Latrophilin-3 disruption: Effects on brain and behavior

Abstract

Latrophilin-3 (LPHN3), a G-protein-coupled receptor belonging to the adhesion subfamily, is a regulator of synaptic function and maintenance in brain regions that mediate locomotor activity, attention, and memory for location and path. Variants of LPHN3 are associated with increased risk for attention deficit hyperactivity disorder (ADHD) in some patients. Here we review the role of LPHN3 in the central nervous system (CNS). We describe synaptic localization of LPHN3, its trans-synaptic binding partners, links to neurodevelopmental disorders, animal models of Lphn3 disruption in different species, and evidence that LPHN3 is involved in cognition as well as activity and attention. The evidence shows that LPHN3 plays a more significant role in neuroplasticity than previously appreciated.

2. Introduction

Latrophilin-3 (LPHN3) is an adhesion G-protein coupled receptor (GPCR) embedded in trans-synaptic complexes in brain regions that are key circuits that mediate learning, memory, attention, and activity. Variants of LPHN3 are linked in some children and adults with attention deficit hyperactivity disorder (ADHD), a finding replicated in cohorts from different countries (Arcos-Burgos et al., 2010; Domene et al., 2011; Gomez-Sanchez et al., 2016; Huang et al., 2018). There are 21 variants of LPHN3 linked to ADHD and these variants occur primarily in those with more severe ADHD symptoms (combined type) and affect the response to psychostimulant medication. There are no known null mutations of LPHN3 in humans, only variants that confer reduced protein expression and increased ADHD risk (Martinez et al., 2016). Beyond the link to ADHD, little is known about the function of LPHN3. The focus of this review is on what is known about the functional effects of LPHN3. We start with an overview of the latrophilin subfamily, including their structure, expression, regulation, synaptic interactions, and binding partners. We then discuss what is known about the functional effects of LPHN3 dysfunction based on model systems in which Lphn3 was genetically modified and the phenotypic changes that were characterized. We identify data gaps to spur further investigation.

3. Structure

Latrophilins are members of the GPCR superfamily (Lelianova et al., 1997). GPCRs comprise the largest receptor family in the human proteome with over 900 members divided into five groups according to sequence homology of the seven transmembrane shared domain (Hamann et al., 2015). Within the GPCR superfamily, latrophilins belong to the 33-member adhesion GPCR family (aGPCR). Structurally, aGPCRs have a long N-terminal domain containing adhesion motifs, a feature not found in other GPCR subfamilies, along with an autoproteolysis GAIN domain and a 7-transmembrane C-terminal (Fig. 1). The 33 human aGPCRs are divided into 9 subfamilies based on transmembrane sequence similarity (Nordström et al., 2009). The latrophilin subfamily has three members: LPHN1-3. All latrophilins have a seven transmembrane domain; latrophilins are also named adhesion G protein-coupled receptor L1-3 (ADGRL1-3) [OMIM 616417]. The N-terminal of LPHNs have a common structure, containing a signal peptide, galactose-binding lectin domain (RBL); olfactomedin domain OLF; a serine, threonine, proline rich region, and a hormone binding site HRM (Krasnoperov et al., 1997; Lelianova et al., 1997; Lu et al., 2015; Matsushita et al., 1999; Tomarev and Nakaya, 2009). The C-terminal of LPHNs are the same, containing a GAIN domain with a GPCR proteolysis (GPS) motif that is an auto-cleavage region. At the opposite end of the 7-transmembrane (7TM) domain is a G protein (Fig. 1).

Figure 1: LPHN3 structure.

Domain structures of latrophilins. LPHNs are G-protein-coupled receptors with unusually large extra- and intracellular sequences. The extracellular N-terminal is composed of five domains. The intracellular sequences have no discernable domain structure. The transmembrane regions of the LPHNs are similar to those of the secretin receptor family. Abbreviations: RBL = Lectin domain, OLF= olfactomedin domain, HRM = hormone binding site, GAIN= G-protein-coupled receptor autoproteolysis-inducing domain, GPS= G-protein coupled receptor proteolytic site, 7TM = 7 transmembrane domain.

LPHNs are dual receptors, in that they are cleaved into functional subunits (Fig. 1). Within the GAIN domain is the GPS site that encompasses a region where LPHN and other aGPCRs are cleaved, thereby generating a molecule containing N-terminal fragments (NTF) and C-terminal fragments (CTF). The generation of these fragments is the result of autoproteolysis (Araç et al., 2012). In a series of experiments, the cleaved fragments co-localized to the cell surface of synapses or were internalized in the synaptic terminal (Krasnoperov et al., 1997; Rahman et al., 2019; Silva et al., 2009; Volynski et al., 2004) where they are presumed to have distinct functions. Studies are needed to elucidate the function of these distinct fragments.

4. Regulation

LPHN genes have multiple splice sites that alter the proteins via insertions or deletions in the extracellular and cytoplasmic domains (Matsushita et al., 1999; Sugita et al., 1998). LPHNs have seven alternate splice sites. Four splice sites are located in the extracellular domain and produce a deletion or insertion at the borders of the lectin and olfactomedin domains (Boucard et al., 2014). These alternate splices modify coupling and G-protein signaling. Splice site 5 results in three variant inserts in the third cytoplasmic loop and these inserts modify G-protein coupling. Little is known about splice site 6. The most divergent splice site is 7. Inserts at splice site 7 cause a frame shift resulting in stop codons after the insert and result in the translation of a truncated protein. This may affect protein stability, targeting, and/or signaling (Sugita et al., 1998). Not all alternative splice sites are found in each gene. LPHN1 and LPHN3 have conserved extracellular sequences suggesting functions that are under evolutionary pressure to remain consistent, whereas LPHN2 is more variable. How these splicing differences are regulated is not well understood (Boucard et al., 2014).

5. Expression

Despite sequence homology between LPHN isoforms, each protein has a distinct spatial and developmental expression pattern. In rodents, LPHN1 is predominately found in brain; low levels of mRNA expression occur in non-neural tissue (Haitina et al., 2008; Sugita et al., 1998). LPHN2 is found at lower levels in brain and expressed in liver and lung. Since LPHN1 and LPHN2 are found outside the brain, they appear to be less specialized than LPHN3. Rodent LPHN3 is primarily expressed in the brain (Ichtchenko et al., 1999; Sugita et al., 1998) with very low levels in heart and kidneys (Boucard et al., 2014). In the brain, LPHN3 is highly expressed in the prefrontal cortex (PFC), caudate nucleus, amygdala, and cerebellum (Arcos-Burgos et al., 2010). Lower levels of expression are found in corpus callosum, hippocampus, occipital pole, frontal lobe, temporal lobe, and putamen. Within the hippocampus LPHN3 expression is highest in the dentate gyrus but it is also expressed in CA1–3 (Arcos-Burgos et al., 2010; Sugita et al., 1998). LPHN3 was not detected in thalamus, medulla, or spinal cord within the rat.

In humans, expression of LPHN2 is ubiquitous and LPHN1/LPHN3 mRNA are enriched in brain, similar to rodent orthologs (Arcos-Burgos et al., 2010; Sugita et al., 1998). LPHN3 is in the same brain regions in rodents as in humans. In contrast to rodents, in humans LPHN3 has high expression in adult and fetal adrenal glands (Xing et al., 2009). LPHN1 mRNA is found in the periphery of humans, whereas in the rodent the ortholog is found primarily in brain with low levels in the periphery (Sugita et al., 1998).

The expression of all three isoforms is tightly regulated during development. In human brain, LPHN3 expression is highest shortly after birth and declines to adult levels during maturation (Arcos-Burgos and Muenke, 2010). In human adrenal glands, the expression of LPHN2 and LPHN3 is higher in fetal tissue compared with adult tissue (Xing et al., 2009). These patterns are similar in rodents. Examination of Lphn1-3 mRNA by in situ hybridization in rat and mouse during development shows that Lphn2 is highest shortly after birth then declines to adult levels (Boucard et al., 2014; Kreienkamp et al., 2000). In contrast, rat Lphn1 mRNA levels increase during brain development. Mouse Lphn3 mRNA increases after birth reaching a peak during synaptogenesis and then declines to adult levels (Sando et al., 2019).

6. Function of Lphn3

LPHNs were discovered as endogenous receptors for α-latrotoxin, a component of black widow spider venom (Latrodectus mactans) (Krasnoperov et al., 1996; Krasnoperov et al., 1997; Kreienkamp et al., 2000; Sudhof, 2001; Valtorta et al., 1984). The venom binds to the hormone binding and GAIN domains in the NTF and stimulates GABA exocytosis (Clark et al., 1970; Krasnoperov et al., 1999; Linets'ka et al., 2002; Longenecker et al., 1970) from presynaptic terminals (Hlubek et al., 2000; Van Renterghem et al., 2000) by creating ionic pores that cause Ca²⁺ influx (Fesce et al., 1986). However, α-latrotoxin only produces exocytosis in LPHN1- and LPHN2-positive cells but not in LPHN3-positive cells (Matsushita et al., 1999).

LPHN3 has multiple synaptic spanning binding partners. The first one identified was a fibronectin leucine rich domain (FLRT) (O'Sullivan et al., 2012). Among the FLRTs, FLRT3 is a single pass transmembrane protein involved in axon guidance, cell migration, and neural development (Leyva-Díaz et al., 2014; O'Sullivan et al., 2012; Yamagishi et al., 2011). LPHN3 binds to FLRT3 at an olfactomedin domain (Lu et al., 2015; Ranaivoson et al., 2015) to form a ligand-receptor trans-synaptic bridge (O'Sullivan et al., 2012). Cell adhesion assays show that LPHN3/FLRT3 binding induces trans-cellular adhesion. Binding experiments also show that FLRT3, LPHN3, and UNC5 (Uncoordinated-5) form a trimeric complex in some regions (Lu et al., 2015).

The LPHN3-FLRT complex regulates glutamatergic synaptic density in CA1 of the hippocampus where it modulates synaptic plasticity (O'Sullivan et al., 2012; Sando et al., 2019). O’Sullivan et al. investigated the function of LPHN3-FLRT3 in cortical pyramidal neurons and found that LPHN3 interacts with FLRT3 during synapse formation influencing terminal density and signaling. Hence, FLRT3 and LPHN3 act as ligand–receptor modulators of excitatory synapses (O'Sullivan et al., 2014; Ranaivoson et al., 2015).

There is evidence of LPHN3 involvement in bidirectional signaling. Sando et al. (2019) hypothesized that LPHN3 acts trans-synaptically with adhesion molecules as ligands to provide specificity to synapse formation. They created a knockout Lphn3 mouse and found that LPHN3 is localized to postsynaptic spines in non-overlapping dendritic domains of CA1 pyramidal neurons. LPHN3 is found on excitatory synapses in S. oriens and S. radiatum layers, corresponding to different presynaptic inputs onto CA1 pyramidal neurons. Deletion of Lphn3 selectively decreased Schaffer collateral function in the S. radiatum and S. oriens. In rescue experiments, selective binding to FLRT3 or teneurins (TEN) disrupted activities required for synapse formation. Coincident binding of these two ligands with LPHN3 was required for signaling to occur.

Cell culture assays show that neither TEN2 nor FLRT3 alone induce excitatory synapse activation. Instead, FLRT3 and TEN2 induce synapse activation only when a splice variant of TEN2 binds to LPHN3, suggesting that simultaneous binding of FLRT3 and TEN2 to LPHN3 is obligatory. Sando et al. (2019) suggest that the location of LPHN3 in CA1 pyramidal neurons in different dendritic domains confers precision on which neurons are activated. Because the function of LPHN3 in synaptic plasticity requires co-binding with FLRT3 and TEN2, this may explain signaling specificity. These data imply a novel molecular mechanism for LPHN3 in synaptic regulation.

LPHN3 presence as presynaptic and FLRT3 presence as postsynaptic also was proposed (O'Sullivan et al., 2012; Sudhof, 2001; Valtorta et al., 1984). This conclusion is based on the observation that latrophilins that act on α-latrotoxin receptors are presynaptic. A postsynaptic localization of FLRT2 was suggested, however, the data for this are not conclusive because the antibody that targets the FLRT2 extracellular region may not be sufficiently selective to prove locality. This will be resolved when a more specific antibody becomes available (Schroeder et al., 2018). Teneurins are proposed to act in synapse formation as heterophilic cell adhesion molecules with LPHN3 (Berns et al., 2018; Hong et al., 2012), which is in agreement with the data of Sando et al. (2019).

To determine if LPHN3 is involved in excitatory synapse formation and maintenance in a different brain region, Zhang and Liakath-Ali et al. (2020), examined LPHN3 in the cerebellum. They found that expression of LPHN2 and LPHN3 is layer specific in the cerebellar cortex (Zhang et al., 2020). Lphn3 is expressed in granule and Purkinje cells, Bergmann glia, and interneurons. Deletion of Lphn2 or Lphn3 alone did not affect synapse formation in this region. However, double knock-outs (KO) of both caused dysregulation of synapse activation. These results demonstrate that although LPHN2 and LPHN3 have distinct functions in the hippocampus, they are redundant in Purkinje cells.

In summary, loss of LPHN3 alters GPCR and neuronal signaling. However, precisely how LPHN3 contributes to these processes is still under investigation. Studies are needed to understand the role the C-terminus of Lphn3 plays in intracellular signaling. One hypothesis is that the classic 7TM domain and C-terminus function as generic GPCRs, because the C-terminal co-purifies with Gα_q/11, which activates phospholipase-C followed by mobilization of intracellular Ca²⁺ that in turn causes release of neurotransmitter (Rahman et al., 1999). Moreover, the C-terminal interacts with multiple postsynaptic density proteins, such as SHANK and ProSAP (Kreienkamp et al., 2000). These proteins provide intracellular stabilization and regulate cytoskeleton, suggesting that LPHNs play a role in intracellular docking or vesicular function by affecting the structure of synaptic active zones.

7. LPHN3 in Neuropsychiatric Disorders

As mentioned, variants of LPHN3 are linked to ADHD. ADHD is the most prevalent neurodevelopmental psychiatric disorder. The United States (US) Center for Disease Control and Prevention (CDC) estimates a 9.4% prevalence of ADHD in children. The US National Institute of Mental Health estimates 4.4% prevalence of ADHD in adults. ADHD is associated with increased risk of adverse outcomes, such as academic underachievement, injuries, traffic accidents, increased healthcare utilization, substance abuse, criminality, unemployment, divorce, and suicide (Biederman et al., 2010; Fleming et al., 2017). ADHD is characterized by symptoms of hyperactivity, impulsivity, and inattention (American Psychiatric Association, 2013). Little is known about the pathophysiology of ADHD. A barrier to understanding this disorder is that ADHD is polygenic and associated with multiple small-effect gene variants.

ADHD is ~80% heritable with ~75% monozygotic twin concordance in cases where environmental risks are not shared (Coolidge et al., 2000; Faraone et al., 2005; Gillis et al., 1992; Gjone et al., 1996; Levy et al., 1997; Matheny and Brown, 1971). Twin and adoption studies show genetic overlap between ADHD and other conditions, including cognitive impairment, autism spectrum disorder, bipolar disorder, and major depressive disorder (Andersson and Tuvblad, 2020; Angold and Costello, 1993; Coolidge et al., 2000; Faraone et al., 2005; Hudziak et al., 2000; Jacob et al., 2007; Leitner, 2014; Martin et al., 2006; Mulligan et al., 2009; Rommelse et al., 2010; Sandstrom et al., 2021; Schiweck et al., 2021; Singh et al., 2006). Genome-wide association studies (GWAS) identify risk factors but not etiology.

Fine mapping linkage studies in the Paisa, a genetically isolated population in Antigua, Colombia, identified variants of LPHN3 associated with ADHD (Arcos-Burgos et al., 2010). Arcos Burgos et al. (2010) conducted a genome-wide linkage study of 16 large families with 433 individuals showing linkage of ADHD to chromosome 4q13.2 and to LPHN3 (Acosta et al., 2016; Arcos-Burgos and Muenke, 2010) an observation replicated in other studies (Domene et al., 2011; Gomez-Sanchez et al., 2016; Huang et al., 2018). The polymorphism was reported to increase the risk of ADHD by 1.2-fold. Those with the LPHN3 polymorphism had more severe ADHD symptoms of the combined type and altered responses to ADHD medications (Arcos-Burgos et al., 2010; Huang et al., 2018). The prevalence of ADHD was estimated to be reduced by ~9% if the effect of the LPHN3 variants could be effectively treated (Arcos-Burgos and Muenke, 2010). An association between LPHN3 and persistence of symptoms into adulthood was also noted (Franke et al., 2012; Ribases et al., 2011).

The spatial and temporal expression of LPHN3 supports its involvement in ADHD. LPHN3 is expressed in regions implicated in ADHD, including amygdala, hippocampus, striatum, and PFC (Arcos-Burgos et al., 2010). Several of these regions are part of dopaminergic systems: nigrostriatal, mesolimbic, and mesocortical pathways. ADHD is hypothesized to have dopaminergic involvement since dopaminergic drugs are used as treatments, and various dopaminergic genes are implicated in small effect associations with the disorder (Faraone et al., 2015). ADHD is a developmental disorder and LPHN3 expression is highest neonatally. Therefore, the timing of ADHD onset and peak expression of LPHN3, including expression of variant forms of LPHN3, show a temporal and spatial fit suggesting noncoincidence (Arcos-Burgos and Muenke, 2010).

Since ADHD is polygenic, interactions between LPHN3 and other ADHD risk genes were investigated. For example, single nucleotide polymorphisms (SNPs) within LPHN3 and genes within the 11q allele double the risk of ADHD (Jain et al., 2012). Adverse mutations within this region, called the NTAD cluster (NCAM1, TTC12, ANKK1, and dopamine (DA) D2 receptor (DRD2)) do not confer an ADHD risk alone. However, the interaction of the NTAD cluster and variants in Lphn3 double the risk of ADHD, indicating interactive risks with other genes. This association was replicated in a large population in Germany and two European-American samples (Kappel et al., 2017). These interactions are associated with increased ADHD severity and persistence (Acosta et al., 2011).

The genes described in the NTAD cluster were associated with different disorders. The haplotype on chromosome 11q spans 166 kbps, from intron 7 of NCAM1, encompasses TTC12 and ANKK1, and is adjacent to DRD2. ANKK1 is a protein that is expressed in placenta and spinal cord, but has not been observed in the developing or adult brain (Neville et al., 2004). NCAM1 is a member of the immunoglobin superfamily and is thought to play a role in normal neural development, function, and plasticity (Rønn et al., 2000; Rønn et al., 2002). Mouse KO models of NCAM2 were proposed as models of schizophrenia based on brain morphological changes and reduced prepulse inhibition of the startle response (Wood et al., 1998; Wood and Toth, 2001). Mouse KOs have hyperactivity, aggression, anxiety, and abnormal social behaviors (Stork et al., 2000). In humans, genetic variations are associated with neural tube defects and bipolar disorder (Deak et al., 2005; Stork et al., 2000). In addition, isoform variations in postmortem brain tissue and cerebrospinal fluid in patients are associated with schizophrenia, bipolar disorder and autism, suggesting that NCAM1 has pleiotropic effects (Vawter, 2000). DRD2 is also a candidate genetic association with ADHD (Esposito-Smythers et al., 2009).

Using mutational analyses of the U.S. population with Lphn3 variants, Domene et al., found that there are 21 variants within the gene of which 14 are reported and 7 that are novel (Domene et al., 2011). Of the 21 variants of LPHN3 identified, 8 are in noncoding and 13 in coding regions (Domene et al., 2011). Mutations within the coding region were found between the hormone receptors and GPCR proteolysis site on the N terminal (Domene et al., 2011) (Fig. 2). SNPs within this region affect mRNA splicing and protein function. SNPs within non-coding regions may affect protein expression. Cell culture studies may help determine if in silico predictions of these haplotypes affect protein function. Domene et al. (2011), showed that LPHN3 polymorphisms cause downregulation of LPHN3 activity but not loss of function, reinforcing the idea that ADHD is the product of verging minor effect gene variants acting in concert. In addition, Martinez et al., identified a transcriptional enhancer element, which is associated with reduced Lphn3 expression in several non-coding SNPs (Martinez et al., 2016). These data suggest that there is evidence of common non-coding variants implicated in the pathogenesis of ADHD.

Figure 2: Schematic depiction of LPHN3 exons and sequence variations in LPHN3.

2A represents the relevant protein domains with 24 different exons with the intronic segments connecting them. 2B represents the protein domains beneath the exons coding each motif. All sequence variants are identified with an arrow with the variant labeled above. The list of the variants can be found in Domene et al. (2010).

LPHN3 variants are also associated with cognitive deficits in humans (Fallgatter et al., 2013). While most ADHD patients do not have cognitive deficits, some do, but whether these are primary or secondary to inattention and/or impulsivity is unclear. To complicate matters, the cognitive disabilities seen in ADHD are heterogeneous, with some cognitive domains more affected in some people than others, whereas a few have non-selective impairments (Nigg et al., 2005; Sonuga-Barke, 2003).

8. Animal models of Lphn3 disruption

To gain Insight into the physiological function of LPHN3, gene targeting was used in different species (Table 1). Because of the infeasibility of modeling 21 LPHN3 variants implicated in ADHD, gene KOs are the main tools used.

Table 1:

Lphn3 null experimental systems

Model	Construct	Hyperactive	Impulsive	Inattentive	Response to medication	DA Changes	Reference
Drosophila	Pan neuronal drivers and shRNA knockdown of lphn	Yes	N/A	NT	↓ activity in hyperactivity to methylphenidate	No change in TH+ cells	(van der Voet et al., 2016)
Zebrafish	Morpholino knockdown of lphn3.1	Yes	Yes	NT	↓ in activity to methylphenidate and atomoxetine	No Δ in DA content ↓ TH positive cells Differential response to DA	(Lange et al., 2012)
Mouse	Gene trap truncation of mucin stalk domain of Lphn3 in Exon 6	Yes	Yes	Yes	↑ in activity to cocaine	↑ DA levels ↑ DAT, DRD2, DRD4, TH ↑ neurite outgrowth ↓ blood Ca2+, calbindin ↑ CamKII	(Mortimer et al., 2019; Orsini et al., 2016; Wallis et al., 2012)
Rat	Exon 3 deletion of Lphn3 with CRISPR/Cas9	Yes	NT	NT	↓ in activity to amphetamine	↓ DARPP-32, TH, DRD1 in striatum ↑ DA conc., ↑ release frequency in striatum	(Regan et al., 2020b; Regan et al., 2019)

^*NT: Not tested

8.1. Drosophila

Van der Voet et al. (2016) used Drosophila melanogaster to construct different genetic KOs. They derived 91 genes from meta-analyses of candidate genes and GWAS based on SNPs. Of the 91 genes, 78 fly mutations resulted in hyperactivity. Three candidate genes were further analyzed: neurofibromatosis −1 (Nf1), DA transporter (DAT), and lphn. For lphn, they used conditional knockdown of the homologous lphn (dCirl) in Drosophila with pan-neuronal drivers. Locomotor activity of male flies on a 12-hour light-dark cycle were recorded for 4 days. The lphn null flies were hyperactive during the day and more severely hyperactive during the night. The greater nocturnal hyperactivity in the lphn knockdown flies raises the question of light suppressed activity, since DA signaling is repressed by light in Drosophila (Shang et al., 2011). Therefore, Van der Voet et al., monitored activity levels during the day in a dark room and found that activity increased to nighttime levels, indicating that the lphn knockdown alters DA signaling. Further, the nocturnal hyperactivity was dose dependently attenuated by food containing 0.5 mg/mL and 1.0 mg/mL of methylphenidate. However, no differences in the number of DA positive neurons were found in knockdown flies, indicating dynamic DA changes rather than of DA content.

8.2. Zebrafish

Zebrafish (danio rerio) are also used to investigate the function of LPHN3 (Lange et al., 2012). Zebrafish have two LPHN3 orthologs: lphn3.1 and lphn3.2. Both have similar expression profiles during development. As the larval development progresses, lphn3.1 and lphn3.2 display shared expression that becomes more prominent in the ventral telencephalon, diencephalon, and in the posterior and ventral spine. In adult zebrafish, expression of lphn3.1 is found along the telencephalic midline, telencephalic parenchyma, anterior thalamus, periaquaductal gray, superior nucleus of raphe, periventricular nucleus of the inferior hypothalamus, and cerebellum. Lange et al. (2012), used the model to assess the developmental and behavioral effects of lphn3.1 disruption. Splice morpholinos (MO) were used to knockdown lphn3.1. Fish at the single cell stage were injected to block the exon 2 splicing site (MO1) and the exon 6 splicing site (MO2).

Lange et al. (2012) assessed swimming 6 days post fertilization. Both the MO1 and MO2 larvae had increased swim distance with MO1 > MO2. The increase in swim distance in the lphn3.1 morphants was from an increase in swim speed during active periods, not a decrease in rest time between swimming bouts. Both morphants had decreased activity during the night, however the lphn3.1 MO1 morphants remained more active than controls. The lphn3.1 morphants also had more activity bursts with significantly faster acceleration than controls, which the authors interpreted as a possible indication of impulsivity. The increased activity bursts were reduced to control levels by treatment with 10 μM methylphenidate or 1 μM atomoxetine. The attenuation of hyperactivity by ADHD medications adds support to a potential link with ADHD.

DA enriched regions in zebrafish are the olfactory bulb, preoptic region, pretectum, posterior tuberculum, and hypothalamus. One of the structures involved in locomotion in fish is the posterior tuberculum, a structure analogous to the substantia nigra pars compacta and ventral tegmental area in mammals, and the region with the highest density of DA neurons (Schweitzer and Driever, 2009). In lphn3.1 morphants this region was disorganized and had reduced numbers of DA neurons. Levels of slc6a3, the gene for DAT, were also lower in the posterior tuberculum of the lphn3.1 morphants compared with controls. There were no effects in morphants on DA or dihydroxyphenylacetic acid (DOPAC) levels or DA turnover. The zebrafish data further implicate lphn3.1 disruption induced hyperactivity arising from changes in DA signaling.

Another zebrafish study used a pharmacological approach (Lange et al., 2018), employing DA agonists (apomorphine, quinpirole, SKF-38393 (SKF)) and antagonists (haloperidol, eticlopride, SCH-23390). Apomorphine, a non-selective DA agonist, dose-dependently increased activity in controls, whereas in lphn3.1 morphants the hyperactivity was attenuated. In response to SKF-38393, a DRD1 agonist, both lphn3.1 morphants and controls had similar increases in activity. In response to quinpirole, a DRD2 selective agonist, controls had decreased activity whereas it had virtually no effect on lphn3.1 morphants. Haloperidol, a non-specific DA receptor antagonist, produced progressive dose-dependent decreases in activity in all genotypes but the decreases were greater in morphants. These data fit with the decreased sensitivity to DA observed with the DA receptor agonist apomorphine. SCH-23990, a DRD1 specific antagonist, also produced a dose-dependent decrease in activity but had little effect on the morphants. Eticlopride, a DRD2 specific antagonist, in control larvae produced a dose-dependent decrease in activity whereas the lphn3.1 morphants had a moderate increase in activity at first (40 min), with the morphants being generally less sensitive to DRD2-like drugs than controls. Hence, lphn3.1 zebrafish morphants have blunted responses to DA agonists and antagonists compared with controls, consistent with a role of DA in changes stemming from deletions of lphn3. The zebrafish data, therefore, support a role of reduced lphn3.1 expression associated with hyperactivity and impulsivity from DA dysregulation.

8.3. Mouse

A third model of Lphn3 disruption is the Lphn3 KO mouse. A gene trap was used to delete the mucin stalk portion of the protein, which is hypothesized to cause dysregulation of intracellular signaling (Wallis et al., 2012). This mouse has no expression of LPHN3 (Mortimer et al., 2019). In an open field, Lphn3 KO mice travelled greater distances in both the center and periphery compared with wildtype littermates (Mortimer et al., 2019; Wallis et al., 2012). In humans, boys with ADHD are more likely to have hyperactivity compared with girls (Mowlem et al., 2019; Rucklidge, 2010), whereas girls more often have inattentive or impulsive characteristics (Biederman et al., 1999). In Lphn3 null mice, there were no differences in hyperactivity between males and females. In mice, stimulant-induced activity was tested to determine if Lphn3 null mutants were differentially sensitive to cocaine (Wallis et al., 2012). Lphn3 KO and WT mice had increased activity after cocaine, but at the highest cocaine dose, the drug-induced hyperactivity was attenuated in KO mice compared with WT mice.

Lphn3 KO mice were also tested for schedule-controlled behavior, working memory, and motor function (Orsini et al., 2016). Lphn3 KO mice had exaggerated reward seeking on a fixed ratio schedule. For working memory, Lphn3 KO mice performed similarly to WT mice (Orsini et al., 2016); since working memory depends on the PFC, this sparing may indicate that LPHN3 is less involved in PFC-related functions.

Another study using the Lphn3 KO mouse (Mortimer et al., 2019), found that Lphn3 KO mice are hyperactive, impulsive, and have gait disturbances. The mice also had impairments in spatial learning and memory in the Barnes maze, increased sociability in a social interaction test, and decreased aggression in a resident-intruder test. At the transcriptomic level, using RNA-sequencing, the number of genes found to exhibit differential expression was small, indicating a pathway of action rather than a general neurological perturbation. Interestingly, the gene-set analysis showed the most differential expression in the PFC, in pathways previously associated with ADHD. Specifically, Slc6a3 encoding DAT was dysregulated in the PFC. These transcriptomic data differ from the behavioral deficits seen by Orisni et al. (2016), who showed no working memory deficits. O’Sullivan et al.(2014) also showed effects of Lphn3 deficiency on cortical synaptic function. The implications of Lphn3 changes in the PFC are poorly understood and require more detailed investigation.

Lphn3 KO mice had increased mRNA expression of the serotonin (5-HT) transporter Slc6a4 and of 5-ht2a, Dat1, Drd4, neural cell adhesion molecule (Ncam), nuclear receptor related 1 (Nurr1), and tyrosine hydroxylase (Th) (Wallis et al., 2012). In dorsal striatum null Lphn3 mice had increased DA and 5-HT levels compared with WT littermates as measured by HPLC (Wallis et al., 2012). The hyperactivity in Lphn3 KO mice is consistent with elevated monoamine content, but increased DA and 5-HT levels are not found in patients with ADHD.

The null Lphn3 mice also exhibited increased neurite outgrowth in the cortex but not in the hippocampus compared with WT mice (Orsini et al., 2016). In hippocampal cell culture from these mice, the loss of LPHN3 reduced the number of glutamatergic synapses (O'Sullivan et al., 2012). Lphn3 KO mice had decreased levels of serum Ca²⁺, overexpression of CamKII, and decreased calbindin. Since Ca²⁺ influx is important for neurotransmitter release, changes in Ca²⁺ influx affect multiple functions, including activity and cognition. Recent data suggest that individuals with ADHD and alcohol use disorders have low serum Ca (Bener and Kamal, 2013; Bener et al., 2014; Kamal et al., 2014; Naude et al., 2012) but what role, if any, LPHN3 plays in Ca regulation is not known.

In a study by Sorokina et al. (2018), they used a polygenic mouse model of ADHD based on selective breeding to identify genes associated with hyperactivity. The high-activity line was bred for 17 generations for increased home-cage activity with a control line maintained in parallel. The hyperactive line was hyperactive in a familiar environment (a hallmark of ADHD), more impulsive, and their hyperactivity was attenuated when given amphetamine. These mice had downregulated expression of Lphn3, Flrt3, and Wnt in striatum. Histologically, the hyperactive mice had decreased synaptophysin, implicating changes in synaptic vesicle function and had a 46% reduction in synapse number in the striatum compared with WT mice. These data support the hypothesis that downregulation of Lphn3 contributes to fewer synapses in the striatum and contributes to an ADHD-like phenotype.

8.4. Rat

Each of these models provides insight into the function of the LPHN3 protein, however, none of the above models is as well suited to test cognitive function as rats. Accordingly, we deleted exon 3 of the Lphn3 gene in Sprague Dawley rats as an alternate approach (Regan et al., 2019).

Constitutive Lphn3 KO male and female rats were hyperactive in a home-cage test that was most pronounced during the dark phase (Regan et al., 2019). Hyperactivity in a familiar, as opposed to novel, environment is a cardinal feature of ADHD (Russell, 2011; Sagvolden and Sergeant, 1998), therefore the hyperactivity in Lphn3 KO rats has a phenotype more consistent with ADHD than the above models. Moreover, this hyperactivity was present during development and in adults, being observed at postnatal day (P)35 and at P50. In a novel environment, Lphn3 KO rats were also hyperactive compared with controls. When administered amphetamine, Lphn3 KO rats remained hyperactive but their increase relative to their higher baseline was attenuated compared with WT rats whose activity increased dramatically. Similarly, children with ADHD have a decline in hyperactivity after treatment with amphetamine or methylphenidate relative to their pre-treatment higher than normal activity levels (Briars and Todd, 2016; Labbe et al., 2012).

Biochemically, no changes in levels of DA, norepinephrine (NE), 5-HT or their major metabolites were seen by HPLC in striatum, hippocampus, or PFC (Regan et al., 2019). On the other hand, several striatal DA markers were altered when assessed by western blot analysis. Both tyrosine hydroxylase (TH) and aromatic L-amino acid decarboxylase (AADC) were increased, suggesting greater DA and/or NE availability, and the DAT was increased suggesting increased DA uptake. DRD1 was decreased, suggesting that DA overflow might be causing a postsynaptic downregulation of DRD1. DA-regulated neuronal phosphoprotein (DARPP)-32 was also reduced in the Lphn3 KO rats compared with controls. DARPP-32 is a downstream effector of postsynaptic signal transduction and may be a response to the DRD1 down regulation.

Lphn3 KO rats also have alterations in spontaneous DA release assessed with fast scan cyclic voltammetry (Regan et al., 2020a). The concentration and frequency of DA release in the striatum was markedly increased. The duration of DA in the synapse was decreased suggesting increased DA uptake that fits with the increased DAT. The data reflect dysregulation of phasic DA release, which may explain the altered DA markers noted above. The increase in DA release is consistent with the increased TH and AADC. The increased DA release may contribute to the downregulation of DRD1 leading to reduction in DARPP-32, whereas the increase in DAT is consistent with decreased DA transients measured by fast scan cyclic voltammetry. Together, the data provide evidence that Lphn3 KO rats have saturated DA phasic neurotransmission (Figure 3). The data imply that a complex set of DA system interactions lie beneath the Lphn3 KO hyperactivity phenotype. This fits with the downward shift in hyperactivity of KO rats after treatment with amphetamine. Hence, the data point toward LPHN3 having a role in striatal DA regulation. These effects agree with data from mouse, Drosophila, and zebrafish that disrupted LPHN3 leads to hyperactivity and may help explain why variant reduced LPHN3 expression in children contributes to ADHD. The increase in DA signaling, however, may not be the only effect of Lphn3 deletion. More data are needed on the developmental role of LPHN3 in brain, including on DA and glutamate neuron ontogeny.

Figure 3: Hypothesis of striatal DA dysregulation in Lphn3 KO rats.

Lphn3 KO rats have several striatal dopamine (DA) markers that were altered when assessed by western blot analysis. Both tyrosine hydroxylase (TH) and aromatic L-amino acid decarboxylase (AADC) were increased, suggesting greater DA and/or norepinephrine (NE) availability. The dopamine transporter (DAT) was increased suggesting increased DA uptake. The dopamine receptor 1 (DRD1) was decreased, suggesting that DA overflow might be causing a postsynaptic downregulation of DRD1. Dopamine-regulated neuronal phosphoprotein (DARPP)-32 was also reduced in the Lphn3 KO rats compared with controls. DARPP-32 is a downstream effector of postsynaptic signal transduction and may be a response to the DRD1 down regulation. Release of DA in Lphn3 KO rats was also altered. The concentration and frequency of DA release in the striatum was markedly increased. The duration of DA in the synapse was decreased suggesting increased DA uptake that fits with the increased DAT. The data reflect dysregulation of phasic DA release, which may explain the altered DA markers. The increase in DA release is consistent with the increased TH and AADC. The increased DA release may contribute to the downregulation of DRD1 leading to reduction in DARPP-32, whereas the increase in DAT is consistent with decreased DA transients measured by fast scan cyclic voltammetry.

Along with associations with hyperactivity, mutations in Lphn3 are linked to cognitive deficits in patients (Fallgatter et al., 2013). Fallgatter et al., showed that patients with Lphn3 mutations associated with ADHD have altered neural activity in the frontal cortex with deficits of cognitive response control during a Go-NoGo task. To investigate the role of Lphn3 deletion on cognition in rats, we tested Lphn3 KO and WT littermates in two mazes that assess different types of learning and memory. We used Morris (MWM) and Cincinnati (CWM) water mazes. The CWM is a test of egocentric, i.e., internal cue-based navigation that is striatally and DA dependent (Braun et al., 2015; Braun et al., 2016; Vorhees and Williams, 2016). This is consistent with functional magnetic resonance imaging studies that implicate striatal changes in ADHD patients who have variant LPHN3 (Rubia, 2018). The MWM, by contrast, is a hippocampal dependent test that assesses spatial/allocentric learning and reference memory (Vorhees and Williams, 2006) and is NMDA-receptor dependent.

The experiment was designed to determine if Lphn3 KO rats can perform comparably to WT rats if given extended training in the MWM. They were also tested with a visual cue to ensure that they can use visual information to navigate. Littermates (5/genotype) from 5 litters were tested in a MWM in two phases: training and extended acquisition. Training consisted of 6 trials with a fixed start and fixed platform positions to render the task as easy as possible. Curtains were closed around the maze making the marked platform the only prominent cue to the location of the 10 cm diameter platform. Although the platform was submerged 2 cm it had an orange ball mounted on a stainless-steel pole that was 10 cm above the water levels and clearly visible. The extended acquisition test consisted of four trials per day using the standard MWM method in which the start positions were from four different pseudorandom positions on each trial of each day (Vorhees and Williams, 2006). The time limit per trial was 90 s. Acquisition also used a 10 cm platform but with no cue attached to it. Path efficiency, latency, and swim speed were analyzed. Data were analyzed using mixed linear model.

ANOVAs.

After WT rats reached asymptotic performance they discontinued tests, while KO rats continued tests to a limit of 12 days to determine if they could eventually reach WT rat levels of performance. To control for litter, only one rat/genotype/litter was used, and litter was a random factor in ANOVAs. Every litter used contained a WT and KO.

For cued training, Lphn3 KO rats had increased latency compared with WT rats [F(1, 8) = 22.28, p <0.01] and an interaction of genotype x day [F(1,88) = 4.34, p < 0.05] (Fig 4A, B). The Lphn3 KO rats had difficulty even in the presence of a proximal cue and constant start and goal positions. However, this training task requires visually egocentric navigation and the CWM data show that KO rats cannot perform that task, hence the two findings are consistent.

Figure 4: Cued training and acquisition phase Morris water maze performance in Lphn3 KO and WT rats.

A. Cued training day 1 by trial. B. Cued training day 2 by trial. C. Learning curve for latency with all groups. On day 6, WT rats reached asympotic performance and received no further testing. The dashed line represents the day 6 data for WT rats. D. Learning curve for path efficiency with all groups. On day 6, WT rats reached asympotic performance and received no further testing. The dashed line represents the day 6 data for WT rats. * p < 0.05, ** p < 0.01, *** p < 0.001 compared with WT controls. N= 5/ group.

On acquisition, KO rats were also impaired. Across the first 6 days, KO rats had increased latency compared with WT rats [F(1, 8) = 22.28, p < 0.01] (Fig. 4C). Path efficiency was similarly impaired in KO rats [genotype: [F(1, 12.2).47 = 56.47, p <0.0001]]; there was also a genotype x day interaction [F(5, 31.5) = 3.11, p< 0.05] (Fig 4D). After WT rats reached asymptotic performance they were discontinued. The impairment in KO rats persisted through the next 6 days and by day 12 there was no evidence of improvement. The pattern of performance of the Lphn3 KO rats differed from WT rats. Three out of five Lphn3 KO rats had persistent thigmotaxis, swimming around the perimeter of the pool without venturing out to search for the platform. Examples of this are shown in Figure 5 alongside the performance of WT rats. The poor performance of the Lphn3 KO rats raises further questions that we are currently investigating. Mortimer et al. (2019), found that Lphn3 KO mice have gait disturbances. Could gait disturbances interfere with swimming ability? This seems unlikely because in an open-field or home-cage, KO rats were hyperactive not hypoactive from ambulatory deficit. Perhaps the gait disturbance does not slow the KO rats but instead causes awkward movements. We see no evidence of that in KO rats, therefore, this may be specific to KO mice. Moreover, the KO rats swim well, they simply do not use surrounding cues to navigate.

Figure 5: Representative track plots of Lphn3 KO rats swimming in the Morris Water Maze.

The perimeter of the maze is seen in orange and the swimming conducted by the rat is in purple. A. Day 6 track plot of a WT rat. B. Day 6 track plot of a WT rat. C. Day 6 track plot of a KO rat. D. Day 6 track plot of a KO rat. E. Day 12 track plot of the same KO rat in C. F. Day 12 track plot of the same KO rat in C.

Are KO rats thigmotaxic because of impaired vision? Lphn3 is expressed in some cells in the retina (Gehring; Moreno-Salinas et al., 2019). To address this, KO rats were tested in acoustic startle using a brief flash of light in a prepulse inhibition paradigm. Light prepulses inhibited acoustic startle in KO rats to the same degree as it did in WT rats. Alternatively, are KO rats thigmotaxic because they are prone to perseverative behavior when confronted with a challenging task? We are testing this by evaluating KO rats in non-swimming tests of learning and memory. Could LPHN3 deletion alter HPA axis function resulting in elevated stress at a level that interferes with attention and learning. We also will test this possibility. There is a report of a link between maternal stress and Lphn3 KO (Choudhry et al., 2012). Is the MWM deficit specific to disrupted LPHN3 in the hippocampus. The Mortimer et al. (2019) finding of Lphn3 KO mouse deficits in a non-swimming spatial test, the Barnes maze, suggests that spatial ability is the principal deficit rather than performance factors.

Spatial learning is associated with long-term potentiation (LTP) and N-methyl-D-aspartate (NMDA) receptors (Jeffery and Morris, 1993). We found that Lphn3 KO rats have decreased LTP in the CA1 in brain slices. NMDA receptors mediate CA1 LTP (Herring and Nicoll, 2016) and we find decreased NMDA-NR1 levels in Lphn3 KO rats compared with WT rats (Regan et al., 2018). Therefore, future studies should examine the effect of LPHN3 deletion on LTP.

8. Summary

LPHN-3 is involved in brain development and function but its functional role in activity, cognition, and other behaviors is poorly understood, despite it being evolutionarily conserved, with orthologs in all vertebrates and many invertebrates. Moreover, LPHN3 is found in brain regions involved in activity and cognition and deleting this gene in four different species all implicate LPHN3 in motor activity and DA signaling, that support indirectly human studies implicating LPHN3 variants in some cases of ADHD. The exact role of LPHN3 in higher cognitive function, attention, anxiety, and executive function remains to be determined and the molecular basis for such changes will be critical in understanding how this protein operates in health and disease.

Highlights.

• Latrophilins are adhesion G-protein-coupled receptors

• Latrophilin-3 (Lphn3) is a trans-synaptic regulator of synaptic function

• 21 gene variants of LPHN3 are linked to ADHD

• Lphn3 disruption in Drosophila, zebrafish, mice, and rats result in hyperactivity

• Lphn3 KO rats show selective cognitive deficits and dopamine abnormalities