The TAO (Thousand and one) kinase subfamily is part of the larger STE20 (sterile 20) kinase family, a diverse group of serine-threonine kinases that participate in a variety of signaling pathways (Dan et al., 2001). The TAO kinase family is highly conserved, and has three members in mammals including TAOK1/PSK2/MARKK (Hutchison et al., 1998), TAOK2/PSK1 (Moore et al., 2000), and TAOK3/JIK (Tassi et al., 1999), but just a single representative TAO kinase in C. elegans (Kin-18; Berman et al., 2001; Spiga et al., 2013; Yin et al., 2016) and Drosophila melanogaster (Tao); (Liu et al., 2010). Structurally, all members of TAO kinases contain three domains including the N-terminal catalytic domain, the central domain, and the C-terminal regulatory domain which regulates catalytic activity (Figure 1A). Among these domains, the catalytic domains of all TAO kinase members are extremely conserved, but the regulatory domains may determine their distinct functions. In addition, each member of mammalian TAO kinases has two isoforms (α and β) except for TAO2, which has a putative third one (TAO2γ) which is mainly distinguished by the regulatory domain (Figure 1A), suggesting a distinct biological function of the different isoforms. Thus, it was reported that the two isoforms (α and β) of TAO2 play different roles in dendritic and dendritic spine development (Yasuda et al., 2007; Richter et al., 2018), the possible isoform-dependent functions in other members should be further explored.

TAO kinases are ubiquitously expressed without obvious tissue specificity based on mRNA transcription and protein translational levels (Duan et al., 2020), although it was reported that TAOK1 and TAOK2 are highly enriched in the brain by Northern blot detection (Hutchison et al., 1998). In the CNS, TAO kinases are widely distributed in all brain regions and the spinal cord without gender differences¹ (Figure 1B). In addition, TAO kinases are modestly expressed but TAOK2 and TAOK3 showed strong expression in the cerebellum (including cerebellar hemisphere) and cortex (including front cortex), implying that the importance of these brain regions are critical for TAO kinase related NDDs.

As TAO kinases belong to the STE20 kinase family, the original functions of TAO kinase are supposed to act as upstream regulators of mitogen activated protein kinases (MAPKs). Upon extracellular stimulations, conventional MAPK cascades including the ERK1/2, p38 MAPKα, β, δ, and γ, and JNK1/2/3 pathways are selectively activated to integrate, amplify, and regulate signal transduction and eventually affect multiple biological processes including cell proliferation, differentiation, migration, and apoptosis (Johnson and Lapadat, 2002; Kyosseva, 2004; Cargnello and Roux, 2011). TAO1 and TAO2 have a lot in common regarding activating p38 MAPK (Hutchison et al., 1998; Chen et al., 1999) and the JNK cascade (Moore et al., 2000; Zihni et al., 2006). TAO3 shows a relatively distinct function that activates ERKs (Zhang et al., 2000) and p38 MAPK but inhibits the JNK cascade (Tassi et al., 1999). It should be noted that TAO3 may inhibit p38 kinases in certain conditions (Tassi et al., 1999). Thus, the TAO kinases family is a central regulator in controlling MAPKs cascades and may be involved in multiple biological processes (Figure 2).

In addition, TAO1 is also known as MARKK (microtubule affinity regulating kinase) due to its ability to phosphorylate MARK, which can further affect microtubule (MT) arrangement by modulating tau (Giacomini et al., 2018) and other related microtubule-associated proteins (MAPs; Drewes et al., 1997), indicating the functional role of TAO1 in regulating MT dynamics. Further research found that TAO1 could interact with Spred-1 to inhibit TESK1, which could modulate actin dynamics by activating Cofilin (Moore et al., 2000). At the same time, TAO2 was also reported to modulate actin filament and MT rearrangement in cultured 3T3 cells (Moore et al., 2000; Mitsopoulos et al., 2003). Although direct evidence remains is scarce, TAO3 might also play a role in the regulation of cytoskeletal dynamics due to the fact that it has been reported that this kinase modulates the JNK signaling (Kapfhamer et al., 2012; Zeke et al., 2016). It is therefore not surprising that Drosophila Tao, the only homolog of mammalian Tao kinases, affects actin and MT dynamics in cultured Drosophila S2 cells (Liu et al., 2010; Pflanz et al., 2015) and primordial germ cells from developing embryos (Pflanz et al., 2015). It should be noted that the MAPK cascade signaling pathway has a tight connection with cytoskeletal dynamics (Šamaj et al., 2004; Komis et al., 2011). Therefore, TAO kinases’ functions may be tightly involved in cytoskeletal dynamics regulation (Figure 2).

A rare de novo microdeletion at 17q11.2, which covers TAOK1, was identified in a patient with a developmental delay and postnatal microcephaly (Xie et al., 2016), suggesting a possible role of TAOK1 to contribute to NDDs. Two genomic microdeletions in the TAOK1 genome (without affecting other genes) in patients with developmental delay (nsv1062993; Cooper et al., 2011), microcephaly, and seizures (Decipher #250045, Decipher database) was reported. Importantly, the deletion nsv1062993 completely overlaps the CNV of #250045, and the two partially overlap the TAOK1 gene. It is assumed that the TAOK1 gene plays a pivotal role in the phenotype of patients (Xie et al., 2016). In addition, TAOK1 was predicted to have a haploinsufficiency score less than 10, suggesting one copy deletion of TAOK1 could cause clinical consequences (Xie et al., 2016). Followed by an integrated meta-analysis that combines de novo mutations from exome sequencing data and CNV morbidity data, TAOK1 with a missense mutation was identified as a candidate for NDDs (Coe et al., 2019). Furthermore, in a large scale whole exome screening (WES) of patients with autism, TAOK1 was identified as one of over 100 putative autism spectrum disorder (ASD) associated genes (Satterstrom et al., 2020). In addition, TAOK1 de novo mutations were identified from eight children with associated NDDs including delayed speech and language development (5/8), autism (2/8), intellectual deficiency (4/8), macrocephaly (3/8), and motor development delay (6/8; Dulovic-Mahlow et al., 2019). Interestingly, half of these de novo mutations locate in the catalytic domain, one truncated mutation is present in the central domain, and the other three are truncated or frameshifted mutations in the regulatory domain. Although the precise functional analysis of these mutations is still lacking, the blood and fibroblast line derived from a patient carrying a variant (c.2366_2367insC) was analyzed. It was found that this variant could decrease cDNA (Sanger sequencing), mRNA (quantitative real time PCR), and protein (western blot) levels of TAOK1. In addition, the mutated mRNA could be stabilized by cycloheximide treatment indicating that nonsense-mediated mRNA decay could explain the reduced abundance of the mutant allele (Dulovic-Mahlow et al., 2019). Interestingly, a recent study collected a cohort of 23 individuals with NDDs from a collaboration facilitated by GeneMatcher (Sobreira et al., 2015) with multiple de novo variants in TAOK1, among which 20 individuals had an intragenic TAOK1 variant and three patients had a chromosomal deletion including TAOK1 (Woerden et al., 2021). Importantly, this study tested several variants with functional assays in a mouse model showing that two variants (c.500T>G and c.943C>T) may act with dominant negative (DN) functions and one variant (c.1643T>C) may have a loss of function (LOF) effect. Thus, these studies provide clear evidence that dysfunction of TAOK1 with either CNVs or mutations with DN or LOF effect will result in NDDs.

TAOK2 is highlighted as a candidate risk gene for NDDs because previous reports showing that microdeletion and microduplication of the 16p11.2 genetic locus are associated with ASDs or schizophrenia (Weiss et al., 2008; McCarthy et al., 2009; Zheng et al., 2013). TAOK2 is one of the 30 genes that are located in the 16p11.2 region. Recently, more psychiatric features like speech/language impairments, intellectual disability, motor/developmental delay, microcephaly and macrocephaly have been identified in patients with 16p11.2 microdeletion and microduplication (Steinman et al., 2016; Rein and Yan, 2020). Although it has been long been proposed that disruption of monogenic TAOK2 function might result in NDDs, the clear clinical evidence has only been presented recently. By performing whole-genome sequencing (WGS) of families with ASD, one frame shift deletion of TAOK2 was identified (Yuen et al., 2017). This was further confirmed in a following study (Richter et al., 2018). Richter et al. (2018) combined WGS and WES to examine over 2,600 families with ASDs and identified 24 different variants in TAOK2, including the frame shift deletion identified by Yuen et al. (2017). They further characterized that three of these variants are known to be de novo including a missense mutation in the catalytic domain (A135P), a frameshift deletion resulting in truncation (P1022*) in the regulatory domain, and a de novo splice site variant (c.563 + 12_563 + 15del) predicted to cause intron seven retention. Among these three de novo mutations, A135P mutation results in lower levels of both phosphorylated and total TAO2 but the P1022* mutation does not affect TAO2 levels in patients with derived lymphoblastoid cells. Combined with more biochemical analysis, A135P and P1022* were confirmed to abolish (LOF) and enhance (GOF) kinase activity, respectively. The impact of the de novo splice site variant (c.563 + 12_563 + 15del), however, is still unknown and requires further characterization (Richter et al., 2018). It should be noted that, patients carrying these de novo mutations showed typical ASDs symptoms and also have a certain degree of disability or delay in language and speech development, implying the broad function of TAOK2 in neurodevelopment.

Although TAOK2 contribution to 16p11.2 CNVs pathophysiology is quite convincing, it should be noted that it is TAOK2 and other genes (e.g., MAPK3, SEZ6L2 and KCTD13) that act collectively to develop much complex and diverse 16p11.2 CNVs phenotypes (Krishnan et al., 2016), compared to individual gene mutations (Richter et al., 2018; Rein and Yan, 2020). In this review, we focus on TAOK2 function only since 16p11.2 CNVs related NDDs have recently been reviewed (Rein and Yan, 2020) and is also out of scope of the current study.

Unlike TAOK1 and TAOK2, studies on the association of TAOK3 with NDDs is relatively scarce. Several GWAS based studies indicate that TAOK3 is likely involved in NDDs. A study with a rare CNVs analysis by WGS suggested that a de novo deletion that affects TAOK3 (and PEBP1) may contribute to schizophrenia (Malhotra et al., 2011) and TAOK3 (but not PEBP1) was further confirmed in a GWAS analysis (Gilman et al., 2012), suggesting LOF of monogenic TAOK3 may contribute to NDDs, at least in schizophrenia. In another GWAS study, TAOK3 was identified as a genetic predisposition to loneliness (Abdellaoui et al., 2019), a status that accompanies NDDs (Kwan et al., 2020; Papagavriel et al., 2020). In addition, two independent GWAS analyses showed that TAOK3 is related to high opioid requirement for patients with advanced cancer pain (Gutteridge et al., 2018) and morphine requirement for postoperative pain in a retrospective pediatric day surgery population (Cook-Sather et al., 2014), suggesting a functional role of TAOK3 in controlling pain under certain conditions. Since the impairment of sensory perception/processing is highly associated with NDDs, like autism (Robertson and Baron-Cohen, 2017), the possible function of TAOK3 in NDDs is further implied. Except for GWAS studies, in a genome-wide-association meta-analysis (GWAMA), TAOK3 was also identified as a susceptive gene for depression (Baselmans et al., 2019). Overall, those studies suggest a possible role of TAOK3 in NDDs.

The direct evidence for monogenic TAOK3 links to NDDs is based on a large scale WES that compared and analyzed over 2,500 affected ASD children and 1,900 unaffected siblings and the parents of each family (Iossifov et al., 2014). In this systematical study, two novel missense de novo mutations TAOK3 (c.1495A>G and c.1894C>T) were identified and validated (Iossifov et al., 2014; Table 1). These two missense mutations encode protein TAO3 with pT199A and pR632W, which are located in the central and regulatory domain, respectively. Although they are likely deleterious mutations as predicted in the gnomAD database, it should be noted that to determine whether these two mutations in TAOK3 are pathogenic or not requires a detailed functional analysis and more clinical samples. Thus, it would be better elucidated if more deleterious mutations of TAOK3 in ASD or other NDDs patients are identified and validated in future studies.

Although TAO kinases at the mRNA level are widely expressed in adult human brains (Figure 1B), TAO2 is preferentially highly expressed in the intermediate zone and the cortical plate of the developing cortex (E18) in mice (de Anda et al., 2012). Interestingly, TAO2α is expressed from early embryonic stages (E10) to adults but TAO2β is only detectable at late stage (from E19) in the mouse brain cortex. Immunostaining with mouse cortical neurons (E17) and cultured 2 days in vitro, TAO2 is found to locate in the growth cone where actin is enriched and its active form pTAO2 was found to localize at the neurite shaft where MT is accumulated (de Anda et al., 2012), suggesting a possible role of TAO2 in regulating neuronal cytoskeletal dynamics. Knockdown of Taok2 decreases basal dendrite arborization and callosal axon projection in the developing cortex (de Anda et al., 2012). Moreover, TAO2 could act downstream of the secreted guidance cue Semaphorin 3A (Sema3A) by interacting with its receptor Neuropilin 1 (Nrp1) to activate the JNK signaling pathway (de Anda et al., 2012).

To further characterize the neurodevelopmental functions of Taok2 in mice, Richter et al. (2018) systematically analyze the neurological phenotypes in Taok2 knockout (KO) mice, which were previously found to control behavioral response to ethanol in mice (Kapfhamer et al., 2013). Magnetic resonance imaging (MRI) of fixed 8- to 10-week-old mouse brains of Taok2 KO and HET (KO/+) brains found that their absolute brain volumes were significantly enlarged compared with wild type (WT) mice but have a relative decrease in the somatosensory cortex and corpus callosum in a gene dose dependent way. Behavioral analysis indicated that the impairments on cognition, anxiety, and social interaction in Taok2 KO mice were consistent with previous clinical studies of ASD patients (Hazlett et al., 2005; Freitag et al., 2009; Hardan et al., 2009; Vaccarino and Smith, 2009; Lai et al., 2014; Sacco et al., 2015). By examining the dendritic morphology in the prefrontal cortex, they found that Taok2 KO and HET mice showed a decrease in basal dendritic complexity, which is comparable to the previous in vitro study (de Anda et al., 2012).

A recent study also clearly showed that overexpression (OE) of human TAO1 and its several NDDs linked variants in cultured primary hippocampal neurons from mice, could significantly reduce dendritic arborization and length (Woerden et al., 2021) although the detailed mechanisms remain to be further explored. Thus, to obtain more physiological evidence, further exploration of its function in neuronal/dendrite development, by generating Taok1 KO mice, is required.

We recently also found that Drosophila Tao could significantly affect peripheral dendritic development of dendritic arborization (da) neurons at whole larval developmental stages and in adults (Hu et al., 2020). da sensory neurons are a well-characterized model system to study neuronal development and functions (Grueber et al., 2002; Williams and Truman, 2004; Shimono et al., 2009; Jan and Jan, 2010; Im and Galko, 2012; Copf, 2015). Using this model, many of the NDDs related genes and their molecular mechanisms have been confirmed and identified (Gatto and Broadie, 2011; Coll-Tane et al., 2019). We found that Drosophila Tao is expressed in all da sensory neurons and its active form pTao is discretely distributed along dendrites, suggesting a local function of Drosophila Tao in regulation of MT and/or actin to affect dendritic development. Loss of Drosophila Tao increases dendritic complexity and MTs dynamic of all da sensory neurons in vivo in a developmental stage dependent way (Hu et al., 2020), which is the opposite to the phenotype of neocortex neurons from an in vitro culture system (de Anda et al., 2012) or TAOK2 KO mice (Richter et al., 2018). Interestingly, this increased dendritic complexity phenotype could be fully rescued by wild type human TAO2 but not by the ASD linked LOF mutation (A135P), suggesting a conserved function of Drosophila Tao and TAO2; but its readout varies in different model systems (King and Heberlein, 2011). Moreover, we also found that disruption of Drosophila Tao in adult sensory neurons caused dendritic over-branching and resulted in impairment of social behaviors, which further confirmed the previous observations that sensory perception is critical for developing ASDs in a mouse model (Orefice et al., 2016, 2019).

Altogether, these different in vitro and in vivo models clearly showed a critical and complex role of TAO kinases in regulating dendrite development which then affects normal brain functions.

Both dendrite arborization and synapse formation are critical for wiring the neural circuitry and establishing normal neural functions (Benson et al., 2001; Jan and Jan, 2001; McAllister, 2007; Colón-Ramos, 2009; Batool et al., 2019). Yadav et al. found that TAO2 localizes to dendritic spines and is required for synaptic maturation in a kinase activity dependent way (Yadav et al., 2017). Combined with an elegant chemical-genetic method and mass spectrometry, the authors identified several candidate substrates of TAO2 including Septin6, Septin7, HADC6, Bai1, Caskin1, CEP170, and MST3, and only Septin7 was further confirmed in a subsequent functional analysis. Septin7 is a GTP-binding protein (Neubauer and Zieger, 2017) that regulates actin (Hu et al., 2012; Mavrakis et al., 2014) and MT remodeling (Bowen et al., 2011; Hu et al., 2012) to control axon/dendrite branching and spine morphology (Xie et al., 2007; Hu et al., 2012). It was found that TAO2 could directly phosphorylate Septin7 and lead to its trafficking to the dendritic spine where it could associate with and immobilize the synapse scaffolding protein PSD95 to promote spine synapse maturation (Yadav et al., 2017). In addition, a peptide pull-down method to identify binding proteins in neuronal lysates labeled by stable isotope labeling by amino acids in culture (SILAC; Zhang et al., 2011; Deng et al., 2019) was employed to identify possible binding components of TAO1/2, showing that Myosin Va could interact with TAO1/2 in a phosphorylation dependent manner. Moreover, endogenous Myosin Va could bind endogenous TAO1 and be phosphorylated by TAO1/2 in neurons (Ultanir et al., 2014). Myosin Va is a motor protein in charge of the intracellular transport of vesicles, organelles, and protein complexes along the actin filaments (Harrington and Rodgers, 1984; Masters et al., 2016; Guhathakurta et al., 2018; Lombardo et al., 2019) and affects microtubule based transport when recruited on the same cargo with the microtubule motor kinesin (Kapitein et al., 2013), suggesting a role of Myosin Va in promoting synaptic formation/maturation when localized in the spine (Ultanir et al., 2014). In addition, the authors found that TAO1/2 could be phosphorylated by mammalian sterile 20 (Ste20)-like kinase 3 (MST3), a homolog to Drosophila Hippo (Harvey et al., 2003). Interestingly, in contrast to the downstream of mammalian Tao Kinase to MST3, it seems Drosophila Tao is an upstream signaling component activating Hippo to regulate tissue growth (Boggiano et al., 2011; Poon et al., 2011, 2018; Huang et al., 2014; Chung et al., 2016). However, whether Drosophila Tao could phosphorylate Hippo to regulate synaptic growth needs to be further explored.

In another study, it was shown that TAO2β instead of TAO2α is essential for activity-induced dendritic spine formation (Yasuda et al., 2007), which rises a possibility that TAO kinases function in an isoform-dependent way. Electroconvulsive or other excitatory stimuli in cultured hippocampal neurons triggers a protocadherin arcadlin that stimulates TAO2β specifically, which in turn activates p38 MAPK through MEK3, resulting in the endocytosis of N-cadherin and the decrease in spine numbers (Yasuda et al., 2007; Sun and Xie, 2012), suggesting an isoform specific function of TAO kinases in spine development. This finding was further confirmed in the Taok2 KO mouse model (Richter et al., 2018). It was shown that dendritic spines in hippocampal neurons from Taok2 KO mice were dramatically decreased compared to WT, possibly by directly affecting the RhoA activity, a kinase which is also preferentially combined to TAO2β and is highly involved in the regulation of spine mobility. Those data indicate that the C-terminal of distinct TAO kinases is critical for their unique physiological functions.

Using Drosophila (neuron-muscle junction) NMJ as a classic system for studying synaptic development (Broadie and Bate, 1995; Menon et al., 2013; Frank, 2014) assists neurobiologist in identifying multiple psychiatric disorder related genes and their underlying molecular mechanisms (Sun and Xie, 2012; Tian et al., 2017). Knockdown or deactivation (hypomorphic allele) of Drosophila Tao increases the number of NMJ (buttons, the single NMJ structure; Politano et al., 2019). However, it seems Drosophila Tao that regulates NMJ development is not dependent on the Hippo/MST pathway as it is in hippocampal neurons (Ultanir et al., 2014) but it could negatively regulate BMP signaling as reduction of Drosophila Tao leads to an increase in both nucleic pMad levels and BMP target gene expression in motor neuron (Politano et al., 2019). However, another study recently showed that knockdown of Tao could decrease the NMJ number (Dulovic-Mahlow et al., 2019). Thus, both studies indicated the important role of Drosophila Tao in synaptic development, while the underlying mechanisms for the opposite phenotypes observed independently remains to be illustrated.

We recently developed a novel model for studying synaptic development in Drosophila larvae (Tenedini et al., 2019) to compensate for the NMJ model. At larvae stage, peripheral Class IV da (C4da) sensory neurons directly come into contact with a pair of interneurons A08n that can form a functional synaptic structure (Town et al., 2014; Hu et al., 2017; Kaneko et al., 2017), which is closer to mammalian synapses when compared with the classic NMJ synaptic system. By using the Syp-GRASP technique (Macpherson et al., 2015) to label C4da-A08n synapses, we found that loss of Tao results in exuberant postsynaptic (comparable to spine structure) specializations and aberrant connectivity during larval growth. Using functional imaging and a behavioral analysis we showed that loss of Drosophila Tao could induce ectopic functional synapses formation of the A08n neuron with other types (C3da) of neurons and resulted in altered behavioral responses in a connection-specific manner (Tenedini et al., 2019), indicating that TAO kinase mutations, like other NDDs susceptive genes, can induce abnormal behaviors partially from improper establishment of neural circuits (Kida and Kato, 2015; Kaiser et al., 2017; Südhof, 2017).

The association between neuronal apoptosis and NDDs is not well documented. A previous study showed that overexpression (OE) of the full length or kinase domain of human TAO1 in human neuroblastoma SH-SY5Y cells resulted in cellular apoptosis which was indicated by elevated caspase-3 activity. However, OE of the regulatory domain of TAO1 in SH-SY5Y did not appear to have an obvious change (Wu and Wang, 2008). Importantly, OE of TAO1 induced elevated caspase-3-like activity and apoptosis of neuroblastoma could be reduced by JNK inhibitor SP600125 to some extent. In addition, OE of rat TAO3 in PC12 (a widely used cell line with the properties of intersecting neurons) resulted in elevation of the expression level of BimEL (Wakabayashi et al., 2005), a protein with apoptotic activity. Recently, it was also reported that TAO1 protected MCAO-induced cerebral ischemic stroke by decreasing the pro-inflammatory factors and apoptosis via PI3K/AKT and MAPK signaling pathways (Li et al., 2019), which further indicates the critical roles of TAO kinases in neuronal apoptosis. Those preliminary studies suggest an underestimated function of TAO kinases in neuronal apoptosis.

Furthermore, TAO2 may play roles in the regulation of neuronal maturation. A recent study showed that BDNF could regulate cortical GABAergic interneuron maturation in a TAO2-JNK signaling pathway in a 16p11.2 duplication mouse model (Willis et al., 2020). Cultured neurons from 16p11.2 duplication mice exhibit an abnormal interneuron developmental phenotype that may be involved in a premature closure of the critical period which is likely to be driven by OE of TAO2 and the subsequent over-activity of JNK, since pharmacological inhibition of TAO kinase could alleviate the 16p11.2 duplication phenotype. Given the importance of parvalbumin (PBV+) interneurons in the regulation of excitatory/inhibitory balance within cortical regions, accelerated GABAergic development by TAO2 over-activity may lead to dysregulated network activity and synaptic connectivity (Willis et al., 2020). In addition, it was reported that OE of human TAO1 could also prevent maturation of cultured primary hippocampal neurons from mice (Woerden et al., 2021), while the role of JNKs is involvement is yet to be determined.

A recent study investigated a possible function of TAOK1 in neuronal migration, a process that is critical for normal brain development (Woerden et al., 2021). By employing the in utero electroporation in mice at embryonic day 14.5 (a well-established time window when immature neurons generated from progenitor cells start to migrate to their final destination within the cortical plate, which will eventually form the cerebral cortex layer 2/3), the authors found that knockdown of Taok1 resulted in a clear migration deficit of the neurons when compared to control neurons from postnatal day 1 (p1) to P7, when only 75% of the Taok1 knockdown neurons were present in cerebral cortex layer 2/3, compared to 95% in control conditions (Woerden et al., 2021). Similar migration deficits were observed when transfection of several NDDs, linked human TAOK1 variants in developing mouse brains including c.500T>G and c.943C>T (see Table 1), suggesting a possible role of TAOK1 in neuronal migration during early human brain development.