Metabotropic Glutamate Receptor 7: A New Therapeutic Target in Neurodevelopmental Disorders

Neurodevelopmental disorders (NDDs) are characterized by a wide range of symptoms including delayed speech, intellectual disability, motor dysfunction, social deficits, breathing problems, structural abnormalities, and epilepsy. Unfortunately, current treatment strategies are limited and innovative new approaches are sorely needed to address these complex diseases. The metabotropic glutamate receptors are a class of G protein-coupled receptors that act to modulate neurotransmission across many brain structures. They have shown great promise as drug targets for numerous neurological and psychiatric diseases. Moreover, the development of subtype-selective allosteric modulators has allowed detailed studies of each receptor subtype. Here, we focus on the metabotropic glutamate receptor 7 (mGlu7) as a potential therapeutic target for NDDs. mGlu7 is expressed widely throughout the brain in regions that correspond to the symptom domains listed above and has established roles in synaptic physiology and behavior. Single nucleotide polymorphisms and mutations in the GRM7 gene have been associated with idiopathic autism and other NDDs in patients. In rodent models, existing literature suggests that decreased mGlu7 expression and/or function may lead to symptoms that overlap with those of NDDs. Furthermore, potentiation of mGlu7 activity has shown efficacy in a mouse model of Rett syndrome. In this review, we summarize current findings that provide rationale for the continued development of mGlu7 modulators as potential therapeutics.


INTRODUCTION
Neurodevelopmental disorders are a group of conditions that present in early life and are characterized by the failure to meet typical developmental milestones. These disorders affect a significant fraction of the population: 15% of children aged 3 to 17 years old were reported to have a developmental disability in the years 2006 to 2008 (Boyle et al., 2011). The current Diagnostic and Statistical Manual of Mental Disorders (DSM-V) categorizes NDDs into six groups: intellectual disabilities (IDs), learning disorders, communication disorders, ASDs, ADHDs, and motor disorders (American Psychiatric Association, 2013). There is often overlap between these groups; for example, 31.6% of patients with ASD also fulfill the diagnostic criteria for ID (Christensen et al., 2016). In addition, NDDs are associated with many co-morbidities, including but not limited to: epilepsy, mood disorders, breathing abnormalities, sleep problems, and gastrointestinal issues (Mannion and Leader, 2013;Doshi-Velez et al., 2014). Individuals with NDDs can struggle to develop interpersonal relationships and face immense challenges in school and in the workforce. Treatment options remain limited and there is a great need to identify novel points of intervention to improve the quality of life of these patients.
A growing body of literature suggests that NDDs arise from complex interactions between the environment and the genome (van Loo and Martens, 2007;Hu et al., 2014). In some cases, NDDs can be traced to genetic abnormalities such as point mutations, gene deletions/duplications, or chromosomal rearrangements. Examples of such disorders include Down syndrome, RTT, Fragile X syndrome, and Angelman syndrome. Although a clear genetic cause is often rare, monogenetic disorders have helped to identify proteins and pathways that are required for proper neuronal development and maintenance. Interestingly, many genes that have been associated with syndromic and non-syndromic NDDs can be clustered into pathways involved in synaptic structure and function (Spooren et al., 2012;Sztainberg and Zoghbi, 2016). In this review, we focus on the metabotropic glutamate receptor 7 (mGlu 7 ), a GPCR that serves as an important regulator of synaptic transmission and plasticity. We will summarize current literature suggesting the involvement of mGlu 7 in NDDs and discuss its potential utility as a novel therapeutic target.
The large extracellular N-terminal ligand binding domain consists of two lobes that sit on top of one another, similar to a Venus flytrap. This structural similarity earned it the name VFD. Glutamate, the endogenous ligand for mGlu receptors, binds to a cleft in between the two lobes of the VFD (Kunishima et al., 2000;Pin et al., 2003;Niswender and Conn, 2010). The mGlu receptors function as constitutive dimers (Pin et al., 2005;El Moustaine et al., 2012), and dimerization primarily occurs at the level of the VFDs (Kunishima et al., 2000;Jingami et al., 2003;Pin et al., 2003;Levitz et al., 2016 (Pin et al., 2005;Muto et al., 2007). It is suggested that glutamate binding to one VFD alone is sufficient for activation (openclosed), but that full activation is achieved when both VFDs are ligand bound (closed-closed) (Kniazeff et al., 2004). Although mGlu 7 has been historically predicted to act as a homodimer, it has also been postulated that the receptor enacts some of its function through hetero-dimerization with other receptors, such as mGlu 8 (Doumazane et al., 2011;Kammermeier, 2015).
The cysteine-rich domain contains nine cysteine residues linked by disulfide bonds that are critical for propagating signals from the VFDs to the rest of the receptor (Rondard et al., 2006;Muto et al., 2007). After glutamate binding, signals are transduced through the cysteine-rich domain to the heptahelical domain where conformational changes allow for G protein coupling (Kunishima et al., 2000;Tateyama et al., 2004;Binet et al., 2007;Muto et al., 2007;El Moustaine et al., 2012). mGlu 7 and the other Group III mGlu receptors couple to G i/o , which inhibits adenylyl cyclase activity and reduces intracellular cAMP concentrations (Niswender and Conn, 2010). Furthermore, mGlu 7 activation can result in K + influx via G βγ -mediated opening of GIRK ion channels, and inhibition of Ca 2+ currents through N-and P/Q-type calcium channels (Millán et al., 2002(Millán et al., , 2003Martín et al., 2007). mGlu 7 is the most widely expressed mGlu receptor in the CNS with relatively high expression in the amygdala, hippocampus, and hypothalamus (Kinoshita et al., 1998). There are 15 splice variants of GRM7, six of which are predicted to be protein coding (Zerbino et al., 2018). The two major isoforms, mGlu 7a and mGlu 7b , differ at their C-termini and it is hypothesized that these distinct C-terminal tails mediate different proteinprotein interactions (Dev et al., 2001). While mGlu 7a and mGlu 7b are primarily expressed in the CNS (Flor et al., 1997;Corti et al., 1998;Kosinski et al., 1999), isoform specificity has been observed in peripheral tissues such as the testes, trachea, uterus, and salivary gland (Schulz et al., 2002). In the CNS, mGlu 7 receptors are primarily localized to presynaptic active zones in neurons where they can act as auto-or heteroreceptors to inhibit the release of their endogenous ligand, glutamate, the main excitatory neurotransmitter or GABA, the main inhibitory neurotransmitter, respectively (Shigemoto et al., 1996;Cartmell and Schoepp, 2000;Dalezios et al., 2002;Somogyi et al., 2003;Niswender and Conn, 2010). Compared to the other Group III mGlu receptors, mGlu 7 exhibits an extremely low affinity for glutamate (high µM to mM as opposed to high nM to low mM for the other Group III mGlu receptors). Because of this low affinity, it has been suggested that mGlu 7 functions as an "emergency brake" in the case of elevated glutamate levels (Niswender and Conn, 2010). This idea is supported by the observation that mGlu 7 knockout mice exhibit spontaneous seizures under certain contexts (Sansig et al., 2001).
To date, only one mGlu 7 -selective allosteric agonist, AMN082 (9, EC 50 = 260 nM), has been reported in the primary literature (Mitsukawa et al., 2005). AMN082 has been used for animal studies involving learning and memory and plasticity in the amygdala among other areas. However, it has been shown that AMN082 exhibits off-target effects, one of which is predicted to be inhibition of the serotonin transporter (SERT) (Sukoff Rizzo et al., 2011;Ahnaou et al., 2016), somewhat limiting its utility in vivo unless coupled with knockout studies.
In contrast to potentiators, there have been several mGlu 7 selective antagonists and NAMs reported in the literature. The antagonist XAP044 (10, IC 50 = 5.5 µM) binds within the VFD and has shown efficacy in both in vivo and in vitro experiments such as anxiety-, depression-, and fear-related behavioral tasks and electrophysiology (Gee et al., 2014). Originally labeled a Group II mGlu receptor antagonist (mGlu 2 and mGlu 3 ), LY341495 (11) was also found to have efficacy at mGlu 4 , mGlu 7 , and mGlu 8 with IC 50s of 22, 0.99, and 0.173 µM, respectively (Kingston et al., 1998) and has been used to study both groups of mGlu receptor since its discovery.
The mGlu 7 NAM 6-(4-methoxyphenyl)-5-methyl-3-pyridin-4-ylisoxazolo[4,5-c]pyridin-4(5H)-one (MMPIP, 12), reported in 2007, has been used for several studies involving mGlu 7 (Suzuki et al., 2007). MMPIP is able to inhibit the response of L-AP4, but its efficacy was later shown to be context-dependent. For example, the potency (IC 50 ) of MMPIP was 70 nM in a calcium mobilization assay utilizing cells expressing G α15 versus 718 nM in a thallium flux assay with cells expressing G αi/o , suggesting that its effects may be dependent on cellular background. Further, MMPIP was not effective in blocking an mGlu 7 -mediated depression of synaptic transmission in electrophysiological studies . ADX71743, reported in 2013, exhibits an IC 50 of 63 and 88 nM at human and rat mGlu 7 , respectively (Kalinichev et al., 2013). However, it also exhibits low activity at mGlu 2 (Kalinichev et al., 2013;Reed et al., 2017) and possesses an electrophilic ketone moiety that could result in covalent modification and subsequent offtarget effects. Most recently, Reed et al. (2017) have successfully developed a series of novel, chemically-distinct mGlu 7 NAMs based upon a phenylbenzamide scaffold. One of the analogs, VU6010608 (13), exhibited modest potency (IC 50 = 759 nM), but was cleared rapidly in rats (64.2 mL/min/kg) and exhibited low levels of brain penetration, making it challenging for in vivo CNS studies (Reed et al., 2017). These existing and emerging tools, coupled with mGlu 7 knockout mice, have provided an initial toolbox to begin elucidation of the function of mGlu 7 in normal and pathological conditions.

mGlu 7 IN SYNAPTIC TRANSMISSION AND PLASTICITY
Inhibition of neurotransmitter release by mGlu 7 is believed to be mediated by the inhibition of N-type and P/Q-type calcium channels through interactions with G βγ , PKC, PICK1, and reductions in intracellular cAMP (Millán et al., 2002(Millán et al., , 2003Perroy et al., 2002;Martín et al., 2007). Millan and Colleagues demonstrated that activation of cerebrocortical mGlu 7 with L-AP4 inhibited N-type calcium channels in a PKA-and PKCindependent manner, suggesting that the inhibition was caused via interactions with G βγ . These authors also demonstrated that mGlu 7 -mediated decreases in cAMP could reduce spontaneous glutamate release in the cerebral cortex (Millán et al., 2002). Additionally, Perroy et al. (2000) demonstrated that P/Q-type calcium channels were inhibited via a PKC-dependent pathway, where G i/o and/or G βγ can stimulate the PLC pathway in cultured cerebellar granule cells. They also showed that the scaffolding protein, PICK1, facilitates the interaction between mGlu 7 and PKC, and is required for receptor-mediated P/Q-type calcium channel inhibition in this context (Perroy et al., 2002). In contrast, Martín et al. (2007) demonstrated that mGlu 7 inhibited hippocampal P/Q-type calcium channels in a PKC-independent manner. The mGlu 7 -mediated inhibition of glutamate release is also dependent on interactions with calmodulin (CaM), where activated CaM allows for the displacement of G βγ from mGlu 7 and the subsequent downregulation of calcium influx into the cell via calcium channel inhibition (O'Connor et al., 1999). Moreover, mGlu 7 's interaction with MacMARCKS (macrophage myristoylated alanine−rich C−kinase substrate) competitively antagonizes CaM-mediated calcium channel inhibition (Bertaso et al., 2006). mGlu 7 's position within the active zone and its ability to modulate neurotransmitter release has led to numerous studies focused on its role in synaptic plasticity. Two major forms of synaptic plasticity include LTP and LTD, which are persistent changes in synaptic strength that are thought to be correlates of learning and memory (Bliss and Collingridge, 1993;Takeuchi et al., 2014). The role of mGlu 7 in synaptic plasticity has been best characterized within the hippocampus at several distinct synapses. mGlu 7 was first reported to mediate a form of LTD occurring in stratum radiatum interneurons within area CA3 (Laezza et al., 1999). At excitatory synapses onto interneurons expressing calcium-permeable AMPA receptors, LTP could be induced by high frequency stimulation and blocked by the Group II and Group III mGlu antagonist, LY341495. Further pharmacological experiments confirmed the specific involvement of mGlu 7 : only a high concentration of L-AP4 depressed synaptic transmission at these synapses and a Group II mGlu agonist showed no effect. A similar form of plasticity was later described at mossy fiber inputs onto SLINs in area CA3 (Pelkey et al., 2005). At SLINs expressing calcium-permeable AMPA receptors, high frequency stimulation of mossy fibers induced an LTD that required mGlu 7 activation and PKC-dependent depression of neurotransmitter release through P/Q-type voltage gated calcium channels (Pelkey et al., 2005(Pelkey et al., , 2006. Interestingly, in slices pretreated with L-AP4, internalization of mGlu 7 receptors revealed the ability of these synapses to undergo LTP instead of LTD in response to the same electrical stimulus. Surface expression of mGlu 7 , therefore, regulates the direction of plasticity at these synapses, making mGlu 7 a "metaplastic switch" that can modulate feedforward inhibition in area CA3. An additional class of interneurons in which mGlu 7mediated plasticity has been implicated is the OLM interneuron population within the stratum oriens of areas CA3 and CA1. At excitatory inputs onto OLM interneurons, mGlu 7 expression is preferentially enriched (Shigemoto et al., 1996) and proposed to be recruited by extracellular-leucine-rich repeat fibronectin type III domain containing 1, or ELFN1 (Tomioka et al., 2014). Sylwestrak and Ghosh demonstrated that ELFN1 knockdown in OLM interneurons decreases short-term facilitation and increases presynaptic release probability. Conversely, overexpression of ELFN1 in parvalbumin interneurons leads to short-term facilitation when these synapses typically undergo short-term depression (Sylwestrak and Ghosh, 2012). In slices from Elfn1 −/− mice, presynaptic release probability, short term facilitation, and LTP are reduced in patch-clamp recordings from OLM interneurons (Tomioka et al., 2014). Although this evidence is indirect, it suggests that mGlu 7 may be involved in these forms of synaptic plasticity since mGlu 7 is likely to be a major regulator of presynaptic release probability at these synapses due to its recruitment by ELFN1.
In addition to its role as an autoreceptor on excitatory terminals, mGlu 7 is also located on the terminals of interneurons within the hippocampus and modulates the release of GABA (Somogyi et al., 2003;Summa et al., 2013). This function of mGlu 7 is required for LTP in wild-type animals at SC-CA1 synapses through a mechanism of disinhibition (Klar et al., 2015). Importantly, deficits in LTP at this particular synapse have been reported in several models of NDDs (Jiang et al., 1998;Moretti et al., 2006;von der Brelie et al., 2006). At SC-CA1 synapses, mGlu 7 is the only presynaptic mGlu receptor present in adult animals and activation of mGlu 7 has been repeatedly shown to reduce field potentials at SC-CA1 (Baskys and Malenka, 1991;Ayala et al., 2008;Jalan-Sakrikar et al., 2014). Klar et al. (2015) demonstrated that mGlu 7 activation by the agonist LSP4-2022 also reduces evoked inhibitory post-synaptic currents recorded from CA1 pyramidal cells. LTP induced by high-frequency stimulation was blocked by ADX71743, but only when GABAergic transmission was intact. Recently, we showed that a chemically distinct mGlu 7 NAM, VU6010608, also blocked LTP induced by high-frequency stimulation at SC-CA1 synapses (Reed et al., 2017). Interestingly, hippocampal slices from Grm7 −/− mice have been reported to exhibit similar levels of LTP when compared to WT controls, but decreased short-term potentiation following high-frequency stimulation (Bushell et al., 2002). In these studies, slices from Grm7 −/− mice showed reduced facilitation during the high-frequency train, an effect that was also seen with ADX71743 by Klar et al. (2015). The presence of LTP in Grm7 −/− slices may be due to compensatory mechanisms during development, such as retained expression of mGlu 8 , which is present at SC-CA1 synapses earlier in development (Ayala et al., 2008). Re-expression of mGlu 8 is not unprecedented as the selective mGlu 8 agonist (S)-3,4-DCPG was recently shown to reduce synaptic transmission at SC-CA1 in slices from pilocarpine-treated rats, but not in those of age-matched controls (Dammann et al., 2018). While further studies will be needed to explain the current discrepancy between genetic and pharmacological approaches, these data indicate that mGlu 7 regulates high-frequency transmission at SC-CA1 synapses. Recently, Martín et al. (2018) demonstrated that prolonged activation of mGlu 7 leads to potentiation of excitatory post-synaptic currents recorded by pyramidal cells in CA1. This potentiation of neurotransmitter release is dependent on PLC and the vesicle release proteins Munc13-2 and Rim1α. These studies indicate that, under conditions of high-frequency stimulation, mGlu 7 activation favors potentiation of excitatory transmission, which could be an additional mechanism by which mGlu 7 modulates long-term plasticity in the hippocampus.
Beyond the hippocampus, a role for mGlu 7 in LTP has also been established within the amygdala. Synaptic plasticity in the hippocampus is believed to underlie associative learning and working memory, whereas plasticity in the amygdala is associated with aversion and emotional learning (Brasted et al., 2003;Sigurdsson et al., 2007). The allosteric agonist AMN082 has been shown to block LTP at thalamo-amygdala synapses in slices from rats and mice (Fendt et al., , 2013. This effect correlates with the ability of direct injection of AMN082 into the amygdala to block the acquisition of fear-potentiated startle behavior in rats  and fear learning in mice (Fendt et al., , 2013. Interestingly, Grm7 −/− mice exhibit a general deficit in fear learning and decreased LTP at thalamoamygdala synapses (Fendt et al., 2013). Reduction of LTP by both an agonist and gene ablation may be explained by AMN082's ability to cause rapid internalization of mGlu 7 receptors (Pelkey et al., 2007). This would suggest that AMN082 can act as a functional antagonist by decreasing surface expression and, therefore, receptor signaling. This hypothesis is further supported by the ability of the mGlu 7 antagonist XAP044 to block LTP within the amygdala, inhibit acquisition of conditioned fear, and reduce anxiety-like behavior (Gee et al., 2014). Together, these studies demonstrate that mGlu 7 promotes plasticity within the amygdala, which is in line with its involvement in behaviors of fear and anxiety.

ROLE OF mGlu 7 IN NDD-ASSOCIATED PHENOTYPES
Core symptoms and comorbidities of NDDs can include, but are not limited to: cognitive impairment, seizures, mood disorders, social deficits, and motor impairments (Mannion and Leader, 2013;Doshi-Velez et al., 2014). Many studies have demonstrated that modulation of mGlu 7 function via genetic and/or pharmacologic techniques is able to mimic some of these phenotypes in animal models, and these studies will be reviewed here. COGNITION mGlu 7 knockout animals (Grm7 −/− ) show deficits in tasks that test cognitive functioning. In a conditioned taste aversion task, which measures amygdala-dependent aversive learning, mice were given saccharin along with an intraperitoneal injection of the control, saline, or LiCl, which evokes malaise. In this task, Grm7 −/− mice did not associate the adverse effects of LiCl to saccharin in comparison to wild-type littermates, exhibiting a deficit in fear learning (Masugi et al., 1999). In addition, Masugi et al. (1999) and Goddyn et al. (2008Goddyn et al. ( , 2015 demonstrated that Grm7 −/− mice exhibit less freezing than wild-type animals in cued and contextual fear conditioning paradigms. Together, these results indicate a role for mGlu 7 in aversion learning, and also suggest that the loss of mGlu 7 causes impairments in these learning paradigms. mGlu 7 has also been demonstrated to play a role in cognitive tasks that do not rely on fearful or aversive stimuli. Callaerts-Vegh et al. (2006) showed that Grm7 −/− mice exhibit impaired short-term working memory in 4-and 8-arm radial maze tasks, committing more errors (visits to previously baited arms or un-baited arms) than their wild-type counterparts. Conversely, Grm7 −/− mice performed similarly to wild-type animals in radial maze tasks when they were modified to assess long-term memory. Furthermore, both Callaerts-Vegh et al. (2006) and Goddyn et al. (2015) have reported that the loss of mGlu 7 causes increased latency to locate a platform in the Morris water maze task of spatial memory. Interestingly, Grm7 −/− mice performed similarly to wild-type animals after increased training and in un-cued trials (Callaerts-Vegh et al., 2006). Together, these data demonstrate that mGlu 7 may play specific roles in tasks involving working and spatial memory.
Pharmacological studies have further confirmed a role for mGlu 7 in learning and memory. Hikichi et al. (2010) showed that administration of MMPIP, an mGlu 7 NAM, to wild-type mice reduced performance in object recognition and location tasks, suggesting that mGlu 7 is also involved in recognition memory. MMPIP also attenuates conditioned taste aversion learning in rats (Klakotskaia et al., 2013). Interestingly, MMPIP improved cognitive performance in Y-maze and object recognition assays in a mouse model of neuropathic pain with no effect on shamtreated animals (Palazzo et al., 2015). As discussed above, MMPIP exhibits cellular background-dependent differences in vitro, and also had no effect in an electrophysiological study of at SC-CA1 synapses in the hippocampus , which may complicate interpretation of in vivo data. Inhibition of mGlu 7 with the antagonist XAP044 also resulted in reduced freezing in mice during a contextual fear conditioning task, further supporting a role for mGlu7 in amygdala function (Gee et al., 2014). Activation of mGlu 7 with an allosteric agonist, AMN082, has been shown to modulate both the acquisition and extinction of conditioned fear, though the results seem to contradict findings from studies performed with XAP044 and Grm7 −/− animals (Fendt et al., , 2013Goddyn et al., 2008;Siegl et al., 2008;Dobi et al., 2013;Gee et al., 2014). Administration of AMN082 impairs the acquisition and enhances the extinction of fear learning (Fendt et al., , 2013Siegl et al., 2008;Dobi et al., 2013), but knockout animals exhibit similar phenotypes in conditioned fear paradigms (Goddyn et al., 2008;Fendt et al., 2013). AMN082 appears to exhibit a task-dependent phenotype, where mGlu 7 activation facilitates between-session extinction, but not within-session extinction in a fear conditioning model (Toth et al., 2012;Fendt et al., 2013). AMN082 was also shown to have effects in social fear; it impaired extinction and recall when administered prior to the social fear extinction task, but not when given before social fear conditioning (Slattery et al., 2017). However, Ahnaou et al. (2016) demonstrated that AMN082 produced similar sleep-wake and hypothermia phenotypes in Grm7 −/− and wild-type mice, suggesting that there may be off-target effects elicited by the compound. Additionally, administration of VU6005649, an mGlu 7/8 PAM, to wild-type mice, increases freezing in contextual fear conditioning (Abe et al., 2017).

SEIZURES
Seizures are often present in patients with NDDs, and mGlu 7 and its interacting proteins have been implicated in seizure activity. Sansig et al. (2001) observed that Grm7 −/− mice suffered from spontaneous sensory stimulus-seizures and were also more susceptible to subconvulsant doses of PTZ and bicuculline than their heterozygous or wild-type littermates. In addition, reduction of mGlu 7 activity with the NAM ADX71743 was sufficient to induce absence seizures (Tassin et al., 2016). Disruption of proteins that interact with mGlu 7 can also induce seizures in mice Tomioka et al., 2014). For example, PICK1 is a PDZ-domain containing protein that interacts with the C-terminus of mGlu 7 . The protein-protein interaction between PICK1 and mGlu 7 is important for stable mGlu 7 cell surface expression, proper trafficking of mGlu 7 to presynaptic active zones, and also for inhibition of P/Qtype calcium channels. Disruption of the interaction between PICK1 and mGlu 7 appears to interfere with mGlu 7 's inhibitory activity via decreased cell surface stability/expression or improper signaling and trafficking, resulting in a seizure phenotype in mice (Perroy et al., 2002;Bertaso et al., 2008;Zhang et al., 2008).
As mentioned previously, ELFN1 is a transmembrane protein that has been demonstrated to recruit mGlu 7 to distinct cell populations in the hippocampus and cortex (Tomioka et al., 2014). Most recently, ELFN1 was also shown to be a trans-synaptic allosteric modulator of Group III mGlu receptors; receptor modulation occurs through an ELFN1mediated alteration of G-protein coupling efficiency to the Group III mGlu receptors (Dunn et al., 2018). Of note, ELFN1 mutations clustered in the region required for mGlu 7 recruitment have been found in patients with epilepsy and ADHD (Dolan and Mitchell, 2013;Tomioka et al., 2014), and ELFN1 knockout (Elfn1 −/− ) animals exhibit a similar seizure phenotype to Grm7 −/− animals (Tomioka et al., 2014). Interestingly, Elfn1 −/− mice also exhibit ADHD-like phenotypes such as hyperactivity and impulsivity. Dolan and Mitchell (2013) showed that Elfn1 −/− animals display hyperlocomotion and increased activity in an open field. Administration of amphetamine to Elfn1 −/− mice was able to attenuate hyperlocomotion, similar to the effects of stimulant therapies for ADHD patients. Tomioka et al. (2014) also demonstrated that Elfn1 −/− mice displayed more spontaneous activity than wild-type animals and also exhibited decreased immobility in a forced swim test, which are behaviors suggestive of hyperactivity. Elfn1 −/− mice spent more time in the open arms during an EPM task compared to wildtype littermates. These data are typically indicative of anxiolytic effects; however, Elfn1 −/− mice showed no preferences between the light and dark boxes of the light-dark box transition task. Based on this finding, the authors hypothesized that the results of the EPM were indicative of impulsivity. Together, these data suggest a role for the ELFN1-mGlu 7 complex in seizures and in other disorders.

MOOD DISORDERS
mGlu 7 modulation has also been demonstrated to impact behavioral models of mood disorders such as anxiety or depression, which are common comorbidities seen in NDDs (Matson and Cervantes, 2014). The amygdala and hippocampus, areas of high mGlu 7 expression, are brain regions known for their importance in anti-anxiety and anti-depressive action (Shin and Liberzon, 2010). In comparison to cognitive tasks, where reductions in mGlu 7 cause deficits, the loss of mGlu 7 has been reported to result in anti-depressive and anxiolytic effects in these domains. For example, Cryan et al. (2003) showed that Grm7 −/− animals spend more time in the open arms than their wild-type counterparts in an EPM paradigm, demonstrating that the loss of the receptor causes anxiolytic activity. In a light-dark box task, the knockout animals have a reduced latency to enter a covered, dark compartment as well as an increased number of transitions into an open, brightly lit compartment than wildtype mice (Cryan et al., 2003). Callaerts-Vegh et al. (2006) demonstrated that Grm7 −/− mice bury fewer marbles than wildtype animals in a marble burying task, which also measures anxiety-like behavior in rodents. ADX71743, the mGlu 7 -selective NAM, causes similar results in EPM, and reduces marble burying in wild-type mice (Kalinichev et al., 2013). Administration of the NAM MMPIP also reduces marble burying, consistent with the Grm7 −/− phenotype (Palazzo et al., 2015). In tail suspension or forced swim tasks, where immobility is indicative of depressionlike behavior, Grm7 −/− mice are less immobile than wild-type animals (Cryan et al., 2003). In wild-type mice, the antagonist XAP044 also increases time in open arms in EPM and decreases immobility in tail suspension, recapitulating data from studies using knockout animals (Gee et al., 2014). In a mouse model of neuropathic pain, the NAM MMPIP also reduces immobility time during tail suspension (Palazzo et al., 2015). The mGlu 7 agonist AMN082 reduces immobility in tail suspension and forced swim tasks, and MMPIP can block the effect of AMN082 (O'Connor and Cryan, 2013; Pałucha-Poniewiera and Pilc, 2013).
In summary, mGlu 7 has been implicated in a range of behaviors in rodent models, many of which mimic those reported in rodent models of NDDs.

GENETIC ASSOCIATIONS BETWEEN mGlu 7 AND NDDs
Genetic associations between NDDs and GRM7, the gene that encodes mGlu 7 in humans, provide a link between experiments in rodent models and the clinical population. ASD affects as much as 1% of the world's population (Lai et al., 2014), and family studies have suggested that the heritability of ASD is about 83% (Sandin et al., 2017), which indicates a strong genetic component. Heterozygous deletions in GRM7 have been identified in three ASD patients by Gai et al. (2012), and in one patient by Liu et al. (2015). The latter patient exhibited language and cognitive impairments as well as hyperactivity, stereotyped behaviors, and deficits in social interaction (Liu et al., 2015). An additional ASD patient with a de novo point mutation in GRM7, resulting in a change from arginine to glutamate at amino acid 622, was reported by Sanders et al. (2012). This mutation affects the third transmembrane portion of the receptor. Yang and Pan (2013) identified the SNPs rs6782011 and rs779867, which encode a C to T change in intron 6 and a T to C or T to G change in intron 5 in GRM7, respectively. These two polymorphisms exhibited significant associations with ASD from a group of 22 ASD patients (Yang and Pan, 2013). In an Iranian cohort of 518 ASD patients, however, only rs779867 was identified as a SNP that associates GRM7 with ASD (Noroozi et al., 2016). rs779867 is a T to C or T to G polymorphism in intron 5 hypothesized to have effects on a MRG protein binding motif. MRG motifbinding proteins are thought to bind chromatin and function in the regulation of gene transcription .
Attention deficit hyperactivity disorder is characterized by inattention, hyperactivity and impulsivity (American Psychiatric Association, 2013). Its estimated prevalence around the world is 7.2% in children and 3.4% in adults (Fayyad et al., 2007;Thomas et al., 2015). A genome wide copy-number variation study revealed that rs7623055, which encodes a G to C or G to T change, was significantly associated with ADHD, and also identified six different deletions in GRM7 in patients with ADHD (Elia et al., 2011). Additionally, rs37952452 was found to have some association with ADHD in a study of 202 patients in Korea, though it was not significantly associated when using a case-control approach (Park et al., 2013). In contrast, neither rs37952452 nor rs7623055 were found to be significantly associated with ADHD in a later study (Akutagava-Martins et al., 2014). Interestingly, ADHD patients with the G/A genotype of rs37952452 showed an improved response to methylphenidate in comparison to those with the G/G genotype (Park et al., 2014).
Rare mutations in GRM7 have also been implicated in undiagnosed NDDs. Whole-exome sequencing in 31 consanguineous Arab families with developmental delay and/or intellectual disability revealed two families with mutations in GRM7. Two brothers in the same family were homozygous for a 461T/C variant, which results in the missense mutation I154T in the ligand binding domain of mGlu 7 . The same study also identified two siblings (brother and sister) who are compound heterozygous for the mutations 1972C/T and 2024C/A, which lead to missense mutations, R658W and T675Y, respectively, in the third transmembrane domain. These four patients share symptoms that include developmental delay, ID, brain malformations and seizures (Charng et al., 2016). In a different set of consanguineous families, exome sequencing identified two female cousins with the homozygous mutation 1757G/A, which results in a premature truncation of mGlu 7 prior to its first transmembrane domain (W568 * ). These patients exhibit seizures, profound ID, microcephaly and leukodystrophy (Reuter et al., 2017). A search of the DECIPHER database (Firth et al., 2009) identified 69 patients with a deletion or duplication that included GRM7, although most of these also affected other genes. Three of these patients had a deletion or duplication restricted to the GRM7 gene and their phenotypes are included in Table 2. mGlu 7 IN MECP2-RELATED DISORDERS Preclinical research in the NDD field has focused largely on mouse models of genetic syndromes due to their high construct validity. RTT is a monogenetic disorder in which mGlu 7 has recently gained particular interest as a potential therapeutic target . RTT is a debilitating NDD affecting 1 in 20,000 births and is characterized by a period of normal development followed by sudden developmental regression and loss of acquired skills at 6 to 18 months of age. Following regression, RTT patients are burdened by lifelong symptoms that include repetitive hand clasping, limited speech, intellectual disability, motor impairment, apneas, and epilepsy (Neul et al., 2010). The majority of RTT cases can be attributed to loss-of-function mutations in the X-linked gene MECP2, which encodes the transcriptional regulator methyl-CpG binding protein 2 (MeCP2) (Amir et al., 1999). Since this discovery, nearly two decades of research have yielded significant insight into the functions of MeCP2 within the brain. Of note, MECP2 mutations have also been identified in patients with ASD and ID independent of a RTT diagnosis (Couvert et al., 2001;Carney et al., 2003), suggesting that pathways involving MeCP2 may underlie NDDs more broadly. MeCP2 is canonically thought to repress gene transcription through binding to methylated CpG dinucleotides and recruiting repressor complexes; however, MeCP2 has also been shown to activate gene transcription and play roles in long-range regulation of chromatin structure, mRNA splicing and micro-RNA processing (Guy et al., 2011). Although MeCP2 is involved in prenatal and postnatal development (Tate et al., 1996; Shahbazian et al., 2002;Bedogni et al., 2016), phenotypes of Mecp2 knockout mice can be reversed if Mecp2 expression is reintroduced in adult animals (Guy et al., 2007). Similarly, ablation of Mecp2 expression in adult mice following normal development is sufficient to recapitulate the phenotype of constitutive Mecp2 knockout mice (McGraw et al., 2011). MeCP2 is thus critical for proper neuronal function throughout life and there exists a therapeutic window to improve disease severity, even at adult stages. These proof-of-concept studies have fueled programs to develop MECP2 replacement strategies, along with parallel efforts to identify targets downstream of MeCP2 dysfunction that may be amenable to pharmacological manipulation. mGlu 7 is one of three mGlu receptors found to be decreased at the mRNA level in a RTT mouse model (Bedogni et al., 2016). These mGlu receptors represent a potential point of access to normalize synaptic function in RTT. Consistent with this initial report, we have shown that mGlu 7 protein expression is significantly decreased in motor cortex autopsy samples from RTT patients compared to those of controls matched for age, sex, and postmortem interval . In global Mecp2 knockout mice, mGlu 7 protein expression is decreased in a brainregion specific manner with a notable reduction in hippocampal synaptosomal fractions. This correlates with reduced depression of synaptic transmission at SC-CA1 synapses by LSP4-2022 in slices from RTT model mice, which can be restored by a PAM. Additionally, pre-application of two structurally distinct Group III mGlu receptor PAMs, VU0422288 and VU0155094, to slices was able to restore deficient LTP at SC-CA1 synapses in RTT model mice. Ablation of Mecp2 selectively from GABAergic neurons is sufficient for LTP impairment (Chao et al., 2010); therefore, rescue of LTP by mGlu 7 potentiation is consistent with the proposed model by which mGlu 7 -mediated inhibition of GABA release is required for LTP at SC-CA1 synapses (Klar et al., 2015).
At the behavioral level, mGlu 7 potentiation by intraperitoneal administration of the brain penetrant PAM, VU0422288, is able to improve performance in assays of cognition in RTT model mice . While many studies in Grm7 −/− mice have implicated a role for mGlu 7 in learning and memory (Hölscher et al., 2005;Callaerts-Vegh et al., 2006;Goddyn et al., 2008), this is the first report of mGlu 7 activity being modulated in a positive direction to reverse a deficit in cognition. VU0422288 is also able to increase performance in a social novelty task and reduce the number of apneas detected by whole body plethysmography . These data suggest that mGlu 7 potentiation may be a valid approach to address multiple RTT-associated symptom domains. It is important to note that these experiments used mice with a global deletion of Mecp2. As RTT is most commonly caused by MECP2 point mutations in humans, it will be important to elucidate the effect of various point mutations on mGlu 7 expression/function to identify patient subpopulations that would be predicted to benefit from an mGlu 7 PAM. mGlu 7 has also been investigated for its therapeutic utility in a mouse of MDS. In contrast to RTT, MDS occurs when the region of the X chromosome containing MECP2 is duplicated or triplicated, and is predicted to account for 1% of cases of unexplained X-linked intellectual disability (Lugtenberg et al., 2009). MDS patients present with infantile hypotonia, autismassociated symptoms, speech impairment, respiratory infections, and epilepsy (Ramocki et al., 2010). This disorder highlights the point that precise regulation of MeCP2 expression is required for normal brain function and that excess MeCP2 protein is detrimental. Fisher et al. (2017) tested whether mGlu 7 protein levels are affected in MeCP2-Tg1 mice, a model for MDS. Contrary to a hypothesis of bidirectional regulation, mGlu 7 protein levels are unchanged in most brain regions in MeCP2-Tg1 mice. Furthermore, neither genetic reduction of mGlu 7 protein levels or administration of the mGlu 7 NAM ADX71743 had any impact on anxiety and fear learning phenotypes in MeCP2-Tg1 mice . These findings suggest that that mGlu 7 expression/function may only be affected by MeCP2 hypofunction and not overexpression. More studies are warranted to understand the molecular interaction between MeCP2 and mGlu 7 expression. This information will inform future drug development of mGlu 7 PAMs for RTT and other NDDs in which MECP2 mutations have been identified.

CONCLUSION
Neurodevelopmental disorders are a prevalent group of disorders with limited treatment options and mGlu 7 represents one potential access point for pharmacological intervention. GRM7 gene disruptions identified in patients with NDDs provide clinical rationale for this approach. Pre-clinical studies in rodent models suggest that decreased mGlu 7 function is sufficient to mimic phenotypes that correlate to NDD symptom domains and that positive modulation of mGlu 7 activity can improve some deficits, specifically in a mouse model of RTT. However, NDDs are highly heterogeneous and are likely the result of unique molecular pathologies that converge to produce similar circuit and behavioral phenotypes. Therefore, further studies are needed to identify and understand which subpopulations may benefit from an mGlu 7 -mediated therapy. In parallel, further development of improved tool compounds will facilitate studies focused on understanding mGlu 7 receptor function in brain circuits and behaviors associated with NDDs.

AUTHOR CONTRIBUTIONS
NF and MS equally contributed to the first draft of this review. All authors read and edited the manuscript prior to submission. FUNDING NF was supported by NIH training grants T32 GM007628 and F31 MH113259. MS was supported by T32 GM007628. CL received funding from William K. Warren, Jr. and the William K. Warren Foundation and is supported by the William K. Warren, Jr. Chair in Medicine. CN received funding from the National Institutes of Mental Health (R21MH102548 and R01MH113543; the latter jointly with CL), the Department of Defense Congressionally Directed Medical Research Program (W81XWH-17-1-0266), and rettsyndrome.org. This review makes use of data generated by the DECIPHER community.
A full list of centers who contributed to the generation of the data is available from http://decipher.sanger.ac.uk and via email from decipher@sanger.ac.uk. Funding for the project was provided by the Wellcome Trust.