Neurons are polarized cells with an axon that can extend several thousand times the diameter of the cell body. Moreover the axon is not a homogeneous expanse but may include thousands of discrete subdomains, such as Nodes of Ranvier and presynaptic terminals, which have complex, specialized protein compositions. Neuron polarity and axonal subdomains are established during development of the nervous system during which neural precursors differentiate and undergo neurite outgrowth with the axon and dendrites extending outwards from the cell body. Membrane expansion is preferentially at the axon tip, which becomes progressively further from the cell body, requiring active delivery of membrane, proteins, and organelles by fast axonal transport (FAT) (Pfenninger and Friedman, 1993; Craig et al., 1995; Futerman and Banker, 1996). It has long been known that blocking axonal transport of vesicles inhibits axon growth and induces growth cone collapse (Martenson et al., 1993). Furthermore kinesin mutants, such as KIF5A^−/− neurons, have reduced outgrowth (Karle et al., 2012). Despite the importance of FAT and selective delivery of cargos to specific locations, the mechanisms for physiological regulation of cargo transport and delivery have remained elusive.

Tau is a microtubule associated proteins found principally in neurons. It was originally isolated with microtubules from brain and was described as essential for microtubule assembly in vitro (Weingarten et al., 1975). The prevailing dogma continues to define the major function of tau in vivo as microtubule stabilization even though several lines of evidence suggest this is not the case. Most importantly, tau null mice have deficits in neurite outgrowth and neuronal migration both in vitro and in vivo while still possessing an extensive microtubule network (Harada et al., 1994; Dawson et al., 2001; Sennvik et al., 2007; Sapir et al., 2012). Furthermore, tau expression is correlated with neuronal maturation (Fiock et al., 2020) and tau protein is enriched on dynamic microtubules at the end of growing neurites (Black et al., 1996). In vitro, wild-type (WT) hippocampal neurons follow a stereotypical timeline of outgrowth (Bartlett and Banker, 1984; Dotti et al., 1988; Fletcher and Banker, 1989). After a brief multipolar stage whereby several short processes grow equally, one process begins to grow rapidly to become the axon while other processes remain stalled. 2–3 days later the short processes restart growth at a slower rate elongating and branching to become dendrites (Dotti et al., 1988). Tau knockdown in primary cerebellar neurons prevented neurite elongation with cells remaining in the early multipolar stage until tau expression was reinstated (Caceres and Kosik, 1990; Caceres et al., 1991). Tau knockout hippocampal neurons also had reduced axonal and dendritic elongation (Dawson et al., 2001).

During studies evaluating biological activities of normal (Morfini et al., 2007) and pathological tau (LaPointe et al., 2009; Kanaan et al., 2011), a phosphatase activating domain (PAD) at the N-terminal of tau was identified. When exposed, the PAD activates a signaling cascade involving protein phosphatase 1 (PP1) and glycogen synthase three beta (GSK3β) (LaPointe et al., 2009; Kanaan et al., 2011). PP1 activates GSK3β by dephosphorylating pSer9 on the inactive form of the kinase. Activation of GSK3β leads to phosphorylation of kinesin light chain (KLC) and trigger the release of vesicle cargoes (Morfini et al., 2002; Kanaan et al., 2011). Exposure of the PAD has been shown to be an early event in multiple tauopathies and is a major contributing factor to neurodegeneration (Kanaan et al., 2011; Kanaan et al., 2012; Cox et al., 2016; Kanaan et al., 2016). However, the presence of a sequestered PAD in normal tau raised the possibility that this domain may have a normal physiological function. The exclusive effect of tau PAD on anterograde FAT raised the interesting possibility for physiological regulation of cargo delivery through transient activation of PP1 and GSK3β. Significantly, there is already evidence that local control of PP1 and GSK3β occurs at sites which require cargo delivery. For example, during axonal outgrowth active GSK3β is enriched at growth cones (Morfini et al., 2002; Morfini et al., 2004). Additionally, it is well known that neurofilaments are highly phosphorylated throughout axons except at nodes of Ranvier (Hsieh et al., 1994; Martini, 2001). This has been attributed to a higher activity of phosphatases in this area, but again the physiological function of this local activation and the mechanism for regulating it are unknown (de Waegh et al., 1992; Hsieh et al., 1994). We hypothesize that activation of a PP1-GSK3β signaling pathway by regulated site-specific PAD exposure is a mechanism for cargo delivery, and is thereby important for neurite outgrowth of developing neurons.

Previously our lab identified a phosphatase activating domain comprising amino acids 2–18, the PAD, at the N-terminal of tau. Exposure of the PAD leads to activation of a PP1-GSK3β signaling pathway resulting in inhibition of anterograde FAT (LaPointe et al., 2009; Kanaan et al., 2011). Studies have demonstrated that PAD exposure is an early event in multiple tauopathies, co-localizing with tau oligomers and AT8 immunoreactivity (Kanaan et al., 2011; Kanaan et al., 2012; Cox et al., 2016; Kanaan et al., 2016), but the question of whether the PAD also has a physiological role was unknown. The selective inhibition of anterograde transport led us to propose that transient, site-specific exposure of the N-terminal PAD, and activation of the PP1/GSK3β signaling cascade, is a mechanism for delivery of select MBO cargos in axons and would therefore be important for neuronal development. In this paper we have established that there is selective exposure of the PAD during hippocampal neuron development at sites of cargo delivery, comparable to the localization of pT205 tau and active dephospho-Ser9-GSK3β. In addition, we have demonstrated that tau PAD exposure and phosphorylation of T205 have a functional role during neurite outgrowth.

Analysis of primary hippocampal cultures by immunostaining revealed that the PAD is exposed throughout development at specific neuronal subdomains such as growth cones. At both early and later developmental timepoints, TNT1 immunoreactivity was observed at the end of growing neurites and within the central portion of growth cones (Figures 2B,C,E) similar to active GSK3β (Morfini et al., 2002; Morfini et al., 2004). Previous studies from our lab concluded that active GSK3β is enriched in growth cones through use of total and inactive phospho-Ser9-GSK3β antibodies (Morfini et al., 2002; Morfini et al., 2004). Using a recently developed antibody which specifically recognizes active dephospho-Ser9-GSK3β (Grabinski and Kanaan, 2016) we were able to confirm that the previous findings that dp-GSK3β is localized in the central region of growth cones (Figures 2A, D) as well as the cell body (Figure 4).

In early development (2 DIV), TNT1 immunoreactivity was enriched along the length of the growing axon but was observed at much lower levels in dendrites and axonal branches (Figure 3A). By seven DIV, PAD exposure was confined to the growth cone regions of the axon but was also visible in some, but not all, dendrites (Figure 3C). This temporal change in localization is significant as hippocampal neurons in culture follow a stereotypical pattern of outgrowth. After a brief multipolar stage, one neurite begins to elongate rapidly whilst the other processes remain quiescent or retract slightly. On approximately day 5 of culture the minor processes resume elongation and acquire dendritic morphology (Fletcher and Banker, 1989). The absence of TNT1 immunoreactivity in minor processes at 2 DIV but its appearance by 7 DIV suggests that the PAD is specifically exposed in actively growing neurites. Moreover these results suggest that, at least during early stages of development, exposure of the PAD is acting in both developing axons and dendrites. Our previous studies found that GSK3β preferentially phosphorylates KLC2 which can associate with homodimers of any of the kinesin heavy chain isoforms (KIF5A, B, or C) to form the conventional kinesin complex (Morfini et al., 2002; DeBoer et al., 2008). Accordingly, the PP1/GSK3β signaling cascade downstream of PAD exposure would be expected to directly target a subset of anterograde FAT cargoes transported by conventional kinesin isoforms containing KLC2. KIF5 motors have been suggested to have functions in both axons and dendrites (Setou et al., 2002; Kanai et al., 2004; Hirokawa et al., 2009), although other studies have found that KIF5 is excluded from dendrites (Jacobson et al., 2006; Karasmanis et al., 2018) and recombinant KIF5-GFP expressed in primary hippocampal neurons preferentially localized to axons (Nakata and Hirokawa, 2003). Conventional kinesin is thought to be involved in transport of mitochondria which are required through the neuron and are especially essential during active outgrowth (Tanaka et al., 1998; Iworima et al., 2016; Smith and Gallo, 2018). There could also be indirect actions of PP1 or GSK3β activation on other kinesin motors. In addition, GSK3β is known to have other phosphorylation substrates in the growth cone such as stathmin 3 and microtubule associated protein 1B (MAP1B) which are involved in modulation of microtubule dynamics (Trivedi et al., 2005; Devaux et al., 2012).

In more mature hippocampal neurons (10 DIV), PAD exposure became limited to domains along the axon (Figure 3D) and there was no TNT1 immunoreactivity in MAP2-positive dendrites. PAD exposure was most prominent in small regions of the axon where it was crossing over a cell body or dendrite (Figure 3D). This is noteworthy as these subdomains are likely to be sites of nascent synapses. These are fluid structures that form rapidly upon axo-dendritic contact in culture (Ahmari and Smith 2002; Washbourne 2015). Interestingly, the protein components needed to form a stable, mature synapse are transported in two classes of vesicles which both display increased pausing at nascent synapses (Ahmari et al., 2000; Bury and Sabo, 2011). Furthermore, experiments indicated that this pausing is most likely due to both vesicle types responding to a common signal (Bury and Sabo, 2011). Pre-synaptic membrane proteins such as synapsin-I, SNAP25, syntaxin-1B have been shown to be transported by conventional kinesin (Szodorai et al., 2009). These results suggest that exposure of the PAD and activation of PP1-GSK3β may facilitate delivery of presynaptic components at nascent synapses.

Since the N-terminal PAD was only detectable at specific subdomains of developing hippocampal neurons, it follows that there must be strict regulation to control its exposure. Tau conformation is known to be regulated through phosphorylation and we have also shown that tau phosphorylation can regulate downstream signaling pathways (Jeganathan et al., 2008; Bibow et al., 2011; Kanaan et al., 2011; Morris et al., 2021). Building on data from Morris et al., 2021 where we demonstrated that pseudophosphorylation of hTau40-T205E is sufficient to inhibit anterograde FAT through PAD exposure and GSK3β activation, we explored the localization of T205 phosphorylation in primary hippocampal neurons observing similar localization to PAD exposure throughout development (Figure 3; Supplementary Figure S1). GSK3β was more widely distributed, but the active form of GSK3β exhibited enrichment in cell regions similar to TNT1 and pT205 (Figure 4). The localization of pT205 at sites of PAD exposure, combined with previous results, supports the hypothesis that phosphorylation of T205 is linked to exposure of the PAD at specific sites during development. This hypothesis is further supported by the dynamics of T205 phosphorylation. pT205 is rapidly dephosphorylated in mouse brain lysates with over 50% of the signal lost within the first 60 s after death indicating this is a dynamic phosphorylation site (Wang et al., 2015). Moreover, unlike most tau phosphoresidues, pT205 is preferentially dephosphorylated by PP1 (Liu et al., 2005) suggesting there may be negative feedback loop for PAD exposure. This is essential as overexposure of the PAD is linked to multiple tauopathies (Kanaan et al., 2011; Kanaan et al., 2016; Combs and Kanaan, 2017) and overexpression of 6D tau, an N-terminal tau construct that cannot fold, in primary hippocampal neurons results in axon degeneration (Morris 2021).

The location of PAD exposure and T205 phosphorylation during hippocampal neuron development is consistent with our hypothesis that site-specific, regulated exposure of the PAD is a mechanism for delivery of specific MBOs from kinesin. Since axonal transport is essential for neurite outgrowth, we investigated the function of PAD exposure and T205 phosphorylation in tau knockout neurons. As previously shown (Dawson et al., 2001), we found tau knockout neurons to have inhibited outgrowth compared to WT neurons (Figure 5; Supplementary Figure S2). Transfection of full length hTau40-WT into tau KO hippocampal neurons significantly increased their axonal and dendritic outgrowth compared to untransfected cells (Figure 6, Supplementary Figures S3D, E). By contrast transfection of hTau40-ΔPAD or hTau40-T205A did not increase outgrowth compared to untransfected tau KO neurons (Figure 6, Supplementary Figures S3D, E). These results clearly demonstrated that exposure of the N-terminal PAD during hippocampal neuron development is important for normal neurite outgrowth. Furthermore, phosphorylation at T205 appears to be necessary for exposure of the PAD. The deficit and improvement of dendritic outgrowth also supports a role for PAD exposure and T205 phosphorylation in dendritic growth cones as was suggested by TNT1 and pT205 immunostaining.

Interestingly, tau KO hippocampal neurons also had deficits in axonal and dendritic branching compared to WT neurons (Figure 5; Supplementary Figures S2F–H) which were not significantly improved by hTau40 transfection (Figure 6; Supplementary Figures S3 F–H). However, the total number and total length of branches was increased by hTau40-WT transfection although it did not reach significance. This appears to have been driven by a slight increase in the length of dendritic branches since the length of axonal branches was unchanged by transfection (Figure 6G; Supplementary Figure S3G). One reason for the limited increase in branch number and length may be the tau isoform used for transfection. The longest isoform of tau (hTau40; 2N4R) was selected to follow on from previously published experiments (Kanaan et al., 2011; Morris et al., 2021). However, in mice, fetal tau primarily consists of the 0N3R isoform until after birth, at which time expression gradually switches to 4R isoforms (Takuma et al., 2003). Moreover, even after this developmental switch, 2N4R is restricted to the soma and does not become enriched in the axon until adulthood (Liu and Götz, 2013). Further experiments will be required to explore the function of tau in neurite branching, as well as the implications of isoform specific functions.

In vivo, tau knockout in mice appears to cause very few overt deficits. Initially the only noted difference was a decreased number and density of microtubules in unmyelinated cerebellar parallel fibers (Harada et al., 1994). However, more rigorous testing showed that tau knockout mice had deficits in motor coordination as well as learning impairments in contextual fear conditioning (Ikegami et al., 2000). By contrast, in humans, tau deficiency is implicated in neurodevelopmental delay. Patients with microdeletion at chromosome 17q21.31, which contains the MAPT gene, all display learning difficulties amongst other symptoms (Shaw-Smith et al., 2006; Dubourg et al., 2011). Acute knockdown of tau in mice also had striking phenotypes of inhibited neurite outgrowth in vitro and inhibition of neuronal migration and dendritic development in vivo (Caceres and Kosik, 1990; Caceres et al., 1991; Sapir et al., 2012). Some functions of tau appear to be compensated by MAP1B in axons and MAP2 in dendrites (DiTella et al., 1996; Takei et al., 2000; Liu et al., 2019). In particular, tau and MAP1B appear to act synergistically as double knockout mice had exacerbated inhibition in axonal elongation and neuronal migration (DiTella et al., 1996; Takei et al., 2000). Neither MAP1 nor MAP2 contain the PAD, however they could have similar scaffolding functions. For example, PP2A/Bα and Fyn interact with proline rich motifs in both MAP2 and tau (Sontag et al., 2012). Furthermore, perfusion of MAP2c into the squid axoplasm model can inhibit anterograde FAT, identical to tau PAD, through an unknown mechanism (S.T. Brady, unpublished data). A previous study in our lab found that inhibition of Cdk5 can also result in activation of PP1-GSK3β which may represent an independent pathway to regulate cargo delivery in growth cones (Morfini et al., 2004).

Overall the results presented in this paper support our hypothesis that activation of a PP1/GSK3β by regulated site-specific exposure of the N-terminal tau PAD is a mechanism for cargo delivery at select subdomains such as growth cones and nascent synapses and is thereby important during neurite outgrowth. Efficient FAT is essential for development and maintenance of neurons, and the selective delivery of cargoes is likely to involve multiple intersecting mechanisms. It was reported that there is local control of PP1 and GSK3β activity at sites of cargo delivery but mechanisms for activation were unknown (de Waegh et al., 1992; Hsieh et al., 1994; Morfini et al., 2002; Morfini et al., 2004). Exposure of the PAD represents one pathway in developing neurons.