's function goes beyond early brain development and may include additional roles in neuronal maintenance throughout the whole life [27]. More provocative findings on MeCP2 function continue to come out; however, none of those discoveries would take place without the use of current experimental mouse models, as discussed below, and those to come in the future.

Widespread Mecp2 Deletions
Mecp2 tm1.1B mice (a.k.a. Bird nulls) The Mecp2 tm1.1B mouse line was generated in the laboratory of Adrian Bird by crossing females with Mecp2 exons 3 and 4 flanked by loxP sites (Mecp2 2loxB ) with males ubiquitously expressing the Cre transgene to delete the loxP-flanked exons [28]. It should be noted that Mecp22 loxB mice already have ~50% lower levels of MeCP2 than wildtype controls (before Cre-mediated recombination), which cause significant neurological symptoms similar to Mecp2 null mice, albeit less severe; thus they are considered a hypomorph Mecp2 allele [29,30]. The progeny after Cre-mediated recombination and deletion of loxP flanked exons 3 and 4, however, completely lacks Mecp2 mRNA and protein expression [28,31]. Hemizygous null mice (i.e. Mecp2-/y males without an X chromosome) do not show any overt phenotypes until 4-6 weeks of age, but rapidly develop several neurological deficits and typically die at 7-8 weeks. Neurological deficits are usually homogenous, including inertia, ataxia, hindlimb clasping when suspended by the tail, irregular breathing, and reduced body weight. It should be noted that these features were observed in mice with a C57BL/6 genetic background. When these mice were crossed to the 129 strain, pronounced weight changes were not observed, but the remaining phenotypes remained unaffected. Heterozygous female mice (i.e. with a mosaic expression of wildtype and null alleles) develop similar RTT-like symptoms, but they have a conspicuously slower and milder disease progression.
Brain specific Mecp2 deletion was obtained by breeding Mecp2 2loxB mice with Nestin-Cre transgenic mice, which deleted Mecp2 exons 3 and 4 in the neuronal and glial cell lineage. These mice displayed a phenotype similar to the widespread MeCP2 deficiency, which supports the notion that RTT originates from its impaired function in the brain. Neuropathological examination of Mecp2 tm1.1B mice in Table 1: Summary of all mouse lines with loss-of-function deletions and single-point mutations in the Mecp2 gene, listing major differences in behavioural assessments. the symptomatic stage did not reveal any major abnormalities such as disorganized cortical lamination and ectopias. However, more detailed analyses showed that cortical growth is stalled as symptoms appear, and that neurons are smaller and more packed at a higher density in layer II/III of somatosensory and motor cortices [32]. In these cortical layers, the length of apical dendrites and dendritic spine density are lower compared to wildtype mice. These impairments are accompanied by the occurrence of immature dendrites and dendritic spines also in layers V and VI of the motor cortex [33]. In motor cortex layer V, however, diverse synaptic phenotypes were found: spine numbers are largely decreased in apical dendrites but not basal dendrites [34,35]. The immunoreactivity levels of the presynaptic marker VGLUT1 (vesicular glutamate transporter-1) and the postsynaptic marker PSD95 (postsynaptic density-95) are lower in hippocampal neurons grown in autaptic microisland cultures [36]. Intriguingly, the density of VGLUT1 puncta is lower in hippocampal area CA1 of perfusion-fixed Mecp2 tm1.1B mice than in wildtypes, but only at 2 weeks of age (presymptomatic stage), and not 5 weeks when overt symptoms start to appear [36], which again suggest that Mecp2 tm1.1B mice have diverse synaptic phenotypes in different brain regions. In the olfactory system, primary sensory neurons of Mecp2 tm1.1B mice have impairments in axonal fasciculation, path-finding, and terminal differentiation [37,38], but on the postsynaptic side, no abnormalities in the mitral cells were observed [39]. The consequences of MeCP2 deficiency in Mecp2 tm1.1B mice are also manifested in postnatal neurogenesis processes, including delayed maturation of newly-born neurons, altered expression of presynaptic proteins, and reduced formation of dendritic spines [40]. However, it does not affect the generation of immature neurons, which is in agreement with an early conclusion that MeCP2 is not involved in neuronal differentiation but only in neuronal maturation and maintenance in a cell-autonomous manner [41,42].

Mecp2tm1.1J mice (a.k.a. Jaenisch mutants)
Unlike the complete loss of MeCP2 protein expression in Mecp2 tm1.1B mice, the deletion of only exon 3 in Mecp2 tm1.1J mice leads to the expression a protein fragment lacking part of the MBD but with an intact but seemingly non-functional C-terminus [43]. Indeed, these mice (in either 129, C57BL/6, or BALB/c backgrounds) show a very similar neurological phenotype to Mecp2 tm1.1B mice, although slightly milder and with a longer lifespan (10-12 weeks). Brain-specific Mecp2 exon 3 deletion by recombination of floxed Mecp2 (Mecp2 2loxB ) with Nestin-Cre also results in phenotypical features identical to that of widespread whole-body deletion, as in Mecp2 tm1.1J mice. At the neuropathological level, Mecp2 tm1.1J mice also have reduced brain weight and neuronal size [43].
More detailed analyses of different brain regions at the cellular level revealed that both mice have similar characteristics, but that Mecp2 tm1.1B mice generally exhibit a more severe phenotype [44,45]. For example, dendritic spine density of all principal neurons of the hippocampus is lower in Mecp2 tm1.1B mice than in wildtype controls, while this feature is only apparent in area CA1 of Mecp2 tm1.1J mice [46]. However, a recent study showed that CA1 pyramidal neurons in the hippocampus of Mecp2 tm1.1J mice have lower dendritic spine density and smaller asymmetric spine synapses than wildtype controls, but only in the presymptomatic stage [47].

Mecp2 tm1Tam mice
Mecp2 tm1Tam mice lack the entire MBD sequence, and carry a non-functional splicing site that prevents splicing and transcription of the downstream Mecp2 sequence, leading to the complete absence of MeCP2 protein in the brain [48]. Male mutant mice stop to grow at 4-5 weeks of age and die at 8-12 weeks (mice in the 129 background typically die earlier than those in the mixed 129/C57BL/6 background), while heterozygous females have the same growth profile as wildtype littermates. These mice showed cerebellumdependent motor ability deficits, and impaired hippocampus/amygdala-dependent behavioral tasks, such as cued and contextual fear learning [48].

Mecp2 Stop mice
To allow a more controlled deletion of Mecp2, its endogenous transcription was disrupted by inserting a lox-STOP-Lox cassette before exon 3 (Mecp2 Stop ) that can be removed by Cre-mediated recombination [26,49]. Similar to Mecp2 tm1.1B mice, these mice (C57BL/6 background) do not show any signs of MeCP2 protein expression, and develop identical RTT-like neurological phenotypes. Male Mecp2 Stop mice typically start to show progressive symptoms at ~6 weeks of age with death occurring at ~11 weeks, while heterozygous females are fertile and display neurological phenotype only after 4-12 months. A recent study described that these mice show thinner neocortex and corpus callosum, smaller cell body size, and shorter length of basal dendrites in layer II/ III pyramidal cells of the motor cortex than their wildtype controls [50]. At the synaptic level, spine density in hippocampal area CA1 is lower compared to wildtype controls. They also show impaired long-term potentiation (LTP) at excitatory synapses in the hippocampus, which may lead to learning and memory deficits [49].

Mecp2 308 mice
Mecp2 308 mice (either C57BL/6J or 129/SvEv backgrounds) were generated by introducing a premature STOP codon in the Mecp2 gene, which leads to the expression of a protein truncated at amino acid 308 containing the part of the MBD, TRD, and nuclear localization signal [51]. Under careful examination, male hemizygous Mecp2 308 mice display more intense curling, and more pivoting and head rising behaviors as early as postnatal day 3 [52]. Symptoms become evident at 6 weeks of age, which include tremors, hypoactivity, kyphosis, myoclonic seizure, and stereotypic forelimb motion. Overall, these symptoms are much milder than those observed in Mecp2 tm1.1B and Mecp2 tm1.1J mice. Also different from those mice, male Mecp2 308 mice have normal body and brain weight, and can survive over 1 year. No obvious histological abnormalities were found in the brain. Detailed examination of neurons and glial cells using several markers, including microtubule-associated protein 2 (MAP2) and glial fibrillary acidic proteins (GFAP), did not reveal any significant differences with agematched wildtype littermates [51]. Consistent with those findings, the same group reported that except for subtle changes of PSD length, neither the dendritic arborization in layer III and V pyramidal neurons of the frontal cortex nor the density and distribution of synaptic vesicles are affected in male Mecp2 308 mice more than 1 year old [53].
Intriguingly, Mecp2 308 mice exhibit upregulation of corticotrophinreleasing hormone (CRH) and its associated heightened anxiety [51,54,55], which differs from other Mecp2 mutant mice [44,48,56], and hypomorph floxed mice [29,30]. The cellular bases for increased anxiety have been related to the amygdala; consistently, both synaptic maturation and developmental elimination are irregularly enhanced in principal neurons of the lateral nucleus of Mecp2 308 mice [57]. Again owing to the potentially aberrant function of truncated MeCP2 protein, behavioral changes are often manifested in a subtle way and giving rise to some conflicting observations, such as for fear learning [51,53] and social interaction tasks [54,58,59].

Specific Spatio-Temporal Deletions
Deletion in forebrain postmitotic neurons: Mecp2 2loxJ ; CamK-Cre93 mice To generate mice with Mecp2 deletion restricted to forebrain postmitotic neurons, Mecp2exon3-floxed (Mecp2 2loxJ , Jaenisch floxed) males were crossed with females expressing Cre under control of the CamK promoter [43]. MeCP2 protein reduction in this conditional mouse line starts at the perinatal phase and reaches a maximal level at postnatal day 21. These mice have a delayed and milder phenotype than those resulting from germline or mediated deletions [43]. Symptoms include hindlimb clasping, impaired diurnal activity, and cue-sensitive Nestin-Cre-fear conditioning; however, context dependent fear conditioning is unaffected [60]. Weight gain, elevated anxiety, and abnormal social interaction were also pronounced in these mice [60]. Histological analyses showed smaller neuronal somata in the cortex and hippocampus, but not in the hindbrain such as cerebellum [43]. One of the interesting conclusions we can draw from these studies is that the cerebellum, a brain region critical for motor control, may not contribute to the manifestation of impaired motor coordination. Loss of Mecp2 in broad forebrain regions and in some regions of the basal ganglia is sufficient to cause the observed motor deficits.

Deletion in hypothalamic neurons: Mecp2 2loxB ; Sim1-Cre mice
Breeding heterozygous Mecp2 2loxB (Bird floxed) females with males (FVB background) expressing Cre under control of the Sim1 promoter generated hypothalamus-specific Mecp2 conditional knockout mice [61]. Immunohistochemical staining demonstrated the absence of MeCP2 protein in several primary regions of the hypothalamus, including the paraventricular (PVN) and supraoptic nuclei (SON). As expected from such cell-specific Mecp2 deletion, these mice recapitulate a subset of the phenotypes observed in patients with typical or atypical RTT. These mice show stereotypical movement, kyphosis, and increased body weight, but lack tremor, impaired motor coordination, and learning and memory deficits. Furthermore, hypothalamus-mediated behaviors are impaired, resulting in enhanced anxiety, and increased aggression in response to stress [61]. On the basis of this study, we can conclude that proper MeCP2 function in the hypothalamus is critical for homeostatic responses to a variety of physiological stimuli, which agrees with heightened anxiety and altered stress in RTT patients, symptoms indicative of hypothalamic dysfunction.

Deletion in dopaminergic and noradrenergic neurons (Mecp2 2loxB ; TH-Cre mice) and serotonergic neurons (Mecp2 2loxB ; PET1-Cre mice)
To specifically remove Mecp2 from dopaminergic and noradrenergic neurons, Mecp2 2loxB (Bird floxed) females were bred with males (FVB/N background) expressing Cre under control of tyrosine hydroxylase (TH) promoter [62]. Deletion of Mecp2 from serotonergic neurons was achieved by Cre-mediated recombination in PC12 ets factor 1(PET1)expressing neurons (C57BL/6 background) [62]. In Mecp2 2loxB ;TH-Cre mice, MeCP2 is significantly decreased in neurons of substantia nigra or ventral tegmental area of locusceruleus. Mecp22loxB;PET1-Cre mice show MeCP2 deficient neurons in the dorsal and medial raphe nuclei. By comparing the phenotype between Mecp2 2loxB ;PET1-Creknockouts, their wildtype counterparts, and Mecp2 2loxB Bird floxed hypomorph mice, Samaco et al. [62] found no obvious differences in several features, including breathing pattern, repetitive behaviors, anxiety levels, motor learning and social interaction. However, conditional knockout mice with Mecp2 deletions in dopaminergic and noradrenergic neurons showed subtle but characteristic phenotypes, such as hypoactivity and aggression, respectively. This study provides strong evidence for the importance of MeCP2 in the regulation of aminergic transmitters, and contributes to our understanding of the role of different transmitters in specific phenotypes.  [63]. Mecp2 2loxB ;Viaat-Cre mice exhibit the stereotypic neurological phenotype observed in Mecp2 tm1.1B and Mecp2 tm1.1J mice, developing symptoms at 5 weeks of age and having a half-survival of 26 weeks. An additional number of features distinct from Bird nulls and Jaenisch mutants include impaired sensorimotor gating and arouse (assessed with acoustic startle response and pre-pulse inhibition tasks) and enhanced social interests, but without altered anxiety and aggression. Interestingly, the expression levels of the GABA synthetic enzymes glutamic acid decarboxylase 65 (Gad65) and Gad67 are reduced in Mecp2 2loxB ;Viaat-Cre mice, with a concomitant reduction in vesicular GABA content and impaired synaptic inhibition. To further dissect MeCP2 function in forebrain interneurons, male mice (FVB background) expressing Cre under the Dlx5/6 promoter were crossed with Mecp22loxB females [63]. It is worth noting that the Dlx5/6 promoter is only expressed at high levels in parvalbumin-positive forebrain interneurons that innervate pyramidal cell somata, but is not expressed in interneurons that target pyramidal cell dendrites, such as calretinin positive and somatostatin-positive interneurons [64]. The phenotype observed in Mecp2 2loxB ;Dlx5/6-Cre mice thus resulted from MeCP2 deficiency only in a subset of forebrain interneurons with a specific functional role. Compared to the more widespread Mecp2 deletion in all interneurons in the Mecp2 2loxB ;Viaat-Cre mice, Mecp2 2loxB ;Dlx5/6-Cre mice show milder symptoms and longer life-span, surviving to over 80 weeks. Except for decreased motility, Mecp2 2loxB ;Dlx5/6-Cre mice show similar phenotypic features presented by Mecp2 2loxB ;Viaat-Cre, including hindlimb clasping, reduced locomotor activity, kyphosis, breathing dysfunction. [63]. It is surprising that unlike other conditional mutants, deletions of Mecp2 in all interneurons or even in a specific subset of interneurons result in a multitude of RTT-like symptoms usually observed in widespread Mecp2 deletions. A tenable conjecture will be that perturbation of MeCP2 function even in small-scale territory of GABAergic interneurons is enough to ruffle the neuronal network and ultimately result in a phenotype similar to that seen in mice with a widespread loss of MeCP2 function.

Deletion in glial cells: Mecp22loxJ; hGFAP-CreT2 mice
Neuronal dysfunction was initially considered the sole basis for RTT etiology, but accumulating evidence indicates that glial cells may also have an important role in disease pathophysiology [12]. Contrary to an initial belief, MeCP2 is expressed in significant levels by glial cells, including astrocytes and oligodendrocytes [22]. Furthermore, astrocytes from Mecp2 tm1.1J mice had detrimental consequences on either wildtype or Mecp2-deficient neurons, an effect also observed by astrocyte-conditioned medium [22]. Consistently, cytokine production altered astrocytic glutamate clearance in Mecp2 tm1.1B mice [65,66]. In addition to astrocytes, microglia also shows impaired function in Mecp2 tm1.1B mice, as they release excitotoxic levels of glutamate [24], and do not respond properly to immunological stimuli [25].
To confirm these unexpected roles of MeCP2 role in glial cells, Mecp22loxJ (Jaenisch floxed) mice were crossed with mice (mixed FVB/N/C57BL/6 background) harboring a tamoxifen-inducible Cre transgene driven by the human GFAP regulatory element (Mecp22loxJ;hGFAP-CreT2). The specific Mecp2 deletion in astrocytes after postnatal day-21 results in a number of classical RTT-like symptoms, including decreased body weight, hindlimb clasping, and irregular breathing, while longevity, locomotion, and anxiety were not affected [23]. The observation of normal dendritic complexity in neurons of these astrocyte-specific Mecp2 null mice appears to contrast with earlier reports in widespread Mecp2 mutant mice; however, it is conceivable that Mecp2 deletion in astrocytes at this later age (i.e. P21) does not seriously compromise already established neuronal morphology. These studies demonstrate that functional loss of glia, an integral component critical for proper brain function, would breach the intimate relationship with neurons and lead to the development of the RTT phenotype.

Viral-mediated Mecp2 deletion: Cre-and shRNA-expressing viruses
To achieve Mecp2 deletion in a spatial as well as temporal manner, viral-mediated approaches have been successfully employed. Targeted Mecp2 deletion achieved by injections of Adeno-Associated Virus (AAV) expressing Cre recombinase in the basolateral amygdala of Mecp2 2loxJ mice (Jaenisch floxed) resulted in increased anxiety and deficits in cue-dependent fear conditioning, without affecting motor coordination, locomotor activity and social behavior [67]. Also, MeCP2 protein expression was knockdown with short hairpin RNAs (shRNAs) designed to target Mecp2 mRNAs. Rat pups that received intraventricular injections of MeCP2 shRNA show altered sensory-motor reflexes and neurobehavioral abnormalities during early development, but only transiently [68]. Even though MeCP2 knockdown by shRNA injections into the dorsal striatum resulted in decreased intake of a drug-of-abuse, RTT-like phenotypes were not apparent [18]. Infusions of MeCP2 shRNA-expressing lentivirus into the Nucleus Accumbens (NAc) had a converse effect (increased drug reward response), but also without occurrence of RTT like symptoms [69]. It is undoubted that new findings would emerge by using these vial mediated knockdown techniques, which can achieve local ablation of gene expression inaccessible to DNA promoter-driven knockout otherwise.

Tamoxifen-driven expression of Cre recombinase: Mecp22loxB; CreER mice
Through its interaction with the estrogen receptor (ER), tamoxifen triggers the nuclear translocation of an engineered Cre-ER complex in CAGGS-CreER mice, allowing the Cre recombinase to interact with loxP sites flanking gene regulatory regions of interest and excise them. Systemic tamoxifen injections allowed Mecp2 deletion at postnatal-day 60 or older Mecp2 2loxB ; CAGGS-CreER mice (C57BL/6J background), which resulted in characteristic RTT-like behavioral deficits and death 13 weeks after treatment onset [27]. A follow-up study revealed that conditional Mecp2 deletion either at 5 weeks or at 11 weeks resulted in similar disease progression and lifespan (16-17 weeks after onset of tamoxifen treatment) [70]. Intriguingly, despite the RTT-like disease progression and fatality, no differences were found at the cellular and molecular levels, including neuronal density and soma size, dendritic complexity, and expression of several synaptic proteins. Another study extending the age of tamoxifen onset to even earlier and later found that, although all mice eventually reached the same disease severity by 10 weeks after tamoxifen onset, mice treated at 3 weeks-of-age develop symptoms more stagnantly in the first week compared to mice that received tamoxifen starting at 11-and 20-weeksof-age [71]. It is also interesting to note that, irrespective of ages for tamoxifen treatments, all mice die at ~40 weeks of age. These studies suggest that MeCP2 levels are critical during an early age (8-14 weeks) when synaptic refinement occurs, as well as later (30-45 weeks) when aging processes initiate. Even though this hypothesis awaits further examination, it is increasingly clear that proper MeCP2 levels are of great importance in the adult brain. The use of tamoxifen-induced Cre recombination has produced fruitful results regarding the required timing of MeCP2 function. In combination with several models of local Cre-ER expression, it has enabled us to map the neuroanatomic origins of phenotypes observed in RTT mouse models, shedding light on the sources of clinical presentations in RTT individuals.

Single Point Mutations
Missense or nonsense MECP2 mutations, including R106W, R133C, T158M, R168X, R255X, R270X, R294X, and R306C, account for more than half of the RTT cases [72]. Thus, mice carrying a single nucleotide change in Mecp2 would represent the closest experimental model of patients with RTT.

Truncated MeCP2: Mecp2 R168X mice
Mecp2 R168X mice (mixed C57BL/6, 129S6/SvEv Tac background) carry a single point mutation, where the codon for arginine at position 168 is replaced with a premature STOP codon [73]. These mice express a truncated MeCP2 protein at ~50% wildtype MeCP2 levels. As seen in other Mecp2 mouse lines, male mutant mice are more severely affected than female heterozygotes. Male Mecp2 R168X mice have a short lifespan (~12 weeks) with neurological features starting as early as 5-6 weeks, while female heterozygotes show symptoms only after six months of age and live longer than 1 year. Symptoms in males consist of forelimb stereotypes, hindlimb clasping and atrophy, hypoactivity, and breathing irregularities, while females have similar but milder signs. Compared with two main mutant mice in phenotypic severity, they are milder than Mecp2 tm1.1B and close to Mecp2 tm1.1J mice. A note added in proof in the Lawson-Yuen et al. [73] paper warned about mistakenly introduced mutations after the STOP codon, which preclude the use of these mice to test read-through compounds that skip STOP codons, such as aminoglycoside antibiotics (i.e. the full-length protein will express several mutations of unpredictable consequences).

MECP2 mutation in males: Mecp2 A140V mice
The Mecp2 A140V mouse model is unique because it mimics atypical RTT cases occurring mostly in males. The Mecp2 A140V mutation is known to retain the methyl-CpG binding function while losing its ability to be associated with the mental disability-related protein ATRX [74]. Mecp2 A140 mice (mixed 129X1/S1, C57BL/6 background) show normal levels of Mecp2 mRNA and protein, which may allow normal lifespan, weight, and motor coordination, as well as breathing rhythms. However, they do show some RTT-like features, such as increased neuronal cell density, and decreased complexity of dendritic branching in layer II/III pyramidal neurons of the somatosensory cortex [75]. [12]. Phosphorylation of MeCP2 sites, especially at serine 421, has been under intense investigation. Several studies suggested a de-repression model in which phosphorylation of MeCP2 following neuronal activity prevents it from binding to target genes and results in expression of genes and the ensuing modulation of dendritic patterning and dendritic spine morphogenesis [76][77][78]. This view has been revised by a recent study using MeCP2 S421A point mutation mice (mixed sv129, C57BL/6 background) and ChIP-Seq [79]. These mutant mice exhibit some cellular phenotype including an increase in the number of dendritic branches in cortical pyramidal neurons, and a shift towards synaptic inhibition in the excitation/inhibition. At the behavioral level, responses to a novel mouse were significantly impaired, but motor activity, social interaction, spatial learning and memory were unaffected. It is surprising that, in this study using ChIP-Seq analysis for genome wide DNA-protein association and ChIP-qPCR for specific association, they did not find any significant changes in MeCP2 binding to target genes. These findings suggest phosphorylation of MeCP2 at serine 421 does not regulate specific genes, but instead plays a role in fine-tuning global gene expression. However, the authors also pointed out that global histone-like manipulation of MeCP2 at serine 421 does not exclude the possibility of facilitating discrete gene expression in the presence of other locus-specific modifications of MeCP2. This notion, although not been tested, could be true. Interestingly, a recent study showed that the additional substitution of serine for alanine at position 424 in Mecp2 S421A;S424A mice (C57BL/6 background) resulted in enhanced contextual fear learning, spatial memory, hippocampal synaptic plasticity (i.e. LTP at CA3-CA1 and mossy fiber-CA3 synapses), and synaptogenesis, while motor activity, coordination and nervousness were unaffected [80]. However, an earlier study using mice (C57BL/6 background) with similar mutations had shown increased locomotor activity [81]. The conflicting observations in these studies suggest complex and diverse consequences of phosphorylation at different MeCP2 sites. Phosphorylation at S421, for example, was revealed to result in completely different outcomes as for phosphorylation at S80, as discussed next.

Phosphorylation mutants: Mecp2 S80A mice
Mice (C57BL/6 background) with serine 80 replaced by alanine in MeCP2 (Mecp2 S80A ) to prevent phosphorylation at this site did not exhibit any RTT-like phenotypes, except for decreased locomotor activity [81]. In striking contrast to phosphorylation at S421, S80 phosphorylation was negatively regulated by neuronal activity. The finding of an opposing regulation of S421 and S80 phosphorylation by neuronal activity raises the intriguing possibility that MeCP2 phosphorylation is engaged differentially under resting and active states.

Mecp2 hypomorphs: Mecp2T158A mice
Threonine 158 in MeCP2 is known to play a crucial role in the stabilization of the MBD and the binding of MeCP2 to its target genes [82]. Since a threonine 158 mutation is common in RTT individuals, Mecp2 T158A mice (sv129, C57BL/6 background) were generated to investigate its role in RTT pathogenesis [56]. Male Mecp2 T158A mice develop RTT-like symptoms at 5 weeks of age and die at ~16 weeks, a similar or slightly milder disease than that seen in Mecp2 R168X mice. Similar to all other mouse models, and likely due to the mosaicism of an X-link mutant allele, female Mecp2 T158A heterozygous mice do not present with symptoms until 17 weeks of age, and live a normal lifespan. In addition to the characteristic RTT-like phenotype, which includes hypoactivity, hindlimb clasping, impaired motor coordination, and learning & memory deficits, Mecp2 T158A mice have an initial body weight loss that later subsides, a different pattern of weight variability than that presented by Mecp2 R168X male mice. Unlike the complete absence of MeCP2 in Mecp2 R168X mice and the normal levels of MeCP2 protein in Mecp2 A140V , Mecp2 S421A /Mecp2 S421A;S424A , and Mecp2 S80A mice, symptomatic Mecp2 T158A mice show ~60% lower levels of MeCP2 protein compared to age-matched littermate controls. This reduction in MeCP2 levels is thought to underlie impaired expression of target genes (e.g. Bdnf).

Transgene Rescue of Mecp2 expression
Ever since the discovery that MECP2 mutations cause RTT, despite the recognized challenge to only target the mutant allele in a X-linked disorder, prevention or reversal of RTT phenotypes has been attempted by means of transgene expression in experimental mouse models. An unexpected corollary of these studies is that elevated levels of MeCP2 are as detrimental to neurological function as their reduction [83].

Neuronal Mecp2 overexpression: Tau-Mecp2; Mecp2tm1.1J mice
An additional copy of the mouse Mecp2 gene was inserted into the Tau locus (encoding the neuron-specific microtubule associated protein Tau) to test for the rescue of RTT-like phenotypes by increasing MeCP2 protein levels in neurons [84]. However, it was found that homozygous Tau-Mecp2 mice (FVB background) exhibit similar symptoms as Mecp2tm1.1J mice (Jaenisch mutants), likely due to excessive levels MeCP2 protein under control of the Tau promoter. Incidentally, a similar neurological phenotype was observed in mice that overexpressed the human MECP2 gene from a P1-derived artificial chromosome containing all regulatory elements [85], demonstrating that MeCP2 overexpression is detrimental to brain function. When Tau-Mecp2 mice (mixed 129, C57BL/6 background) were bred with Mecp2 tm1.1J (Jaenisch mutants), their progeny showed a significant reduction of all RTT-like phenotypes examined [84], suggesting that the mild increase in MeCP2 levels from one copy of Tau locus was sufficient to prevent disease onset but not too high to be deleterious.
A construct containing the synthetic CAGGS promoter, a loxP flanked STOP cassette, and mouse Mecp2 cDNA was targeted to the Col1a1 locus and subsequently transferred into ES cell to generate CAGGS-Mecp2 stop mice (C57BL/6 background) [86]. The lack of Mecp2 expression due to a STOP codon results in the characteristic RTT-like phenotypes in CAGGS-Mecp2 stop mice. These mice were crossed with Mecp2 tm1.1J mice (Jaenisch mutants), and the resulting offspring were bred with different Cre-expressing transgenic mice. Removal of the STOP codon by Cre-mediated recombination leads to reactivation of the CAGCS-Mecp2 transgene and significant MeCP2 protein expression, which was higher in Nestin-Cre (neural lineage) and Tau-Cre (neuronspecific microtubule associated protein) mice than in CamK-Cre mice (specific to forebrain neurons). Mice with higher Mecp2 transgene expression showed longer lifespan than Mecp2 tm1.1J mice without overt RTT-like phenotypes, while mice with lower Mecp2 transgene expression showed slight improvements compared to Mecp2 tm1.1J mice. For example, CamK-Cre rescue mice lacked the characteristic low brain weight and small neuronal soma size observed in Mecp2 mutant mice [86].

CreER-mediated Mecp2 reactivation by systemic tamoxifen treatment: Mecp2 StopB ; CreER mice
As above, lack of Mecp2 expression due to a STOP codon results in characteristic RTT like phenotypes in Mecp2 StopB ;CreER mice (C57BL/6 background). Excision of the STOP cassette by tamoxifeninduced nuclear translocation of the CreER complex results in the reexpression of Mecp2 under control of native regulatory elements [26]. Compared to the rapid progression of RTT-like phenotypes in male Mecp2StopB;CreER mice, mice receiving tamoxifen starting at 3-4 weeks after birth showed a significant delay in symptom progression and a longer lifespan of up to 17 weeks of age. Similar delayed onset symptoms and extended lifespan were observed in female Mecp2 StopB ;CreER mice after tamoxifen treatment.

MECP2-TREEGFP; CamK-tTA/Eno2-tTA; Mecp2 308 mice
A MECP2-TRE-EGFP construct was designed to allow for bi-directional regulation of both human MECP2 and EGFP (for identification of expressing cells) under the control of a central TRE (tetracycline-responsive element) [87]. TRE activates gene transcription once its tetO recognition sequence is bound by tTA (tetracycline transactivator), which is given to mice orally in the drinking water. Progenies resulting from the crossbreeding of mice containing the MECP2-TRE-EGPF transgene with those containing the CamK-tTA or Eno2-tTA transgenes express MeCP2 in neurons of the forebrain and the striatum and cerebellum, respectively. These inducible mice (C57BL/6 background) were crossed to Mecp2308 mice to test for the prevention of RTT-like symptoms. Surprisingly, cell specific expression of Mecp2 failed to prevent most of the phenotypes characteristic of Mecp2308 mice, except for a marginal increase in motor activity in CamK rescue mice (forebrain neurons). However, earlier studies described that specific manipulations of Mecp2 expression in forebrain neurons resulted in detectable, albeit mild, behavioural consequences [43,60,86]. The expression of a partially functional truncated protein in Mecp2308 mice could contribute to this discrepancy. Also, normal levels of Mecp2 transgene expression may increase MeCP2 protein dosage too much, and thus compromise any potential beneficial effects. Indeed, using a similar strategy, female heterozygous Mecp2 tm1.1B mice with optimal Mecp2 transgene expression showed improved rearing and locomotor activity [88].

CreER-mediated Mecp2 reactivation in glial cells by systemic tamoxifen treatment: Mecp2 Stop ; hGFAP-Cre mice
Glial-specific reactivation of endogenous Mecp2 silenced by a loxPflanked STOP codon was achieved by excising it with a tamoxifeninducible CreER placed under control of the human hGFAP promoter [23]. Despite lack of Mecp2 expression in neurons, Mecp2 reactivation exclusively in glial cells extended the lifespan to 7.5 months compared with only 3 months in oil-treated Mecp2 Stop ;hGFAP-Cre mice (C57BL/6 background). Tamoxifen-treated Mecp2 Stop ;hGFAP-Cre mice show increased activity levels, decreased anxiety, while the characteristic irregular breathing (apnea) frequency is also improved compared to oil-injected controls. Consistent with these improvements in behavioural features, the size of neuronal somata dendritic arborization, and VGLUT1 protein expression were all significantly increased in tamoxifen-treated Mecp2 Stop ;hGFAP-Cre mice.

Introduction of wildtype microglia into Mecp2 null brain: bone marrow transplants in Mecp2 Stop ; Lysm-Cre mice
Since Mecp2-deficient microglia has adverse effect on wildtype neurons in co-cultures [24], it is reasonable to speculate that repopulation of Mecp2-containing microglia would improve RTTlike symptoms in Mecp2 deficient hosts. Indeed, a recent study demonstrated that transplantation of wildtype bone marrow into presymptomatic Mecp2 null mice (postnatal day 28) after whole-body irradiation-mediated immune ablation resulted in engraftment of microglia-like myeloid cells into the brain parenchyma and arrested the progression of neuropathology including recovery of body and brain size, enhancement of lifespan, improvement of gait and locomotor activity (but not hindlimb clasping), and reduction of apneas and breathing irregularities [25]. However, no significant rescue was observed when bone marrow transplants were performed during the time RTT-like phenotypes begin to appear (postnatal day 40 or 45). Phenotypic prevention also failed if irradiation did not include the brain, which leads to reconstitution of only peripheral but not brain microglia-like myeloid cells. To further support this strategy, Lysm-Cre mice (mixed 129, C57BL/6 background), which express Cre under control of the Lysm promoter, were crossed with Mecp2 stop mice to generate progeny with normal Mecp2 expression by removal of STOP codon only in myeloid cells, granulocytes, and microglia. Such cellspecific Mecp2 reactivation was sufficient to improve breathing patterns, prolong lifespan, and increase both body weight and locomotor activity. Furthermore, long-term treatment of Mecp2Stop;Lysm-Cre mice with annexin-V, a pharmacological manipulation that blocks phagocytic activity of microglia, failed to prevent RTT-like disease progression, strongly suggesting that micrglia dysfunction contributes to RTT pathophysiology.

Concluding Remarks
The various experimental mouse models described above have enabled researchers to discover novel mechanisms at the molecular, cellular, and circuit levels that contribute to MeCP2 dysfunction in RTT, providing remarkable platforms for testing a wide variety of genetic, pharmacological, and physiological interventions to improve the quality of life in this debilitating disease. Notwithstanding, and considering the significant differences between research rodents and humans at all levels, there is great need to establish consistent and effective human-based models. Undoubtedly, the use of patientspecific human induced pluripotent stem stem cells (iPSCs) is by far the closest model experimental research and manipulations [89]. Neuronal differentiation of human iPSCs derived from RTT individuals has been successfully achieved [89][90][91][92][93][94], which one day should allow highthroughput screening of pharmacological libraries. With RTT research evolving into full-fledged endeavors in an ever-growing number of research laboratories, we hope that the scientific knowledge gained from these and new animal models will be translated into rational therapies available for RTT individuals in not too distant future.