The FMRP regulon: from targets to disease convergence

The fragile X mental retardation protein (FMRP) is an RNA-binding protein that regulates mRNA metabolism. FMRP has been largely studied in the brain, where the absence of this protein leads to fragile X syndrome, the most frequent form of inherited intellectual disability. Since the identification of the FMRP gene in 1991, many studies have primarily focused on understanding the function/s of this protein. Hundreds of potential FMRP mRNA targets and several interacting proteins have been identified. Here, we report the identification of FMRP mRNA targets in the mammalian brain that support the key role of this protein during brain development and in regulating synaptic plasticity. We compared the genes from databases and genome-wide association studies with the brain FMRP transcriptome, and identified several FMRP mRNA targets associated with autism spectrum disorders, mood disorders and schizophrenia, showing a potential common pathway/s for these apparently different disorders.


INTRODUCTION
Protein synthesis at subcellular sites is a well-conserved mechanism that allows the rapid expression of specific genes in response to localized cues (Xing and Bassell, 2013). During transport, mRNAs are stabilized via association with multiple and different trans-acting factors, such as RNA-binding proteins (RBPs) and non-coding RNAs, forming ribonucleoparticles (RNPs) that vary in size and composition during cell cycle and development.
In highly polarized cells, such as neurons, mRNAs are transported from the nucleus to dendrites and axons where these molecules undergo local translation and degradation (Steward and Schuman, 2003;Bramham, 2008;Cajigas et al., 2010;Doyle and Kiebler, 2012;Hornberg and Holt, 2013) according to their subcellular localization and cellular inputs (Bramham, 2008).
RNA-binding proteins recognize and bind mRNA targets through regulatory elements in the 5 and 3 untranslated regions (UTRs) (Pichon et al., 2012), and in some cases the coding regions are also involved in these interactions (Anko and Neugebauer, 2012). Binding to mRNAs is mediated through well-known RNAbinding motifs, which are often present in multiple copies (Clery et al., 2008) and typically bind short RNA sequences (Anko and Neugebauer, 2012). Several RBPs cooperate for the binding of mRNA, thereby increasing the specificity of this interaction (Matlin et al., 2005;Ule and Darnell, 2006). The actin cytoskeleton might well facilitate RNA recognition, as this structure associates with RBPs and coordinates the binding of these proteins to mRNA (Percipalle, 2009). However, individual RBPs bind to several mRNAs. The multi-targeted binding property of RBPs has led to a model of regulated gene expression in eukaryotes termed "the post-transcriptional operon" (Keene, 2007).
The fragile X mental retardation protein (FMRP) is a widely studied RBP in the brain. Silencing of the FMR1 gene encoding FMRP leads to fragile X mental retardation syndrome (FXS), the most common cause of inherited intellectual disability (Bagni et al., 2012). A majority of the clinical cases of FXS reflect a lack of FMRP due to a large trinucleotide CGG-repeat expansion in the 5 UTR of the gene, resulting in FMR1 gene silencing. Rare cases have been reported to carry partially deleted or mutated FMRP (De Boulle et al., 1993;Mila et al., 2000;Coffee et al., 2008;Collins et al., 2010). The FMR1 gene and FMRP have also been associated with the pathogenesis of other disorders, such as fragile X-associated tremor ataxia syndrome (FXTAS), premature ovarian failure (POF), and autism spectrum disorder (ASD) (Bagni et al., 2012).
Here, we briefly reviewed the structure and function of FMRP, a multifunctional RBP that regulates the transport, stability and local protein synthesis of hundreds of RNAs in the brain. We further discuss how anomalies in the expression of FMRP alter the condition of its targets and ultimately, highlight a subset of FMRP target mRNAs dysregulated in autism spectrum disorders (ASDs), mood disorders (MDs) including bipolar disorder (BD), major depressive disorder (MDD), attention deficit hyperactive disorder (ADHD), and schizophrenia (SCZ).
Structural studies of the FMRP domain have contributed to the understanding of the molecular function/s of this protein.
FMRP homodimerises and interacts with several cytoplasmic and nuclear proteins involved in mRNA metabolism and cytoskeleton-remodeling proteins (Bagni and Klann, 2012). Among the best characterized FMRP-interacting proteins are FIGURE 1 | FMRP exon structure comprising its functional domains. Upper frame: The red box at the N-terminus of exon 1 indicates the location of the CGG triplet repeat within the 5 UTR of the mRNA. The four RNA binding domains are: the N-terminus, the two K homology domains (KH1 and KH2) and the RGG box. Middle frame: FMRP domains interacting with NUFIP1, CYFIP1, CYFIP2, FXR1P, FXR2P, TDRD3, and SMN proteins. The FMRP amino acid sequence involved in these interactions is shown between the brackets. The nuclear localization signal (NLS) and the nuclear export signal (NES) are also indicated. Lower frame: The FMRP RNA binding domains and the RNA/mRNA targets directly bound are indicated.

CELLULAR AND MOLECULAR FUNCTIONS OF FMRP
Although in neurons, FMRP has been localized in the nucleus, cell body and dendrites (Willemsen et al., 1996), the cytoplasmic function of FMRP has been the most studied. FMRP forms large cytoplasmic RNPs containing several proteins and RNAs, and this protein is involved in the transport, stability and translation of several mRNAs (Bagni et al., 2012). One report suggested FMRP might also function as splicing enhancer (Didiot et al., 2008). Additionally, Drosophila FMRP has been related to the RNA-editing pathway (Bhogal et al., 2011).

REGULATION OF mRNA STABILITY
Initial studies performed in Fmr1 KO mice have revealed that the absence of FMRP alters the abundance of hundreds of mRNAs in the brain (Brown et al., 2001;Miyashiro et al., 2003;Gantois et al., 2006); a few mRNAs were found to be down regulated in all three studies. Further analyses on specific mRNAs showed that dysregulation occurred in specific brain areas and/or subcellular compartments, suggesting that FMRP might regulate the same mRNA in multiple ways (Miyashiro et al., 2003). FMRP modulates the stability of certain mRNAs by preventing or sustaining mRNA decay (De Rubeis and Bagni, 2010). As an example of the two opposite activities on different mRNAs, it has been shown that hippocampal FMRP protects PSD-95 mRNA from decay (Zalfa et al., 2007) in an activity-dependent manner; however, FMRP protein also facilitates the decay of nuclear RNA export factor 1 (NXF1) mRNA in mouse neuroblastoma (N2a) cells (Zhang et al., 2007). Furthermore, FMRP regulates PSD-95 mRNA stability in the hippocampus (Zalfa et al., 2007) and regulates translation at cortical synapses (Muddashetty et al., 2007). PSD-95 mRNA is an important player in synaptic plasticity and is affected in ASD (Feyder et al., 2010) and SCZ (Toro and Deakin, 2005).
The cortical region of the Fmr1 KO mouse brain shows the reduced expression of different GABA A receptor subunits (El Idrissi et al., 2005;Gantois et al., 2006), consistent with evidence of imbalanced GABAergic signaling in FXS patients. Taken together, FMRP-RNPs might play different roles in several brain regions and regulate mRNAs through different mechanisms according to the developmental stage and subcellular localization.

REGULATION OF mRNA TRANSLATION
The translational dysregulation of FMRP mRNA targets significantly contributes to the FXS phenotype (Bagni et al., 2012;Darnell and Klann, 2013). Initial studies performed in lymphoblastoid cells derived from FXS individuals showed an increased translation rate in several FMRP targets (Brown et al., 2001). The increased translation of FMRP mRNA targets was also observed in Fmr1 KO mice specifically at synapses, consistent with the idea that FMRP functions as a repressor of translation (Muddashetty et al., 2007;Narayanan et al., 2007;Napoli et al., 2008;De Rubeis et al., 2013).
One of the most affected and best characterized signaling cascades in fragile X is the mGluR (Bear et al., 2004). Upon mGluR receptor activation, FMRP-mediated translational block is released and protein synthesis can ensue. In the absence of FMRP, the increase in protein synthesis results in a receptor imbalance; an increase in the mGluR1 and mGluR5 activity and the reduced insertion of AMPA receptors at the surface that leads to enhanced mGluR long-term depression (mGluR-LTD) (Bear et al., 2004). mGluR-LTD is a form of synaptic plasticity that involves mRNA targeting and local protein synthesis and degradation (Bear and Malenka, 1994), and this condition can be induced through the application of (S)-3,5-dihydroxyphenylglycine (DHPG) (Wisniewski and Car, 2002) in a protein synthesisindependent manner (Huber et al., 2002). In Fmr1 KO mice, DHPG-induced LTD is strongly increased and these electrophysiological phenotypes established the "mGluR theory" in FXS (Bear et al., 2004).
FMRP activity is regulated through posttranslational modifications. DHPG-induced LTD also activates FMRP synthesis at synapses (Antar et al., 2004;Ferrari et al., 2007;Kao et al., 2010), which in turn is quickly degraded through the ubiquitinproteasome system (Hou et al., 2006). The effect of FMRP on protein synthesis is influenced by the phosphorylation status of FMRP (Ceman et al., 2003), via the mTOR pathway (Narayanan

FXS AND COMMONALITIES WITH OTHER DISEASES
FXS is the most common monogenic cause of ASD, and 30% of patients with FXS present autistic behaviors (Bagni et al., 2012). Early studies performed on heterozygous females carrying the fragile X mutant gene showed a greater frequency of psychopathologies associated with schizophrenia spectrum diagnoses (Reiss et al., 1988). Furthermore, carriers of premutated FMR1 alleles (reduced FMRP levels) have been associated with a significant degree of psychiatric disorders (Bourgeois et al., 2009). Recently, low FMRP levels have been detected in the postmortem brain from subjects with SCZ, BD and MDD (Fatemi et al., 2010;Kelemen et al., 2013;Kovacs et al., 2013) and in blood samples from schizophrenia patients . Some individuals that display psychoses also carry FMR1 full and pre-mutations (Jonsson et al., 1995;Ashworth et al., 1996;Khin et al., 1998).
It is not known whether decreased levels of FMRP are the cause or the consequence of the development of these disorders. However, it is tempting to speculate that the loss or reduced function of FMRP might lead to a dysregulation of particular FMRP target genes associated with ASD, SCZ, and MD, suggesting the correlation of certain FXS features with these neuronal disorders. Because the GABAergic system is dysfunctional in these disorders  and the lack of FMRP affects the expression of some GABA receptor subunits (D'Hulst and Kooy, 2007), it is reasonable to hypothesize that FMRP reduction might explain the alterations of proteins associated with the GABAergic system in these different neurological diseases. Indeed, recent findings showed that a selective activator of GABA B receptor reversed some FXS associated pathologies (Henderson et al., 2012).
It cannot be ruled out that certain proteins, which are risk factors for ASD, SCZ and/or MD, work together with FMRP and might disrupt the function of this protein in a disease context. Recently, it has been observed that topoisomerase Top3β, a risk gene for SCZ and ASD (Iossifov et al., 2012;Xu et al., 2012;Stoll et al., 2013), binds to FMRP and modifies the function of this protein in vitro, thereby supporting normal neurodevelopment and averting mental disorders (Xu et al., 2013). In addition, the authors observed that the disruption of either Top3β or Fmr1 genes in Drosophila led to a dysregulation of ptk2, which is genetically associated with SCZ (Walsh et al., 2008). Notably, CYFIP1 has been associated with ASD (Sahoo et al., 2006) (Doornbos et al., 2009;Van Der Zwaag et al., 2010;Von Der Lippe et al., 2010;Talebizadeh et al., 2013), SCZ and epilepsy (Sahoo et al., 2006;Tam et al., 2010;Zhao et al., 2012). In addition, we have recently shown that the CYFIP1 interactome contains many novel proteins associated with ASD, SCZ, and MDD, providing new perspectives to define the regulatory pathways shared by neurological disabilities characterized by spine dysmorphogenesis (De Rubeis et al., 2013), a common feature of several neuropsychiatric disorders (Penzes et al., 2011).
Over the last 10 years, several hundred putative FMRP mRNA targets have been identified in the brain FIGURE 2 | Venn diagram of FMRP mRNA targets associated with autistic spectrum disorder (ASD), schizophrenia (SCZ) and mood disorder (MD). FMRP neuronal target genes (1169) were compared with de novo gene disruptions (nonsense, splice and frameshifts) reported in GWAS are associated with ASD, as previously described (Iossifov et al., 2012;Neale et al., 2012;O'roak et al., 2012b;Sanders et al., 2012), the 528 genes from the SFARI (https://sfari.org/resources/sfari-base) and the 304 genes from the AutDB databases (http://autism.mindspec.org/autdb/Welcome.do). Approximately 120 FMRP target genes were associated with ASD (in black). Among the 120 target genes, included 35 genes were identified from the GWAS (in red) and 100 genes were identified from the two databases (in black). Fifteen genes showed overlap between the GWAS, the SFARI, and AutDB databases (in black/red color). The 1169 FMRP target genes were compared with the genes associated with SCZ and MD in the NHGRI GWAS database (http://www.genome.gov/gwastudies/) and 26 common FMRP targets were shown.
As represented in Figure 2, according to the GWAS, 35 FMRP target mRNAs are associated with ASD (in red), while the SFARI and AutDB databases have revealed that 100 FMRP target mRNAs are candidate genes for ASD (in black). Fifteen genes overlap between the results obtained in the GWAS and the SFARI and AutDB databases (in black/red color). This analysis shows that approximately 10% of the neuronal FMRP targets identified, in the above-mentioned studies, overlap with the genes associated with ASDs (120 out of 1169).
We also compared the 1,169 FMRP target mRNAs with 176 genes associated with BD, attention deficit-hyperactivity disorder (ADHD), mood disorder (MD), and SCZ (GWAS compiled by the National Human Genome Research Institute catalog http://www.genome.gov/). Twenty-six (out of 176) FMRP target mRNAs were also identified in this cohort (Figure 2, in gray). Because a few genes in this group (10) were also detected among the FMRP targets in the ASD group, it is reasonable to hypothesize that ASD, SCZ, and mood disorders (BD, MDD, ADHD) share certain common signaling pathways.

FUTURE PERSPECTIVES
Post-transcriptional studies have revealed that the FMRP regulon controls disease-related proteins that affect both neurodevelopment and adult brain plasticity and homeostasis. The emerging wave of genetic association studies has revealed a large number of risk genes for several neurodegenerative diseases and neurodevelopmental disorders, such as SCZ, ASD, and BD (http://www. genome.gov/gwastudies).
The risk genes for neurodevelopmental disorders, identified through GWAS, were compared with the list of the FMRP targets, and the results suggest that several pathways are dysregulated in FXS and might account for specific FXS phenotypes.
As the FMRP acts as a protein synthesis repressor, it is reasonable to propose that the FXS phenotype might reflect the overexpression of specific genes. However, FMRP not only regulates gene expression at the translational level, but it also influences the stability of several mRNAs. Furthermore, to determine the functional association of the FMRP regulon with the repertoire of genes altered in individuals carrying ASD, SCZ, and MD, it is important to investigate the dosage of these genes in individuals with FXS. Moreover, FXS is a neurodevelopmental disorder, and the absence of FMRP could affect the expression of specific targets at different developmental stages and in different brain areas. Further studies on FMRP targets and the FMRP interactome at specific developmental stages would help to determine the cause of these disorders and develop further strategies to ameliorate FXS. P., et al. (1996). Linkage analysis of the fragile X gene FMR-1 and schizophrenia: no evidence for linkage but report of a family with schizophrenia and an unstable triplet repeat. Psychiatr. Genet. 6, 81-86. doi: 10.1097/00041444-199622000-00008 Bagni, C., andKlann, E. (2012).