The KLDpT activation loop motif is critical for MARK kinase activity

MAP/microtubule-affinity regulating kinases (MARK1-4) are members of the AMPK family of Ser/Thr-specific kinases, which phosphorylate substrates at consensus LXRXXSXXXL motifs. Within microtubule-associated proteins, MARKs also mediate phosphorylation of variant KXGS or ζXKXGSXXNΨ motifs, interfering with the ability of tau and MAP2/4 to bind to microtubules. Here we show that, although MARKs and the closely related salt-inducible kinases (SIKs) phosphorylate substrates with consensus AMPK motifs comparably, MARKs are more potent in recognizing variant ζXKXGSXXNΨ motifs on cellular tau. In studies to identify regions of MARKs that confer catalytic activity towards variant sites, we found that the C-terminal kinase associated-1 (KA1) domain in MARK1-3 mediates binding to microtubule-associated proteins CLASP1/2; but this interaction is dispensable for ζXKXGSXXNΨ phosphorylation. Mutational analysis of MARK2 revealed that the N-terminal kinase domain of MARK2 is sufficient for phosphorylation of both consensus and variant ζXKXGSXXNΨ sites. Within this domain, the KLDpT activation loop motif promotes MARK2 activity both intracellularly and in vitro, but has no effect on SIK2 activity. As KLDpT is conserved in all vertebrates MARKs, we conclude that this sequence is crucial for MARK-dependent regulation of cellular polarity.


Introduction
5' AMP-activated protein kinase (AMPK) family members share a high degree of sequence similarity within the N-terminal kinase domains [1], but their regulatory C-termini are far more divergent. Phosphorylation of a conserved threonine residue within the activation loop is necessary for the activation of all AMPK family members, whereupon they regulate various cellular processes including metabolism, growth, and polarity [2]. The tumor suppressor liver kinase B1 (LKB1 or STK11) activates all 13 AMPK family members: AMPKα1/α2 (PRKAA1/ 2), BRSK1/2, NUAK1/2, SIK1-3, MARK1-4 [3], as well as the related SNRK [4]. With the exception of AMPKs and BRSKs, the interaction with LKB1 is mediated by the ubiquitin-associated (UBA) domain, which is located directly adjacent to the kinase domain [5]. Some AMPK subfamilies can also be activated by other signal-dependent kinases. For example, Forskolin (Sigma-Aldrich), 10 mM Doxorubicin (Cayman Chemical; light protected), and 2 mM Carfilzomib (PR-171; Selleck Chemicals).
MARK2 truncation constructs and plasmid backbone exchanges (3xFLAG / 3xHA) were archived by restriction cloning. Single and cumulative site-directed mutagenesis were used to introduce point mutants. The MARK2 and SIK2 hybrid proteins, the ΔMSL mutant, and the MARK2 SIK2 AL (activation loop) mutants were generated by Fusion PCR, which was performed as previously described [36].

Overexpression & immunoprecipitation (IP)
Experiments were performed in 6 well plates by reverse transfecting HEK293T cells (
Protein and peptide identification and protein quantitation were done with Integrated Proteomics Pipeline-IP2 (Integrated Proteomics Applications). Tandem mass spectra were extracted from raw files using RawConverter [38] with monoisotopic peak option. Peptide matching was done against a Uniprot H. sapiens protein database with reversed sequences and recombinant proteins using ProLuCID [39,40] with a fixed modification of 57.02146 on C and differential modification of 79.9663 on STY. Peptide candidates were filtered using DTA-Select [38,41].

Immunofluorescence
HEK 293T cells (0.75 × 10 6 cells) were plated in Poly-D-Lysine coated glass bottom dishes (MatTek Corporation) and reverse transfected with Lipofectamine1 2000 (Invitrogen) using 1 μg of MARK2 and SIK2 plasmid DNA (pUbC-3xFLAG backbone with IRES GFP). 24 h post-transfection cells were fixed with 4% paraformaldehyde and incubated with primary antibodies (FLAG M2). Samples were incubated with secondary antibodies conjugated with Alexa Fluor1 -568 (goat anti-mouse) and in certain experiments counterstained with DAPI (Cayman Chemical Company) before image acquisition was performed (LSM 710; Carl Zeiss). Representative images from at least two independent experiments are shown.

Luciferase reporter assays
The luciferase reporter plasmid contains the EVX promoter fragment (220 bp, 2x CRE halfsites) in the pGL4 backbone (Promega) [10]. Luciferase reporter assays were performed in 96 well plates by reverse transfecting HEK293T cells (100,000 cells). For each well 80 ng of DNA was used: 10 ng of EVX reporter plasmid, 10 ng of N-terminal FLAG-tagged CRTC3 plasmid, 20 ng of N-terminal FLAG-tagged MARK2/SIK2 plasmids, and 40 ng of empty pUbC-MCS-His 6 -TEVsite-3xFLAG-STOP-IRESeGFP (MCS = multiple cloning site) plasmid [10]. 24 h post transfection 10 μM Forskolin was added and cells further incubated for 4 h. DMSO served as the control treatment (each well 1% DMSO final). Next, 10 μl of Bright-Glo™ (Promega) was added per well and luciferase activity measured in a GloMax1 multi microplate reader (Promega). The reporter assay was performed twice, but representative data is shown.

Sequence alignment
Amino acid sequences were aligned using MegAlign and Clustal W method (DNASTAR v7).

Statistical analysis
Data are either presented as the mean ± SEM. or ± SD. Statistical analysis was performed using Microsoft Excel (Microsoft Corporation) or PRISM (GraphPad) and graphical presentations were generated using SigmaPlot (Systat Software Inc.) or PRISM.

Comparison of MARK2 and SIK2 activities on cellular zXKXGSXXNC motif phosphorylation
The closely related SIK and MARK families phosphorylate conventional AMPK substrate motifs within CRTCs (CRTC1-3) and class IIA HDACs (HDAC4, 5,7,9) (Fig 1A). By contrast, phosphorylation of tau and MAPs at KXGS motifs appears to be more MARK-selective [19]. Closer inspection of the KXGS motifs in tau and MAP2/4 reveals that all KXGS sites contain a Asn at position +3, and 10 of the 11 sites have a hydrophobic residue at position +4 (C = I, L, V). In contrast to the LXRXXSXXXL AMPK consensus, the KXGS motifs have a hydrophilic residue (z = Q, T, S, R, K) in place of the characteristic hydrophobic residue at position -5, resulting in a refined zXKXGSXXNC motif.
Recent studies have implicated other AMPK family members, including the SIKs [30], in the phosphorylation of these variant zXKXGSXXNC sites. SIKs and MARKs show considerable sequence homology within the N-terminal kinase and UBA domains; and both kinase families are activated by LKB1 phosphorylation (Fig 1B). By contrast, the C-terminus of SIK2 enables inhibition by cAMP/PKA while the C-terminus of MARK2 mediates KA1-dependent and aPKC-regulated membrane localization (Fig 1B and 1C). Based on these properties, we considered that MARK-specific regions might modulate zXKXGSXXNC-directed specificity by bringing tau and MAP2/4 substrates into physical proximity with the catalytic domain. To test this notion, we expressed full-length FLAG-tagged MARK2 and SIK2 in HEK 293T cells and evaluated protein-protein interactions by immunoprecipitation-mass spectrometry (IP-MS) ( Fig 1D). Across all experiments, only MARK2 showed strong interactions with CLIP-associating proteins 1 and 2 (CLASP1/2), a family of microtubule-binding proteins [46]. We confirmed this result by co-immunoprecipitation (co-IP) of cells overexpressing MARK2 or SIK2, followed by detection of endogenous CLASP1 using Western blot analysis ( Fig 1E).
Regulatory differences of MARKs and SIKs are encoded by their divergent C-terminal sequences, which dictate membrane localization and cAMP sensitivity, respectively [11,14].
To determine the impact of these regions on catalytic activity, we generated hybrid proteins exchanging the regulatory domains of MARK2 and SIK2 (Fig 2A). We employed a cyclic-AMP response element (CRE)-based reporter assay to evaluate the repressive activity of these MARK2/SIK2 hybrid proteins on CRTC-dependent luciferase gene expression [49] (Fig 2B   Fig 1. Comparison of MARK2 and SIK2 reveals differences in zXKXGSXXNC motif phosphorylation and interactions with CLASP proteins. A) Confirmed and putative phosphorylation sites of MARKs and SIKs [9][10][11][12][13]45]: Common AMPK family substrate motifs in cAMP-regulated transcriptional coactivators (CRTCs: CRTC1-3) and class IIA Histone deacetylases (HDACs: HDAC4,5,7,9) as well as zXKXGSXXNC motifs in tau (numberings are according to Homo sapiens CRTC/HDAC proteins, Tau-441, and isoform 1 of MAP2/4). # indicates motifs that were previously shown to be phosphorylated by both MARKs and SIKs. B) To scale schematic depiction of H. sapiens MAP/microtubule affinity-regulating kinase 2 (MARK2 1-788) and salt-inducible kinase 2 (SIK2 1-926). MARK2 and SIK2 share homology inside their N-terminal kinase domain (KD) and are both activated by LKB1, but encode distinct C-termini. C) Representative immunofluorescence image of HEK 293T cells transfected with FLAG-tagged MARK2 and SIK2, which were expressed from the constitutive Ubiquitin C promoter (UbC). These plasmids coexpress EGFP via an internal ribosome entry site (IRES). Cells were stained for the FLAG epitope and either imaged for EGFP (MARK2) or counterstained with DAPI (SIK2). (scale bar indicates 20 μm). D) Tables show the recovery of N-terminally FLAG-tagged MARK2 (F-MARK2; hMARK2 TV4) and SIK2 (F-SIK2 AA, with AA = S358A S587A) subjected to mass spectrometry (MS) analysis following the co-IP (SPCs = spectral counts). Recovery of endogenous 14-3-3 ε (YWHAE) and CLIP-associating proteins (CLASP1/2) are indicated. Samples were either treated with the adenylyl cyclase activator Forskolin (Fsk; 10 μM for 1h) or the DNA-damaging agent Doxorubicin (DOX, 10 μM for 4h) prior to cell lysis and IP-MS protocol. In case of SIK2 AA one sample was not only subjected to trypsin but also to pepsin proteolytic digest. E) Western blot analysis of the co-IP of FLAGtagged GFP, MARK2, and SIK2 with endogenous CLASP1 and 14-3-3 proteins. Prior to the co-IP, cells were treated for 1 h with either DMSO or 10 μM Fsk for 1h. F) Western blot analysis of the corresponding input, resolving MARK activation loop phosphorylation (P-MARK AL; blue arrow indicates endogenous HEK 293T cell signal at~85 kDa) as well as phosphorylation of AMPK and zXKXGSXXNC motifs inside CRTCs and tau. EGFP is coexpressed via an IRES from all expression plasmids. and 2C). The MARK2 kinase domain was inactive upon fusion with the C-terminus of SIK2 (M2-S2), because neither M2-S2 nor its respective activation loop mutant (T208A S212A [50]) had any effect on CRE activity. In contrast, the SIK2 kinase domain remained active following fusion to the C-terminus of MARK2 (S2-M2), where it efficiently repressed CRE reporter activity. Consistent with our previous results that C-terminal phosphorylation by PKA and 14-3-3 binding inhibits SIK2 catalytic activity [11], the S2-M2 hybrid was active in repressing CRE reporter activity even upon Fsk stimulation. We confirmed that exposure to Fsk increased SIK2 phosphorylation at PKA sites S358 and S587 ( Fig 2D) [11,34]. Consistent with its inhibitory effect on wild type SIK2, Fsk promoted dephosphorylation of CRTCs and class IIA HDACs. Although the S2-M2 hybrid protein was destabilized in comparison to SIK2 and had weaker membrane localization relative to MARK2 (Fig 2E), S2-M2 phosphorylated CRTC/ HDAC in a Fsk-resistant manner (Fig 2D). Taken together, these results indicate that SIK2 and MARK2 are distinguished by their relative sensitivity to cAMP, subcellular localization, and interaction with CLASP1/2 proteins. These differences may contribute to the increased cellular potency of MARK2 in phosphorylating variant zXKXGSXXNC motifs in tau and other substrates.

MARK1-3 interact with the microtubule-binding CLASP1 protein via their C-terminal KA1 domain
The MARK family consists of four members, whose activities are modulated by LKB1/ TAOK1-and aPKC-mediated phosphorylation (Fig 3A). MARKs also harbor a~100 amino acid conserved KA1 domain, which typically mediates membrane localization [14]. To investigate the relationship between CLASP1 interaction, subcellular localization, and zXKXGSXXNC motif phosphorylation, we performed overexpression studies with members of the MARK family (MARK1-4) (Fig 3B and 3C). In co-IPs, MARK1-3 associated with CLASP1 and MARK1 and MARK2 showed stronger interactions with CLASP1 than MARK3 (Fig 3B). Out of two non-overlapping truncation mutants of MARK2 (2-307 and 307-788), only the C-terminal domain (MARK2 307-788) bound CLASP1. All four MARKs mediated efficient phosphorylation of both zXKXGSXXNC motifs inside tau (sites S262, S293, and S324) as well as the AMPK motifs in the CRTCs (Fig 3C), with MARK2/4 appearing more potent than MARK1/3. Consistent with the importance of the UBA domain for LKB1-mediated MARK activation (T208 in MARK2) [5], MARK2 2-307 containing only the kinase domain was not phosphorylated within the activation loop (AL; P-MARK AL antibody) and consequently, was unable to phosphorylate cellular substrates (Fig 3B and 3C). In immunofluorescence studies of HEK 293T cells, we confirmed the association of MARK1-3 with cell membranes [14] and the interaction of MARK4 with microtubules and microtubule organizing centers (MTOCs) [51] (Fig 3D). MARK2 showed the strongest membrane binding, which was unaffected by the catalytically inactivating Thr to Ala activation loop mutant (MARK2 T208A S212A).
Although the interaction with the microtubule-associating CLASP1 proteins was limited to MARK1-3, all MARKs hyperphosphorylate zXKXGSXXNC motifs on cellular tau. MARK4 on the other hand binds by itself to microtubules and might therefore not require CLASP1 interaction for proximity-dependent zXKXGSXXNC phosphorylation. Sequence comparison of MARK1-4 revealed that the C-terminal KA1 domain is highly conserved among MARK1-3 and more variable or even absent in MARK4 isoforms [42] (Fig 4A). To further evaluate the effect of CLASP1 binding for zXKXGSXXNC motif phosphorylation, we generated consecutive C-terminal truncations of MARK2 (Fig 4B and 4C). Deletion of its extreme C-terminus (MARK2 2-745) severely impaired the binding of MARK2 to endogenous CLASP1 ( Fig 4B). As MARK2 2-688 and 2-575 mutants had no further inhibitory effect on CLASP1 binding relative to MARK2 2-745, we conclude that the distal 45 amino acids of the KA1 domain are required for complex formation. Despite the loss of CLASP1 binding, MARK2 mutants were still capable of phosphorylating zXKXGSXXNC and AMPK motifs comparably (Fig 4C). The MARK2-CLASP1 interaction was subsequently confirmed by coexpression of FLAG-tagged CLASP1 and HA-tagged MARK1-4 ( Fig 4D). Consistent with the observed interaction patterns for endogenous CLASP1, overexpressed CLASP1 bound most strongly to wild type MARK2 relative to MARK1/3 and the MARK2 2-745 mutant showed dramatically impaired complex formation. Since the KA1 domain has previously been implicated in membrane binding [14], we studied our C-terminal truncation mutants by immunofluorescence (Fig 4E). Although MARK2 2-745 still bound predominantly to the plasma membrane, the MARK2 2-575 mutant localized completely inside the cytoplasm. Thus, although MARK family members exhibit distinct subcellular localization and interactions with the microtubule-associating protein CLASP1, these associations have no detectable effect on catalytic activity or zXKXG SXXNC motif phosphorylation.

The kinase domain of MARK2 mediates efficient zXKXGSXXNC motif phosphorylation
Realizing that the C-terminal KA1 domain is dispensable for MARK2-mediated zXKXGS XXNC phosphorylation, we evaluated effects of other MARK2 regions on catalytic activity. MARK family members contain a UBA domain, which is necessary for LKB1-mediated activation inside cells [5] (Fig 5A). Consequently, the smallest active mutant (MARK2 2-370) contained the catalytic and UBA domains (Fig 5B and 5C). All three MARK2 truncations were phosphorylated at their activation loop and had comparable activity on both AMPK and zXKXGSXXNC substrate motifs. These results indicate that a protein fragment containing kinase and UBA domains of MARK2 (MARK2 2-370) is sufficient for cellular phosphorylation of zXKXGSXXNC motifs.
Sequence comparison of MARKs (MARK1-4) with the other AMPK family members revealed a 22 amino acid MARK-specific linker (MSL; MARK2 amino acids 308-330) ( Fig  6A), which is located between the N-terminal kinase and C-terminal UBA domains. Fig 6B) disrupted MARK catalytic activity as well as CLASP1 and 14-3-3 binding (Fig 6C and 6D). Because these interactions are mediated by sequences outside the MSL, we hypothesize that loss of this region must disrupt protein folding. Nevertheless, ΔMSL still bound predominantly to membranes ( Fig  6E), supporting the notion that the KA1 domain is partially functional.

Replacement of the MSL with a flexible GGSGGS linker (ΔMSL;
We generated MARK2 and SIK2 hybrid proteins to further define the role of the MSL for zXKXGSXXNC-directed MARK2 activity (Fig 6F). M2-S2 UBA carries the kinase domain of MARK2 with the linker sequence and UBA domain of SIK2, while S2-M2 UBA encodes the kinase domain of SIK2 with the MSL and UBA domain of MARK2 (Fig 6G). The kinase domain of MARK2 was not functional in the context of SIK2-derived sequences (M2-S2 UBA versus activation loop mutant); but the kinase domain of SIK2 was active following fusion with

The role of the KLDpT motif inside MARKs activation loop
Having seen that the kinase domain of MARKs has higher zXKXGSXXNC-directed activity relative to SIKs, we examined the sequences within the kinase domain of MARK1-4 and SIK family members. We noticed that a number of MARK2 isoforms lack the unstructured N-terminus [42] (Fig 7A). Around the activation loop, MARK1-4 also contain residues that are conserved within this subfamily, but that differ in SIK family members (Fig 7B). In turn, we examined effects of deleting the unstructured N-terminus or mutating residues inside the activation loop of MARK2 to corresponding residues in SIK2 (Fig 7C).
Deletion of the unstructured N-terminus (MARK2 48-788) had no effect on AMPK and zXKXGSXXNC motif phosphorylation. But an activation loop mutant of MARK2 (MARK2 K205L D207A) was more strongly impaired in its ability to phosphorylate tau at zXKXG SXXNC sites S262, S293, and S324 than at AMPK motifs inside CRTCs and class IIA HDACs. Although the activation loops of MARK2 and SIK2 differ at eight residues (Fig 7D), mutation at any of these residues did not further disrupt MARK2 catalytic activity relative to MARK2 K205L D207A (Fig 7E). However, replacement of the complete activation loop with the SIK2 sequence (MARK2 SIK2 AL) rescued cellular activity against zXKXGSXXNC and AMPK motifs. The impairment of MARK2 activity by K205L D207A was also evident within the stand-alone kinase domain fragment (MARK2 2-370), confirming that the motif-specific catalytic inhibition is independent of the regulatory C-terminus. In contrast to MARK2, the corresponding activation loop mutant inside SIK2 (L172K A174D) affected neither zXKXGSXX NC nor AMPK motif phosphorylation. As the KLDpT to LLApT activation loop mutants in MARK1/3/4 also impaired zXKXGSXXNC phosphorylation (Fig 7F), these results support the Dependency of MARK kinase activity upon the activation loop notion that the molecular mechanism underlying this process is MARK-specific and conserved.
We purified the full-length proteins of MARK2, SIK2, and their respective activation loop mutants from HEK 293T cells to determine whether the KLDpT208 motif is also required for in vitro activity of MARK2 (Fig 8). To obtain appropriate unphosphorylated zXKXGSXXNC and AMPK substrates, we purified tau (Tau-383/0N4R) and a CRTC2 fragment (147-297 of mCRTC2) following their heterologous expression in Escherichia coli. In the corresponding in vitro kinase assay, we quantified the parallel phosphorylation of S274 AMPK motif inside CRTC2 (P-CRTC3 S273) and S262 zXKXGSXXNC motif inside tau (P-Tau S262 T263). Consistent with the cellular data, the K205L D207A mutant within the MARK2 activation loop disrupted the in vitro phosphorylation of zXKXGSXXNC and AMPK motifs (Fig 8A-8C). In contrast, the L172K A174D mutant had no detectable effect on the in vitro activity of SIK2 ( Fig  8D-8F). Collectively, these results demonstrate that the KLDpT208 activation loop motif promotes catalytic activity of MARK2. Dependency of MARK kinase activity upon the activation loop Discussion AMPK family members have been shown to regulate both canonical AMPK and variant zXKXGSXXNC motifs [2,30]. We found that the MARK family has an intrinsically higher potency regarding the phosphorylation of zXKXGSXXNC motifs in comparison to SIKs. The ability for MARKs to recognize AMPK and variant motifs appears to be hardwired into their kinase domain, and more specifically, to the KLDpT activation loop motif. Consistent with its importance for MARK activity, the KLDpT motif is conserved among all vertebrate MARKs (Fig 9A) [42]. Indeed, KLDpT is also present in PAR-1 kinases, the invertebrate paralogs of MARKs (Fig 9B). Although the KLDpT motif is unique to the MARK activation loop (Fig 9C), other AMPK family members contain related motifs: LLEpT in BRSK1/2 and KLTpT in SNRK. Although introduction of KLDpT into SIK2 did not alter specificity, future studies should address the extent to which KLDpT-related sequences influence catalytic activity of BRSK1/2 and SNRK against conventional and variant zXKXGSXXNC motifs.
The consensus LXRXXSXXXL motifs inside CRTCs and class IIA HDACs display a greater degree of flexibility than the variant zXKXGSXXNC motifs in tau and MAP2/4. Indeed, these  canonical motifs can either naturally lack critical residues at positions -5, -3 or +4 or tolerate mutations at these positions, for example the Ala mutation at -5 Leu of L -5 NTGGpS273LPDL +4 in mCRTC3 had no effect on SIK-or MARK-mediated phosphorylation [10]. A crucial difference between consensus and variant motifs is their biological role, as the AMPK motifs in CRTCs and class IIA HDACs also function in parallel as cellular 14-3-3 binding sites [48], while zXKXGSXXNC phosphorylation alone is sufficient to displace tau and MAP2/4 from microtubules [24,52]. The 14-3-3 consensus site is described as RSXpSXP [53], and indeed the +2 Pro is found in six out of the seven AMPK motifs in CRTCs and class IIA HDACs. Consistent with their importance for 14-3-3 interaction, while the +2 Pro is critical in L -5 NTGGpS273LP +2 DL +4 , at the site missing the +2 Pro L -5 NR -3 TNpS162DSAL +4 the -3 Arg was required for phosphorylation and 14-3-3 binding of mCRTC3 [10]. Therefore, the AMPK-like motifs might allow for greater deviations from the consensus as long as they enable efficient 14-3-3 protein binding.
Previous studies documented the influence of activation loop residues on catalytic activity as well as substrate specificity [54], with point mutants having either beneficial or deleterious effects on activity [55,56]. The KLDpT motif (MARK2 K205L D207A) we identified could modulate activity either via ubiquitous effects on catalytic activity or via modulation of substrate specificity. The KLDpT motif incorporates the phosphorylated activation loop threonine (T208 in MARK2), which is required for cellular activity [3,50]. Accordingly, changes within the surrounding residues might affect the stoichiometry of threonine phosphorylation and correspondingly the catalysis of the phosphoryl transfer reaction. The phosphorylated KLDpT activation loop sequence may also influence MARKs substrate specificity by enhancing zXKXGSXXNC and LXRXXSXXXL recognition. The structure of substrate-bound MARK has not been elucidated, but MARK2 with unphosphorylated activation loop (T208) has been crystallized in complex with an inhibitory pseudosubstrate (CagA) that resembles a classical AMPK motif (Fig 9D) [43]. Notably, the KLDT208 motif makes no direct contact with CagA in this structure. The structure of another inactive MARK2 (K82R) carrying the activation loop mimic T208E, shows a different outward-facing activation loop conformation in an otherwise rigid kinase domain (Fig 9E) [44]. These results suggest that the MARK2 activation loop retains a high degree of conformational flexibility, which may allow it to participate in substrate recognition. In addition, computational modeling studies of PKA support the notion that catalytic activity and substrate specificity are not mutually exclusive; the phosphorylated Thr inside its activation loop modulates PKA activity in the phosphoryl transfer reaction as well as in the active site conformation [57].
In the course of our zXKXGSXXNC motif studies, we characterized the KA1 domainmediated interaction of MARK1-3 with CLASP1/2. Indeed, the interaction of MARK3 with CLASP1 and CLASP2 by co-IP has been reported [58] and the MARK2-CLASP2 complex has also been documented in IP-MS experiments [59]. In the former study, CLASP1/2 were found to serve as in vitro substrates of MARK3, which phosphorylated these proteins at a conserved and classical AMPK motif (hCLASP1/2: L -5 QR -3 SRpS600/370DIDV +4 ) [58]. Although CLASP1 interaction and KA1-dependent membrane localization were dispensable in cells expressing MARK2, these molecular properties may be critical in promoting the localized displacement of tau and MAPs from microtubules. Recent studies with a MARK1"mini" protein consisting of the kinase & UBA domains fused to KA1 revealed that the KA1 domain has autoinhibitory effects on MARK catalytic activity [60,61], under conditions when KA1-phospholipid interactions are absent [60]. These results suggest that cell membrane-bound MARK1-3 and microtubule-bound MARK4 are catalytically active, where they mediate the displacement of MAP2/4 and tau from microtubules via zXKXGSXXNC phosphorylation. Following activation of aPKC [15] and other kinases [16], MARKs undergo phosphorylation and 14-3-3 protein binding, which promotes their translocation to the cytoplasm and thereby limits MARK cellular activity. The mechanism by which the phosphorylated MARK-14-3-3 complex interferes with the KA1-mediated phospholipid interaction while at the same time allowing KA1 to bind to and inhibit the N-terminal kinase domain, is unclear. Future studies should reveal signals that modulate cell polarity by controlling MARK subcellular localization and activity.