Control of cortical synapse development and plasticity by MET receptor tyrosine kinase, a genetic risk factor for autism

Abstract

The key developmental milestone events of the human brain, such as neurogenesis, synapse formation, maturation, and plasticity are determined by a myriad of molecular signaling events, including those mediated by a number of receptor tyrosine kinases and their cognate ligands. Aberrant or mistimed brain development and plasticity can lead to maladaptive changes, such as dysregulated synaptic connectivity and breakdown of circuit functions necessary for cognition and adaptive behaviors, which are hypothesized pathophysiologies of many neurodevelopmental and neuropsychiatric disorders. Here we review recent literature that supports autism spectrum disorder (ASD) as a likely result of aberrant synapse development due to mistimed maturation and plasticity. We focus on MET receptor tyrosine kinase, a prominent genetic risk factor for autism, and discuss how a pleiotropic molecular signaling system engaged by MET exemplifies a genetic program that controls cortical circuit development and plasticity by modulating anatomical and functional connectivity of cortical circuits, thus conferring genetic risk for neurodevelopmental disorders.

1. Disrupted synapse development and plasticity as a core pathophysiology of NDDs.

Dysregulated synapse development is posited to underlie the pathophysiology of many neurodevelopmental and neuropsychiatric disorders, including autism spectrum disorder (ASD), schizophrenia and major depression. These complex brain disorders all share a major developmental etiology, and a common feature that disease onset and progression occurs during early brain development and maturation (Geschwind and Levitt, 2007; Abrahams and Geschwind, 2008; Walsh et al., 2008; Zoghbi and Bear, 2012). There have been no unifying neuropathological or neurobiological features that define these disorders, yet it is commonly agreed that genetics, the environment, and interactions between the two play a large role in the pathogenesis of these disorders. Autism, for which diagnosis is solely based on behavior, manifests a triad of behavioral anomalies including impaired communicative (both verbal and non-verbal) skills, deficits in social skills, restricted interests and repetitive behaviors. ASD is a complex disorder, and as such, identification of causative genes has been hampered by many inherent problems, such as multiple gene effects/interactions, variable penetrance for each individual gene, genetic heterogeneity, environmental factors, and gene-environment interactions. Yet autism has evidently the strongest genetic components of all the developmental neuropsychiatric disorders; there is an 82-92% concordance rate for autism among monozygotic twins as compared with the ~10% concordance rate for dizygotic twins (Bailey et al., 1995; Abrahams and Geschwind, 2008; Constantino et al., 2013).

To date, an enormous number (> 1000) of gene loci have been documented to contribute to the risk of developing ASD (Sebat et al., 2007; Piggot et al., 2009; Pinto et al., 2010; Aldinger et al., 2011; Abrahams et al., 2013; Ebert and Greenberg, 2013). This genetic heterogeneity includes de novo mutations that produce syndromic forms of autism (e.g. FMR1 - Fragile X syndrome (Santoro et al., 2012), MECP2 - Rett syndrome (RTT,Ip et al., 2018)), non-coding variations (e.g. MET gene (Campbell et al., 2006)), copy number variations, and chromosome abnormalities (e.g. , duplication of UBE3A - Angelman syndrome with autism traits (Glessner et al., 2009); deletion of 16p11.2 in autism (Kumar et al., 2008; Weiss et al., 2008)). Many autism risk genes encode proteins that are enriched at the excitatory synapse, a highly specialized junctional structure that contains at least >1500 different proteins (Coba, 2019; Koopmans et al., 2019). These risk gene-encoded proteins may scaffold pre- and postsynaptic assembly, control neurotransmitter release, gate synaptic receptor trafficking, and affect activity-dependent circuit connectivity (Zoghbi and Bear, 2012). In addition, autism risk genes have broader functions spanning neuronal growth, cytoskeletal organization, cell projection and motility, synaptic signaling, and activity-dependent plasticity (Levitt and Campbell, 2009; Pinto et al., 2010), all of which are critical to shaping neuronal connectivity and guiding behavior. Unsurprisingly, disruption of ASD risk genes has been shown to lead to synaptic dysfunction, altered circuit connectivity, and aberrant excitatory/inhibitory circuit networks (Rubenstein and Merzenich, 2003; Tabuchi et al., 2007; Ebert and Greenberg, 2013). Therefore, the role of specific gene products needs to be studied in the context of developmental timing, specific neural circuits and brain regions, and molecular and cellular substrates. The two cardinal features of ASD, heritability and heterogeneity, while imposing a major challenge for the identification of causative genes, also offer unprecedented opportunity to gain mechanistic insights into disease pathophysiology.

2. The human MET gene emerges as a autism genetic risk factor.

The human MET gene (OMIM 164860; on chromosome 7q31) was first reported as a risk factor for autism by Campbell et al. (2006) in a family-based, candidate gene-approach study. It was revealed that variation in the 5’ transcriptional regulatory region of the MET gene confers differential risk to autism; specifically, the rs1858830 “C” allele is over-represented in autism cases. Compared with the GG genotype, the relative risk (RR) for autism diagnosis was 2.27 (95% CI 1.41–3.65) for the CC genotype and 1.67 (95% CI 1.11-2.49) for the CG type. In multiplex families with more than one autistic proband, the rs1858830 “C” allelic association is even stronger. This promoter variation is functional in that it affects binding for specific transcription factors (SP1, PC4 and MeCP2) (Campbell et al., 2006; Plummer et al., 2013) and regulates MET gene transcription in vitro in cell lines. Following this initial report, a subsequent study examining MET expression levels in the postmortem brain tissues of autism and control cases found decreased MET transcripts and protein levels in ASD individuals compared to matched controls (Campbell et al., 2007), further supporting that reduced or hypomorphic MET signaling may contribute to the genetic risk.

Complementary to these original reports, additional genetic and pathophysiological evidence further support that dysfunctional MET signaling contributes to ASD risk; the rs1858830 “C” variant was later found to be associated with autism risk in a different cohort (Jackson et al., 2009). Other independent family-based association analyses also found a significant association between a single nucleotide polymorphism (rs38845 and rs38841) in MET gene intron 1 and autism in different ASD cohorts including both Japanese and Caucasian populations (Sousa et al., 2009; Thanseem et al., 2010). The strong association of MET with autism risk is further strengthened by the findings that genes in the MET pathway (Campbell et al., 2009), such as HGF, which encodes the only known ligand for MET (Naldini et al., 1991b), are associated with autism (Russo et al., 2009; Russo, 2014). In addition, case-control and family-based association analysis revealed that PLAUR gene is significantly altered in the ASD brain. The PLAUR promoter variant rs344781 T allele was found associated with ASD (Campbell et al., 2008). The human PLAUR gene encodes a urokinase plasminogen activator receptor (uPAR), which is required to process the HGF precursor into an active form (Figure 1). The SERPINE1 gene, which encodes plasminogen activator inhibitor-1, was also found significantly associated with autism (Campbell et al., 2008). Further evidence exists that supports MET as a candidate risk gene. First, MET is located on human chromosome 7q31, under a linkage peak identified in multiple whole genome scan studies of autism (IMGSAC, 1998, 2001; Yonan et al., 2003). Second, MET signaling is known to engage pleiotropic cellular events, including neural specification, invasive growth, neurite extension, dendritic spine morphogenesis, and brain circuit connectivity, which are critical histogenic events in brain development (Maina et al., 1997; Powell et al., 2001; Ieraci et al., 2002; Gutierrez et al., 2004; Qiu et al., 2011; Ma et al., 2019). Hypomorphic MET signaling is also known to affect interneuron migration and proliferation of granule cells in the cerebellum (Ieraci et al., 2002; Powell et al., 2003; Eagleson et al., 2005). Third, MET protein expression is also reduced in Rett syndrome brains; MeCP2 binds to a region of the MET promoter containing the ASD-risk ‘rs1858830C’ in human neural progenitor cells, and enhances MET expression (Plummer et al., 2013). Further, MET transcription is attenuated by RTT-specific mutations in MeCP2. In female RTT patients, the postmortem temporal cortex show dramatically reduced MET gene expression. Therefore, MeCP2 seems capable of regulating MET expression in allele-specific manner. Lastly, MET signaling also plays a role in gastrointestinal repair and immune function (Beilmann et al., 2000; Tahara et al., 2003; Arthur et al., 2004; Ido et al., 2005; Okunishi et al., 2005), disruption of which are often co-morbid conditions in ASD cases. Taken together, these findings further support that MET signaling pathway may be centrally positioned in the genetic and molecular networks that contribute to ASD genetic risks.

Figure 1. MET receptor tyrosine kinase engages intracellular molecular signaling pathways and controls neuronal morphological development.

Human MET protein is proteolytically processed to yield an extracellular α-subunit (50 kD) and a transmembrane β-subunit (145 kD) that are linked together by a disulfide bond. The beta subunit contains extracellular, transmembrane and intracellular signaling domains. The extracellular domain in both subunits contain a SEMA domain, the beta chain has a MET-related sequence (MRS), glycine-proline rich PSI domain, and four immunoglobulin-like domains (Ig domain). The intracellular beta subunits contain tyrosine kinase domain and binding sites. Upon activation by its sole ligand HGF, which is proteolytically processed by urokinase-type plasminogen activtor (uPA) into active form from pro-HGF, MET RTK dimerizes, and the Tyr¹²³⁴ and Tyr¹²³⁵ within the tyrosine kinase domain are phosphorylated. This activates the intrinsic kinase activity of the receptor. Two phosphorylated tyrosine residuals (Tyr ¹³⁴⁹ and Tyr ¹³⁵⁶) located in the multi substrate docking sites near the c-terminus recruit adaptor proteins such as Src, SHC, Gab1, Grb2, or PI3K. For the purpose of clarity, this figure only depicts signaling pathways relevant to dendrite and spine development, which include potential activation of Ras > Erk1/2; PI3K/AKT/mTOR pathway that promotes dendritic morphogenesis. In addition, small GTPase cdc42 and rac1 activation likely engage LIMK1/cofilin to control spine morphogenesis, actin dynamics and structural plasticity of the developing synapse. Note other potential molecular signaling events (??) are not depicted. Disrupted MET signaling, such as the hypomorphic ‘low activity’ rs1858830 ‘C’ allele, in developing neurons may collectively affect a large number of developmental events leading to increased risk for autism.

In addition to these genetic findings, human imaging studies also suggest functional MET variants can impact brain circuit connectivity. Rudie et al. (Rudie et al., 2012) reported that MET impacts structural and functional networks in the human brain by examining the functional ASD risk variant (rs1858830 “CC”) on cortical network functions in ASD and control subjects. The relationship between MET risk genotype and functional activation patterns to social stimuli (e.g. emotional faces) was revealed by functional magnetic resonant imaging (fMRI) (Rudie et al., 2012). The study found that the risk “CC” allele was capable of predicting atypical fMRI activation and deactivation patterns in response to social stimuli, and was highly correlated with reduced functional connectivity in the temporoparietal junctions, areas known to have high levels of MET expression in humans and non-human primates (Judson et al., 2011; Mukamel et al., 2011). The rs1858830 “C” risk allele also exhibits the largest alterations in other ASD-related endophenotypes; for instance, the rs1858830 ‘C’ allele modulates the severity of social impairment in ASD, such that ASD individuals carrying this risk allele have more severe social deficits phenotypes than those who do not (Campbell et al., 2010). In addition, the rs1858830 “C” risk allele also modulates binding of MeCP2 to the MET promoter, and MET expression is severely depressed in the postmortem temporal cortex of RTT syndrome brains (Plummer et al., 2013). Taken together, these genetic, neurobiological, and clinical findings establish MET as a prominent candidate risk gene for ASD.

3. MET signaling and its pleiotropic nature.

The developing brain is controlled by a myriad of molecules including a large family of growth factors and their receptors (for a review, see Furge et al., 2000). Receptor tyrosine kinases (RTKs) are cell surface receptors which bind to polypeptide growth factors, hormones, cytokines (Robinson et al., 2000), and elicit responses with broad roles in developmental and postnatal physiological processes. RTKs are known to regulate cell survival, differentiation, neurogenesis and migration, synaptic connectivity, maturation and plasticity (Peng et al., 2013). RTKs consist of an extracellular ligand-binding domain, a single transmembrane domain, and an intracellular domain containing a catalytic tyrosine kinase motif that can be either autophosphorylated or transphosphorylated by various interacting proteins. Upon activation, RTKs bind to signaling molecules and recruit adaptor/effector proteins to initiate downstream cellular responses through various signaling pathways (Park and Poo, 2013).

MET RTK is known to play a pleiotropic role in cell proliferation, differentiation and survival, motogenesis and morphogenesis in many tissue types (Maina et al., 1998; Birchmeier et al., 2003). MET was first identified as a proto-oncogene (Cooper et al., 1984; Bottaro et al., 1991). The only known ligand for MET receptor, hepatocyte growth factor (HGF, also known as scatter factor), is a polypeptide growth factor (Naldini et al., 1991a). Both MET and HGF are critical for embryonic development of mesenchymal-derived organs, including muscle, liver and placenta. As such, genetic knockout of Met or Hgf in mice leads to early embryonic lethality, resulting from liver malformation, parenchymal cell loss, and disruption of placenta trophoblast cells and muscles (Bladt et al., 1995; Schmidt et al., 1995; Huh et al., 2004).

Human MET protein is produced as a ~170 kD single-chain precursor (Cooper et al., 1984; Faletto et al., 1992), which is then proteolytically processed to yield a highly glycosylated extracellular α-subunit (50kD) and a transmembrane β-subunit (145kD) (Tempest et al., 1988; Furge et al., 2000) (see Figure 1). Both subunits are linked together by a disulfide bond. The beta subunit shares a common topology with many other RTKs by including extracellular, transmembrane and intracellular signaling domains. The extracellular domain in both subunits contain a SEMA (homologous to semaphorins) domain. The beta chain has a MET-related sequence (MRS), glycine-proline rich repeats, and four immunoglobulin-like (Ig) domains. The intracellular beta subunit contains a tyrosine kinase domain and multi substrate docking sites whose function depends on several critical tyrosine residues. Upon activation by HGF, human MET dimerizes - Tyr¹²³⁴ and Tyr¹²³⁵ within the activation loop of the tyrosine kinase domain are transphosphorylated. This activates the intrinsic kinase activity of the receptor (Naldini et al., 1991a). In addition, two phosphorylated tyrosine residuals (Tyr ¹³⁴⁹, Tyr ¹³⁵⁶) located in the multi substrate docking sites near the c-terminus, recruit downstream Src homology-2 (SH2)-domain containing adaptor proteins (Ponzetto et al., 1994) such as Src, SHC, Grb2, or PI3K to interact with the multisubstrate docking site. Additionally, many biological effects of MET activation are mediated through the large scaffolding protein Gab1, which is tyrosine phosphorylated to recruit many binding effector proteins including SHP2, PI3K, and PLC-γ (Faletto et al., 1992; Gual et al., 2000). Both adaptor proteins Grb2 and Gab1 mediate the majority of the MET signaling outcomes during ontogenic events for many organs, a process involving pleiotropic activation downstream pathways including Ras, Src kinases, Rho family GTPases, ERK/MAPK, PI3K/AKT/mTOR, STAT3, JNK/SAP kinases, and phospholipid pathways (Ponzetto et al., 1994; Maina et al., 1996; Maina et al., 2001). Collectively, MET signaling mediates a diverse array of cellular events, including cell fate specification, proliferation, cell cycle progression, polarity and morphology, actin cytoskeleton reorganization and motility, angiogenesis and organ regeneration (Takaishi et al., 1994; Royal et al., 2000; Tahara et al., 2003; Arthur et al., 2004; Ido et al., 2005), tumor invasion (Birchmeier et al., 2003) and immune responses (Beilmann et al., 2000; Okunishi et al., 2005; Roccisana et al., 2005).

4. MET RTK in neuronal development, insight from basic neurobiology.

Although MET function is best understood in cancer biology, MET and HGF are also expressed in the developing nervous system of a number of mammalian species including rodents (Achim et al., 1997; Maina et al., 1997; Maina et al., 1998; Judson et al., 2009; Wu and Levitt, 2013), monkeys (Judson et al., 2011), and humans (Mukamel et al., 2011; Hamasaki et al., 2014), where they influence many neurodevelopmental events. It is currently unclear to what extent these signaling events are operating in neurons during brain development. Yet, studies have shown that MET signaling is required for neuronal lineage commitment (Streit et al., 1995), suggesting MET/HGF plays a role during the early steps of neural induction. In addition, postnatal proliferation of cerebellar granule neurons were shown to require full level of HGF/MET signaling; MET is expressed in granule cell precursors, which respond to HGF with proliferation. A Grb2-binding incompetent, hypomorphic Met mutant results in reduced granule cell proliferation, foliation defects and smaller size of the cerebellum (Ieraci et al., 2002). HGF has also been shown to act as a motogen and guidance signal for gonadotropin hormone-releasing hormone-1 neuron migration (Giacobini et al., 2007), an effect that requires molecular cross talk between MET and the AXL receptor tyrosine kinase (Salian-Mehta et al., 2013). HGF/MET signaling has been shown to modulate migration of a few specialized neuron types, including trans-telencephalic migration of interneurons during forebrain development (Powell et al., 2001), and migration of olfactory interneuron precursors in the rostral migratory stream (Garzotto et al., 2008). These processes seem to depend on MET-Grb2 coupling, and the ensuing signaling through ERK, Rac1/p38, and PI3K/AKT (Segarra et al., 2006). HGF also increases the numbers of neurites of peripheral sensory neurons (Maina et al., 1998), cooperates with nerve growth factor (NGF) to facilitate sympathetic neuron axonal outgrowth, and promotes sensory neuron target innervations (Maina et al., 1997). HGF is also reported as an axonal chemoattractant and a neurotrophic factor for spinal motor neurons (Ebens et al., 1996), and a chemoattractant for developing cranial motor axons (Caton et al., 2000). Taken together, the pleiotropic signaling capacity of MET suggests broad functional significance across the protracted neurodevelopmental timeline, and is consistent with the computationally-hypothesized enrichment of ASD risk genes related to growth factor signaling (Wittkowski et al., 2014).

To gain better insights on the role of MET/HGF signaling, it is important to ascertain the normal spatiotemporal patterns of MET/HGF expression levels in the developing brain (preferably in multiple species). This would be informative to understanding the biological role of dynamic MET signaling at cellular, circuit and system levels. An early study showed that MET protein is detectable in human brain tissues using Western blot, and immunohistochemistry (IHC) staining of MET revealed extensive labeling of both grey and white matter, including microglial cell components (Di Renzo et al., 1993). Another earlier study (Jung et al., 1994) using in situ hybridization demonstrated that in the developing and adult mouse brain both HGF and MET transcripts are expressed, and HGF mRNA is primarily localized in neurons of the cortex and hippocampus, as well as in cerebellum granule cells, while MET mRNA is more abundantly distributed to the hippocampus CA1, the septum, and the cortex. In addition, both HGF and MET mRNA transcripts can be detected as early as embryonic days 12 and 13. Judson et al. (2009) systematically investigated the expression pattern of Met transcripts and proteins using in situ hybridization, Western blot, and IHC labeling. The study revealed that Met mRNA expression in the forebrains of mice lasts throughout late embryonic and postnatal development (embryonic day E17.5 to postnatal day P35), and the expression of MET protein was temporally regulated across multiple forebrain structures. Peak MET protein level occurs around postnatal day (P) 7, being relatively stable during the second postnatal week, but is precipitously down-regulated after P21 to very low levels at adult stage. This transient expression pattern was confirmed in the developing cortex and hippocampus by a recent study (Peng et al., 2016). Therefore, peak levels of MET expression coincide with rapid neurite outgrowth and synaptogenesis, but is downregulated prior to synapse maturation, circuit refinement and heightened plasticity. The study by Judson et al. revealed that MET is expressed by discrete subtypes of long-projecting neurons of the forebrain (at P7-14), both at the neuropil and more strongly distributed to axons throughout the anteroposterior cortical axis, with the corpus callosum evidently possessing the highest level of MET expression. MET expression also follows distinct laminar patterns in sensory cortices, in that the layer IV somatosensory cortex distinctly shows no MET staining due to lack of MET expression in the subcortical structures, including the thalamus and the dorsal striatum (Judson et al., 2009). Another study (Eagleson et al., 2013) employing biochemical, ultrastructural, in situ proximity ligation assay (PLA) approaches showed that MET is enriched at synapses during development in both presynaptic and post synaptic compartments, but synaptic distribution and brain region-specificity is dynamically regulated by developing age.

Although MET is predominantly expressed in dorsal pallium structures, ascertaining the heterogeneity of MET⁺ neurons was facilitated by a recent study (Kast et al., 2019), in which a MET^GFP transgenic mouse line, constructed using the bacterial artificial chromosome (BAC) approach, allows co-expression analysis of class-specific molecular markers and connectivity. This study revealed that MET is expressed by a subset of sub-cerebral and a larger number of intra-telencephalic projection neurons, while is excluded from most layer (L) 6 corticothalamic neurons. This work suggests anatomically defined cortical projection classes can be further subdivided based on their molecular profiles that likely influence synaptic maturation and circuit wiring. These findings on MET expression in cortical projection neurons are consistent with the reported coalescence of functional ASD gene networks in circuits involving both superficial and deep layer glutamatergic projection neurons (Parikshak et al., 2013; Willsey et al., 2013). In addition to dorsal pallium projection neurons, distinct brainstem circuitry also strongly expresses Met and/or Hgf. Utilizing in situ hybridization and IHC staining, Wu et al. (Wu and Levitt, 2013) reported a highly specific pattern of MET expression in a subpopulation of brainstem neurons in cranial motor nuclei, the dorsal raphe, Barrington’s nucleus, and the nucleus of solitary tract. Taking advantage of the newly developed MET^GFP mice, Kast et al. reported that MET is specifically expressed in a subset of 5-HT neurons within the caudal part of the dorsal raphe nuclei (DRC). MET expression is almost exclusively in the DRC as a condensed nucleus, while other DR subdivisions contain few or sparse set of MET⁺ neurons. These MET⁺ DRC 5-HT neurons provide serotonergic inputs to the ventricular/subventricular region, medial and lateral septum, and the ventral hippocampus, but do not innervate the dorsal hippocampus or entorhinal cortex. Yet, how MET expression affects the development and function of these 5-HT neurons is currently unknown. One recent study (Okaty et al., 2015) reported that loss of Met in 5-HT neurons in mice (Met^fx/fx::ePet1^cre) results in impaired social interest, which potentially reflects autism conditions. MET⁺ 5-HT neurons may influence certain aspects of complex behaviors, which can be mapped to specific subtypes of 5-HT neurons. This study, however, did not address how loss of MET impacts 5-HT neuron functions or their projections. Considering the growing evidence linking 5-HT neurons with autism (Dougherty et al., 2013; Muller et al., 2016; Walsh et al., 2018), it would be interesting to examine how loss of MET affects the molecular, cellular and circuit functions of these neurons and how that is related to more broad autism endophenotypes.

Another recent study (Kamitakahara et al., 2017) found that a subpopulation of neurons in the dorsal motor nucleus of vagus and nucleus ambiguus expresses MET. This subpopulation of neurons innervates a broad area of the gastrointestinal system. MET⁺ neurons were also found in vagal motor nuclei in the brainstem of non-human primates. These data indicate brainstem MET⁺ neurons are anatomically positioned to regulate structures implicated in ASD pathophysiology and to control automatic function. Given gastrointestinal disturbances are common in children with ASD (Coury et al., 2012; Chaidez et al., 2014), this study suggests that both neurological and gastrointestinal functions potentially share similar biological vulnerability. This hypothesis is also supported by the shared genetic burden for psychiatric and co-occurring medical conditions frequently affecting both the brain and the gut organs (Plummer et al., 2016).

5. Control of glutamatergic synapse development by MET signaling.

The conserved temporal pattern of MET protein expression during early cortical development indicates MET signaling may be ideally positioned to regulate neuronal growth, synaptogenesis, morphology development and circuit connectivity. Because this transient nature of MET expression is also preserved in cultured primary hippocampal neurons (Peng et al., 2016), MET signaling seems to be a highly conserved, intrinsic, cell-autonomous mechanism that regulates key aspects of neural development, including both morphology and function. An early study has shown that HGF treatment in cultured hippocampal neurons leads to activation of MET, and increased number of dendrites and total dendritic length, effects that are mediated by AKT and glycogen synthase kinase 3 beta (GSK-3P) activation (Lim and Walikonis, 2008). Using immunocytochemistry and ultrastructural immunoelectron microscopy approach, the same research group also reported that MET is clustered at excitatory synapses and co-localizes with NMDA receptor subunit GluN2B and PSD-95 protein at a relatively mature stage (Tyndall and Walikonis, 2006). These reports are seemingly at odds with the observation of reduced levels of MET as cortical circuits mature (Judson et al., 2009; Peng et al., 2016), but may be explained by more localized synaptic MET expression in mature cortical circuits (Eagleson et al., 2013). Another study found that exogenous HGF application to cortical organotypic slice cultures increases dendritic growth and branching of pyramidal neurons, an effect that is blocked by HGF antibody or an dominant-negative mutant MET receptor construct (Gutierrez et al., 2004). It is important to note that MET expression is highly dynamic even at the cellular level. Using subcellular fractionation in developing cortical tissues, Peng et al. reported that in cortical tissues of P0 mice, the growth cone-enriched fraction contains elevated levels of MET protein. This subcellular MET distribution was also revealed in cultured hippocampus neurons that express MET during early development (Peng et al., 2016), as these neurons showed enhanced MET immunostaining signal in the early growth cone structure.

The selective MET distribution in early neuronal growth cone seems capable of promoting early neuronal growth. Qiu et al (2014) found that HGF (50 nM) application to cultured primary hippocampal neurons leads to faster growth of both dendrite and axon compartments. In addition, transfecting primary hippocampal neurons with MET cDNA (gain-of-function) or interference RNA (RNAi, loss-of-function) exerts an opposing effect on neuronal morphology that is development stage-dependent. In cultured primary hippocampal neurons, enhancing MET signaling during the second week in culture leads to profuse dendritic protrusions, while reducing MET signaling leads to sparser dendritic protrusions. At later developmental stages, numerous dendritic spines are developed as a result of enhanced MET signaling, while fewer spines exist as a result of MET RNAi. More importantly, the supernumerary spines associated with enhanced MET signaling are generally smaller in size, while the less number of spines associated with RNAi are consistently larger. These findings on spine head size and density in cultured neurons are in general conformity with the findings in vivo, as transfecting the developing embryonic hippocampus CA1 neurons with Met cDNA or RNAi using in utero electroporation yielded similar spine phenotypes. Since dendritic spine head size and geometry are associated with their AMPA receptor content and maturation status (Matsuzaki et al., 2001), these observations on the bi-directional regulatory effects strongly suggest MET signaling may be an intrinsic mechanism controlling many key aspect of glutamatergic synapse development, including spine/synapse number, and the timing of excitatory synapse maturation.

How does pleiotropic MET signaling broadly shape multiple aspects of the excitatory synapse development? Eagleson et al. (2016) reported that HGF-induced MET autophosphorylation peaks during synaptogenesis, with a striking reduction in activation just before synapse pruning (~P17). Further assessing the roles of intracellular signaling revealed that HGF-induced phosphorylation of MET engages ERK1/2 and AKT, which account for both increases in total dendritic growth and synapse density (Figure 1). Remarkably, MAPK/ERK pathway inhibition significantly reduced the HGF-induced increase in dendritic length, while inhibition of the PI3K/AKT pathway selectively reduced HGF-induced increases in synapse density. This study reveals a key role for MET activation and the critical intracellular signaling pathways in shaping distinct biological outcomes of neocortical neuron growth and synaptogenesis. In addition to the MAPK/ERK and AKT pathway, Peng et al. (2016) found that application of HGF to acutely prepared P9-10 cortical slices (utilizing the endogenous molecular signaling components) leads to MET activation, measured by phosphorylation of the MET kinase activity site (Tyr^{1232, 1233} of mouse MET), a process that depends on MET and PI3K kinase activity (inhibited by PHA665752 and wortmannin, respectively). In addition, acute activation of cortical MET leads to Rho family small GTPase activation, measured by the increased level of GTP-bound form of cdc42. Cdc42 activation seems to partially account for the enhanced dendritic growth and spine morphogenesis associated with MET signaling. Considering the well-established role of Rho-family small GTPases (including rac1, cdc42 and rhoA) in mediating neurite growth and spine morphogenesis (Tashiro et al., 2000; Govek et al., 2005; Nadif Kasri et al., 2009; Tolias et al., 2011; Ba et al., 2013), this study established a critical missing molecular link between MET activation and glutamatergic synapse morphological development. Revealing the role of small GTPases also lends support to an ASD genetic study that reported an enrichment of genetic variations that disrupts “functional gene sets involved in cellular proliferation, projection and motility, and GTPase/Ras signaling” (Pinto et al., 2010). At the functional level, perturbing the normal timing of MET signaling by Met cDNA transfection in cultured mouse hippocampus neurons results in altered maturational measures, including reduced amplitude of miniature excitatory synaptic currents (mEPSCs), and reduced functional synapses (measured by the number and proportion of synaptic puncta that contain both presynaptic and postsynaptic proteins [synapsin I/PSD95], or both NMDA and AMPA [GluN1/GluA1] receptor markers). All of this evidence suggests MET has an essential role in not only hardwiring circuits during early histogenic events, but engaging intracellular pathways and physiological processes that converge to regulate glutamatergic synapse development.

6. MET signaling shapes the developmental synaptic and circuit connectivity.

Circuit connectivity-based etiology for autism has gained strong support, with numerous functional and imaging studies documenting alterations in both local and long-range circuit connectivity among different functional regions in autism patients (Just et al., 2004; Geschwind and Levitt, 2007; Just et al., 2007; Kana et al., 2009; Sahyoun et al., 2010; Hong et al., 2011; Shukla et al., 2011). For example, studies by Just et al. (Just et al., 2004; Just et al., 2007) reported less synchronous activity of language processing areas in ASD patients in response to a semantic comprehension task, in which impaired synchrony is mostly observed in frontoparietal areas. Similarly, compromised synchrony was also found during social processing tasks in autism (Kana et al., 2009). ASD brains also show reduced fractional anisotropy (FA) in major long-range fiber tracts in a diffusion tensor imaging (DSI) study (Shukla et al., 2011), alluding to a global deficit in functional connectivity in ASD brains. When presented with face recognition or linguistic reasoning tasks, ASD brains show reduced functional connectivity with frontal cortical regions (Sahyoun et al., 2010). This hypo-functioning in long-range circuits is consistent with existing anatomical evidence, in that reduced corpus callosum volume has been reported in some ASD patients (Just et al., 2007; Hong et al., 2011; Shukla et al., 2011; Thomas et al., 2011). In contrast to a reduction in long range connectivity, local brain area connectivity are frequently reported to be increased (Geschwind and Levitt, 2007; Keown et al., 2013; Lynch et al., 2013; Supekar et al., 2013; Itahashi et al., 2015). The molecular and cellular basis for these altered connectivities are not fully understood, yet animal model studies reveal that disrupted excitation/inhibition (E/I) balance may be an underlying mechanisms. E/I balance in neural circuits is a type of homeostatic synaptic plasticity that is frequently reported disrupted in neurological and psychiatric conditions (Rubenstein and Merzenich, 2003; Nelson and Valakh, 2015; Lo et al., 2016a). Other circuit mechanisms, including developmental wiring of circuits, specificity of connectivity, synaptic strength, molecular constituents at synaptic sites, and activity-dependent plasticity changes may all play an important role.

By controlling the timing of excitatory synapse development, MET signaling may shape brain circuit functional connectivity. As described above, the human MET risk ‘rs8912463C’ allele is capable of modulating responses to social stimuli (Rudie et al., 2012), suggesting social processing circuits may be altered with hypomorphic MET signaling. To approach the question of how MET shapes circuit connectivity, it would be informative to investigate potential synaptic and circuit changes in the absence of MET signaling in the developing forebrain in animal models. Because germline knockout of Hgf or Met in mouse is embryonically lethal (Bladt et al., 1995; Schmidt et al., 1995; Huh et al., 2004), a brain- and cell-type specific conditional knockout of Met would be necessary. This approach has been utilized in previous studies to explore the circuit and behavioral phenotypes associated with Met loss of function. Judson et al. (Judson et al., 2009) crossed two genetically-modified mouse lines to achieve Met conditional knockout (cKO). One of the lines encompasses Met conditional targeting (Met^fx/fx) (Huh et al., 2004), in which Met gene was modified to contain loxP sites flanking exon 16. Exon 16 encodes a peptide containing a critical ATP-binding site (Lys¹¹⁰⁸). When crossed to mice expressing cell-type-specific cre recombinase, exon 16 can be deleted. This deletion inactivates the intracellular tyrosine kinase activity of MET protein, which is essential for its function. Crossing Met^fx/fx mice to a dorsal pallium excitatory neuron-specific cre driver line, such as the Emx1-IRES-cre mice (Gorski et al., 2002), leads to the cKO of Met gene (Met^fx/fx/Emx1^cre) in excitatory neurons in the mouse neocortex. Using this genetic approach, Judson et al. conducted a detailed morphology study in defined neuronal types obtained from Met^fxlfxlEmx1^cre cKO mice and wild-type littermate controls (Judson et al., 2010). The study reported morphological deficits in pyramidal neurons from cortical regions (e.g. anterior cingulate cortex), including a significantly increased (by ~ 20%) spine head volume, a marked reduction in apical dendritic arborization, and a decreased cortical volume that can be sampled by Met^fx/fx/Emx1^cre L5 projection neurons. Another MRI imaging study reported macroscopic changes in cortical structures in cKO mice (Smith et al., 2012). The study found evidence of continuous cortical growth across postnatal development, in that multiple cortical structures including the caudal hippocampus, dorsal striatum, thalamus, and corpus callosum are larger in adult cKO mice compared with littermate controls, which suggests that mistimed development of the forebrain and continued neurite growth may lead to circuit over-connectivity and maintenance, or failed synapse pruning.

The Met cKO mouse model shows additional construct and face validity to ASD. Hypomorphic signaling (Eagleson et al., 2005) or inactivation (Powell et al., 2003) of Met in mice disrupts interneuron migration, and leads to region-specific loss of GABAergic interneurons, in which the fast-spiking parvalbumin-expressing subpopulation may be preferentially affected. It has been previously reported that in thalamocortical slices prepared from Met cKO mice, the E/I ratio is biased toward excitation due to decreased GABA_A receptor mediated inhibition (Lo et al., 2016b), an effect that may be due to insulin fails to increase GABA_A receptor-mediated response in the barrel cortex of Met cKO mice. Another recent study employed resting-state functional magnetic resonance imaging (rs-fMRI) and in vivo high-resolution proton MR spectroscopy to examine neuronal connectivity (Tang et al., 2019). The study focused on somatosensory thalamocortical circuitry because somatosensory disturbances are frequent reported co-morbid conditions with ASD, and may be responsible for altered sensorimotor gating and for processing the saliency of social behaviors. The study found that Met cKO or heterozygote mouse brains showed impaired maturation of large-scale somatosensory network connectivity, including sex-dependent cortical network features and glutamate/GABA balance. This study suggests that aberrant functioning of the somatosensory thalamocortical system may underlie the somatosensory behavioral phenotypes associated with autism.

Met cKO also manifest disrupted synaptic connectivity at cellular levels and across different brain regions. Qiu et al. (2011) investigated potential intracortical circuit connectivity changes in Met^fd/fx/Emx1^cre cKO mice using laser scanning photostimulation (LSPS) combined with glutamate uncaging. This technique involves preparation of mouse cortical brain slices in which local circuit connectivity of interest can be preserved, followed by whole cell patch clamp recording on a defined neuron type. For example, intracortical circuit connectivity onto anterior frontal cortex L5 neurons that project to the striatum can be mapped after these neurons are retrogradely labeled by fluorescent latex microspheres. The brain slices are bathed in MNI-caged glutamate, which can be uncaged by focal high-energy UV laser positioned by a pair of x-y scanning mirrors. If one or a group of presynaptic neurons are synaptically connected with the patch clamped neuron, then uncaging glutamate at the soma location of these presynaptic neurons elicit action potentials in these neurons, which will result in excitatory postsynaptic current (EPSC) responses in the recorded neuron (Dantzker and Callaway, 2000; Shepherd and Svoboda, 2005; Suter et al., 2010). By uncaging glutamate at hundreds of locations (e.g. a 16x16 array) in brain slices, the connectivity map to the postsynaptic neuron can be constructed, which contains both location and strength of synaptic inputs from local intracortical circuits. If a genetic manipulation leads to altered synaptic connectivity that is hardwired during development, LSPS mapping may be an ideal tool in resolving such alterations. Although LSPS mapping can be ideally applied for studying local connectivity which has to be preserved in a brain slice preparation, when combined with optogenetic tools the technique also provides a feasible approach to map long-range circuits (e.g. channelrhodopsin 2 assisted circuit mapping-CRACM, (Petreanu et al., 2007)). Using LSPS mapping, Qiu et al. found that synaptic inputs from L2/3 neurons onto the anterior frontal cortex L5 corticostriatal (CS) pyramidal neurons are increased by approximately two-fold as a result of Met loss-of-function; this increase is cell type specific, as L2/3 synaptic inputs onto L5 corticopontine (CPn) neurons did not change. Interestingly, the increased connectivity is sublayer-specific, as only neurons with soma located in L5A show increased connectivity. The enhanced connectivity is also manifested at the synaptically-connected L2/3 > L5 neuronal pairs, indicating stronger unitary connections in local brain circuits. Although it is unclear from this study whether pre- or postsynaptic mechanisms contribute to increased synaptic connectivity, the findings may be related to dendritic spine phenotypes observed inMef^x/fx/Emx1^cre mice (Judson et al., 2010), in which increased spine head volume suggest more glutamate receptor content and an accelerated maturation of frontal circuits (Qiu et al., 2014). Therefore, even considering the potentially reduced synaptogenesis in Met cKO mice, an early maturation may lead to increased functionally connectivity in cortical circuits.

Another recent study (Peng et al., 2016) examined the effects of MET on prefrontal projection neuron connectivity by altering MET expression through in utero electroporation (IUEP), and found that post-synaptic inhibition of MET signaling (RNAi) in single L5 PFC neurons leads to decreased L2/3 to L5 connectivity, while enhanced MET signaling leads to overall increased connectivity. Strikingly, the increased connectivity is mostly explained by altered synaptic input topology, as ectopic L5 connection is prominent in neurons with Met cDNA over-expression. The results suggest that the balance of circuit connectivity and refinement may be altered when MET is disrupted. It is likely that activity-dependent synapse pruning is altered with persistent postsynaptic MET signaling. The effects of manipulating MET signaling in single prefrontal L5 neurons are in contrast to the Met cKO brains (Qiu et al., 2011), in which increased L2/3 inputs, but no ectopic connectivity to a specific anterior frontal L5 projection neuron populations (i.e. L5A corticostriatal neurons) were observed . As discussed above, the enhanced L2/3 > L5 connectivity in cKO mice cortical circuits may be due to an early maturation of glutamatergic transmission, because increased dendritic spine size, glutamate receptor content, decreased proportion of silent synapses were found in the cKO hippocampal circuits (Qiu et al., 2014).

Overall, combined with the LSPS mapping findings (Qiu et al., 2011), the enhanced/ectopic synaptic connectivity may be reminiscent of hyper-connectivity of local brain regions seen in ASD patients (Supekar et al., 2013; Uddin et al., 2013).

7. MET signaling as a regulator of synaptic plasticity.

In addition to the effects on early neuronal growth and morphological development, MET signaling is functionally coupled to synaptic function. Treatment of cultured neurons with HGF induces tyrosine phosphorylation of MET and enhances clustering of GluN2B, GluR1, and CaMKII synaptic proteins (Tyndall and Walikonis, 2006). Enhancing neuronal activity in cultured hippocampal neurons also increased HGF immunoreactivity (Tyndall et al., 2007). Tissue plasminogen activator, which is required for HGF activation, is released in a neuronal activity-dependent manner (Thewke and Seeds, 1999), suggesting MET signaling can be recruited by activity-dependent mechanisms, and may be capable of responding to activity-dependent neural plasticity changes. In hippocampal slices prepared from adult mice, HGF application enhances phosphorylation of NMDA receptor subunit GluN1 (Ser 896/897), augmented NMDA receptor-mediated currents, and elevates the activity-induced long-term potentiation (Akimoto et al., 2004). Kawas et al. (2013) found that MET protein levels are especially high in brain regions known to undergo extensive synaptic remodeling and plasticity, such as the CA1 region of hippocampus. Therein, MET activation increases the dendritic spine density and number of synapses. Using specific MET antibody combined with atomic force microscopy (AFM), the authors found that the activated multimeric form of MET is concentrated in the dendritic spine compartment while the inactivated monomeric form of MET is more prominent on the soma of neurons. Therefore, although the adult brain cortex possesses relatively low levels of MET, the signaling seems more restricted to subcellular compartments to modulate multiple aspects of synaptic physiology, including synaptic transmission and plasticity.

It is important to note that the cellular and molecular pathways engaged by MET, including the pleiotropic signaling components of PI3K, AKT, mTOR, STAT3, small GTPases, and ERK1/2, etc. are also shared by many other receptor tyrosine kinases expressed in the developing brain, including those receptors for neurotrophins (including receptors for nerve growth factor/NGF, brain-derived neurotrophic factor/BDNF, and neurotrophin 3/4) (Reichardt, 2006). By activating their RTKs, these neurotrophins are well documented to determine neural plasticity outcomes (McAllister, 1999; Park and Poo, 2013).

Functional study of the developing hippocampus circuits in Met^fx/fx/Emx1^cre cKO mice (Qiu et al., 2014) revealed that mouse CA1 neurons at early development (P12-14) show increased mEPSC amplitude, increased AMPA/NMDA receptor (A:N) current ratio, reduced ifenprodil-sensitive, GluN2B-mediated current at the CA3 > CA1 synapse; while the paired pulse responses are not altered. Biochemical analyses utilizing the synaptic membrane surface protein biotinylation revealed the CA1 synaptic protein profiles were altered in cKO hippocampus: increased GluA1, GluN2A, and decreased GluN2B proteins were found. Consistently, the proportion of silent synapses is significantly reduced in Met cKO mice. Silent synapses are those which contain only NMDA receptors, but acquire AMPA receptors to enable synaptic transmission in an activity-dependent manner. Silent synapses are abundant at nascent, immature synapses during early cortical circuit development, but are gradually diminished by incorporating AMPA receptors as synapses mature (Liao et al., 1995; Lu et al., 2001; Lu and Constantine-Paton, 2004). These results suggest Met cKO primarily affects synapse maturation by affecting the postsynaptic molecular composition. In the absence of MET signaling, cortical circuits may mature earlier. The decreased GluN2B content at CA3 > CA1 synapse may indicate decreased synaptic plasticity, as GluN2B-containing NMDA receptors exhibit high calcium permeability compared with GluN2A-containing receptors (Erreger et al., 2005; Shipton and Paulsen, 2014).

The impact of MET on the plasticity of mature neurons might be a direct impact on the plasticity of in mature neurons/synapses, or could be a consequence of the impaired morphogenesis (i.e. developmentally encoded). The synaptic and cellular correlate of plasticity, long term potentiation (LTP) and long term depression (LTD), was investigated in Met cKO by a recent study. Ma et al. reported that basal synaptic transmission at CA > CA1 synapse is enhanced at an earlier stage (P12-14) in Met cKO mice, while this difference no longer exists in adult (P56-70) mice (Ma et al., 2019). Strikingly, both LTP and LTD magnitudes are significantly increased in cKO mice at P12-14, but are markedly decreased at P56-70, during which wild type control littermates show robust LTP/LTD. These findings were further corroborated by the bidirectional regulation of LTP/LTD in single developing CA1 neurons, as enhancing MET signaling by IUEP delivery of Met cDNA decreased LTP magnitude of single CA1 neurons, while abating MET signaling by RNAi increased LTP magnitude.

In a separate behavioral study, Met^fx/fx/Emx1^cre cKO mice were found to display hypoactivity in both open field and T-maze tests, and exhibit deficits in spontaneous alternation in a Y maze test (Thompson and Levitt, 2015). When Met is deleted in all neuronal lineages (Met^fx/fx/Nestin^cre), these conditional null mice exhibited contextual fear conditioning deficits. Heterozygote mice (Met^fx/+ /Nestin^cre) also spent less time in the closed arms of an elevated plus maze, indicating an elevated anxiety phenotype. These data, taken together, may have implications on activity-dependent synaptic plasticity, memory, learning behavior, including the socio-cognitive learning that is impaired in autism patients (Leekam, 2016; Velikonja et al., 2019).

8. Conclusions, unanswered questions and future studies.

The developing brain undergoes many developmental milestones, and is capable of restructuring synaptic connections in response to experience. Neurogenesis and circuit wiring are initially established by genetic programs and then actively refined by the environmental inputs in which the individual is exposed to. This experience-dependent sculpting of neuronal circuits requires synaptic plasticity. There is now increasing evidence that neurodevelopmental disorders, such as ASD, can be viewed as disorders of impaired synaptic plasticity and cortical critical period (LeBlanc and Fagiolini, 2011). MET receptor tyrosine kinase exemplifies a pleiotropic, yet non-redundant genetic program that controls multiple aspects of brain development. Disruption of MET-mediated molecular signaling has been shown to contribute to genetic risk of autism. MET expression peaks during early postnatal development in the mouse forebrain, coinciding with a period in which cortical circuits undergo connectivity, refinement, and maturation. Currently, many of the high confidence candidate genes for autism identified for rare/syndromic forms of ASD are an integral part of synaptic molecular machinery and thus directly involved in circuit connectivity and synaptic function. MET signaling may be a unique mechanism that is capable of regulating a multitude of neurodevelopmental events, including neurogenesis, dendritic growth, spine/synapse morphogenesis, circuit connectivity and refinement, and synaptic plasticity changes. Therefore, efforts in deciphering the functional significance of MET at the molecular, cellular and system levels are central to understanding how MET contributes to ASD pathophysiology.

There are some unanswered questions that likely steer future research. For example, 1). Distinct molecular pathways that MET engages in neurons to shape various developmental outcomes (e.g. dendritic arborization vs. spine development, plasticity-related changes, etc.), and the brain region specificity of MET signaling; 2). Molecular mechanisms controlling the timing of MET expression across cortical domains. Specifically, the transcriptional regulatory mechanisms of Met and how they can be regulated by synaptic activity need to be established. In addition, the functional significance of rapid down-regulation of Met by the third postnatal week needs to be clarified. The recent finding that FOXP2, a well-known risk factor for language dysfunction (Lai et al., 2001), regulates MET expression (Mukamel et al., 2011) further supports the view of MET signaling as part of a complex molecular network implicated in ASD risk. 3). How MET signaling regulates critical period plasticity, considering the signaling profoundly alters synaptic glutamate receptor contents. Because the developing visual cortex expresses Met (Judson et al., 2009), it would be interesting to examine a visual cortex plasticity readout, such as the ocular dominance plasticity during critical period. This is particularly interesting in the context that impaired critical period plasticity has been reported for animal models of neurodevelopmental disorders including Angelman syndrome, Fragile X syndrome, and Rett syndrome (Dolen et al., 2007; Tropea et al., 2009; Yashiro et al., 2009; Harlow et al., 2010; Sato and Stryker, 2010). 4) How brain-circuit specific Met expression is related to behavior, particularly behaviors relevant to the most replicated ASD-related endophenotype – disrupted social behaviors. Answers to these questions may reveal how a conserved intrinsic cortical genetic program controls many key aspects of circuit development, and shed light on novel circuit-based interventions towards many neurodevelopmental disorders.

Significance.

Synapse development and plasticity are controlled by highly regulated molecular signaling events, disruption of which are posited to underlie neurodevelopmental and neuropsychiatric disorders. The timing of these events is likely regulated by many genes, disruption of which often confer risks to neurodevelopmental disorders (NDDs), including autism. One of such genes is human MET, which encodes MET receptor tyrosine kinase, and is capable of initiating pleiotropic signaling to shape a wide variety of developmental outcomes. Here, we review literature that sheds light on the role of MET signaling in cortical circuit development, maturation, and plasticity. Understanding how this pleiotropic signaling shapes brain development is critical to understanding disease pathophysiology and devise circuit-based interventions toward NDDs.