MARK4 and MARK3 associate with early tau phosphorylation in Alzheimer’s disease granulovacuolar degeneration bodies

Tau protein was identified as the principal component of Alzheimer’s disease (AD) neurofibrillary tangles (NFTs) in the 1980s [1–3]. This finding led to the identification of many kinases that have the ability to phosphorylate tau and the description of more than 45 phosphorylation sites present on paired helical filament (PHF) tau [4–6]. Indeed, phosphorylation at specific tau epitopes is critical for neurofibrillary tangle formation and subsequent neurodegeneration [7, 8]. One of the earliest sites on tau to become phosphorylated in the AD brain is Ser²⁶²[9], which is located in the KXGS motif in one of the microtubule repeat domains, and that can be phosphorylated by MARK1-4 [10–13], a family of four highly conserved kinases [10, 12, 14].

It is known from in vitro studies that MARK phosphorylation of tau at the Ser²⁶² site causes detachment of tau from microtubules and their subsequent destabilization makes tau available for further phosphorylation by other kinases [10, 15], and MARK phosphorylation can induce mis-sorting of phosphorylated tau [16]. Tau Ser²⁶² phosphorylation and mislocalization are early events in a mouse model of tau pathology [17], and studies in Drosophila have demonstrated a crucial role of the MARK phosphorylation site on tau for neurodegeneration [18, 19].

A previous study that examined MARK expression in the human brain reported increased MARK1 expression in AD, but lacked a robust confirmation of the isoform-specificity of the antibody used [20]. We recently succeeded in developing and identifying specific antibodies towards each of the four MARK isoforms. Using these specific antibodies and a monoclonal antibody towards unphosphorylated tau we were able to demonstrate an increased interaction of MARK2 and MARK4 in AD hippocampal tissue compared to controls using the in situ proximity ligation assay [13, 21].

Granulovacuolar degeneration bodies (GVDs) are double membrane vacuoles present in neurons, having an immunohistochemical signature that suggests that they derive from the autophagic system [22]. GVDs also stain for cytoskeletal proteins such as neurofilament, tubulin tau and tau kinases [1, 23–28]. GVDs have been shown to be more frequent in AD brains compared to in age-matched controls [29], and a recent study suggests that GVD accumulation is specific to AD, since GVD frequency correlated with every measure of AD severity but was not different in any other non-AD tauopathies compared to control brains [30].

In the present study we characterized the intracellular localization of the four MARK isoforms and investigated whether their expression levels were elevated in the hippocampus in AD. We observed abundant neuronal mRNA expression of all MARK isoforms in both AD and NDE cases. At the protein level we determined that MARK1 and MARK2 were abundantly expressed in neuronal cytoplasm, but that expression levels did not increase in AD. In addition to a general cytoplasmic expression that did not change in AD, MARK3 was detected in a minor fraction of GVDs that are evident in neurons in AD. The expression of MARK4 was below the detection level in normal brain tissue, but was highly present in a phosphorylated form in GVDs in AD, where it colocalized with tau Ser²⁶² phosphorylation.

The mRNA levels of all MARKs were similar between AD and NDE and did not correlate with pathological changes. This indicates that transcription of MARK genes is unaffected in AD. Whereas the uniform neuronal expression of MARK1, 2, and 3 mRNAs was reflected by a cytoplasmic expression of the corresponding protein in neurons, there was no noticeable cytoplasmic MARK4 protein expression and hence MARK4 was entirely absent in healthy neurons. This absence was not due to lack of MARK4 affinity, as the antibody detected MARK4 in co-transfected HEK cells. In a previous study we observed low cytoplasmic MARK4 expression in NDE cases using the in situ proximity ligation assay [13]. This is likely due to the higher sensitivity achieved using the proximity ligation method, but could also reflect the recognition of neurons in which pathological MARK phosphorylation has begun. It is worth noting that both fresh-frozen and formalin-fixed paraffin embedded tissues were used, with identical staining pattern. The very low MARK4 protein expression in NDE cases was confirmed by Western blotting, which did not detect full-length MARK4. This apparent absence of MARK4 protein may be due to that the expression level is below the limit of detection using the present methodologies, but could also perhaps be due to some unknown post-translational modification at the epitope site that disturbs both immunohistochemical and biochemical detection in the post-mortem tissues.

In contrast to the very low expression in NDE, we observed a prominent expression of MARK4 protein in AD tissues. It is interesting that MARK4 is specifically elevated, particularly since a locus near the MARK4 gene has recently been described to have genome-wide significance for AD [34]. MARK3 and MARK4 granules coincided with the presence GVDs (as determined by CK1δ⁺ granules), which are defined as electron-dense granules within double membrane-bound cytoplasmic vacuoles, and that are believed to be part of the autophagic system [22, 32]. Interestingly, it was demonstrated in a recent report that the level of GVDs increases steadily with AD progression, correlating with Braak stage, CERAD score, amyloid beta phase and cognitive state, but does not correlate in other non-AD tauopathies [30]. Another recent study corroborates the view that neurofibrillary and GVD pathologies are tightly associated [35].

The distribution of MARK3 and MARK4 granules in CA-neurons correlated well with CK1δ granules and all three markers increased in a Braak-stage dependent manner, except for a small attenuation in CA1-CA2 at Braak stage VI. This attenuation is likely due to the severe neuron loss observed in this area at this Braak stage, since the data was expressed in neurons/mm² and not percent of total neurons. Interestingly, only 32% of intracellular granules were MARK4⁺CK1δ⁺ and as few as 10% were MARK3⁺CK1δ⁺. Furthermore, MARK4⁺ granules were significantly larger in size than CK1δ⁺ or MARK4⁺CK1δ⁺ granules. Taken together, this suggests that although MARK3 and MARK4, along with CK1δ are sequestered in a GVD-dependent manner, CA-neurons utilize different intracellular pathways to sequester these kinases. The relatively large size of the MARK4⁺ granules as compared to the CK1δ⁺ granules could possibly indicate that MARK4⁺ granules represent a more mature form of GVDs.

The exact role of these events in neuropathological progression in AD remains to be elucidated. It is well established that there is a spatiotemporal progression of neurofibrillary pathology in the brain, with certain areas being affected before others. The entorhinal cortex becomes affected at an early stage with subsequent appearance of NFTs in the CA1-field [36]. Conversion of individual neurons into NFTs is also suggested to occur in a hierarchical sequence of phosphorylation events, with Ser²⁶² phosphorylation appearing early in the formation of a NFT when the neuron is in a pretangle stage [9, 37]. Herein we demonstrate that the first appearance of Ser²⁶² phosphorylation in the CA-field is paralleled by an activation of the GVD system and particularly by strong MARK4 expression. In CA1 in the early Braak stages, tau Ser²⁶² is initially only evident in a very limited number of neurons, and although present in the cytoplasm is highly concentrated in the GVDs. According to the GVD-staging proposed by Thal et al. [30], GVD pathology first appears in CA1-neurons and only in later stages in the entorhinal cortex. The fact that GVD formation is first apparent in CA1-neurons could be due to a higher sensitivity of these neurons to tau phosphorylation, or alternatively a higher ability to engage autophagic GVD processes as a defence mechanism. In concordance with this latter hypothesis, induction of autophagy has recently been shown to be able to reduce levels of phosphorylated tau in neurons [38]. It can be hypothesized that this early phosphorylation of tau Ser²⁶² is driven by increased activities of cytoplasmic MARK2, MARK3 and very low levels of MARK4. This hypothesis is partly supported by the increased association of MARK2 and MARK4 to tau in the somatodendritic compartment, shown in a previous study [13]. Phosphorylation at tau Ser²⁶² will lead to subsequent phosphorylation at other tau epitopes which will cause detachment from microtubules and make tau available for aggregation into NFTs [10].

MARK4 in the GVDs was highly phosphorylated, indicative of the presence in a high activity form. MARKs contain numerous kinase-regulated sites [39, 40] and, for example, phosphorylation at the Thr²¹⁴ site (of MARK4 sequence) can increase MARK activity >50 times [33]. However, the exact mechanism involved in activation of MARK4 in AD, and the possible link to amyloid beta oligomers [41] remains to be established.

It is possible that MARK3 and 4 are phosphorylated and activated within the GVDs, but alternatively the activation may already have occurred in the neuronal cytoplasm. If so, there appears to be a very effective sequestering by the GVDs since we could not detect any phosphorylated and activated MARK4 or MARK3 in the cytosolic compartment. MARK4 in GVDs appears to undergo proteolysis since only shorter bands were observed in the immunoblots and these bands had increased in intensity in AD. That the 40–65 kDa bands represent proteolytic products within the GVDs is also consistent with the lack of similar bands in HEK cell lysates. However, it cannot be excluded that they to some extent represent non-specific cross-reactive bands.

The detection of bands representing truncated protein evident in the NDE cases probably reflects the rare presence of GVDs at these early Braak stages. A similar pattern was observed for MARK3, whereby an increase of truncated forms of MARK3 in immunoblots corresponds to an increased expression in AD by appearing in a minor fraction of GVDs. In contrast, full- length MARK3 was not elevated in AD and therefore presumably reflects the cytoplasmic component that is similar between NDE and AD cases.

It is possible that MARK4 and to some extent MARK3 become sequestered and that they phosphorylate tau already present in the GVDs at Ser²⁶², due to the closer proximity between substrate and enzyme [13], as well as due to the activated state of MARK3 and MARK4. Alternatively, Ser²⁶² may already be phosphorylated in the tau molecules sequestered into the GVDs, and sequestering of tau phosphorylated Ser²⁶² into GVDs would confer prevention of subsequent phosphorylation at other epitopes and tangle formation. In fact, subsequent phosphorylation at Ser²⁰²/Thr²⁰⁵ (AT8), Thr²¹²/Ser²¹⁴ (AT100), and Ser⁴²² are absent in the GVDs, in spite of the presence of many tau phosphorylating enzymes such as CK1δ [26, 28], GSK3β in its active form [27] and CDK5 [42]. It should be noted that our results on this matter conflict slightly with those of Leroy et al.[27] who found not only p-tau Ser²⁶² but also AT100 immunoreactivity (although to a lesser extent) in GVDs. An early study also identified GVD immunoreactivity using the Tau-1 antibody in dephosphorylated brain tissues, suggesting the presence of the p-Ser¹⁹⁹ epitope [1, 43]. In further support of this hypothesis, Ikegami et al. could not find PHF-structures upon ultrastructural examination of GVDs, indicating the presence of pre-PHF tau [44]. It should be noted though that total tau does not accumulate into GVD as a result of such possible sequestering. Moreover, it is interesting to note that CaMKII, which accounts for a large portion of normal brain p-Ser²⁶² activity [45] has not been found in GVDs in three immunohistochemical studies of AD hippocampus [46–48].

The cytoplasmic phosphorylation of tau steadily increases with time, likely driven by cytosolic MARK2. This notion is supported by an increased MARK2-tau interaction in this compartment [13, 21], as well as a complete lack of immunoreactivity in GVDs. Hence a continuous cytoplasmic presence of active MARK2 may allow a slow but steady continuous cytosolic phosphorylation at other p-tau sites, a process that GVD-mediated autophagy cannot prevent. The presumed protective effect of GVDs ultimately fails at later Braak stages, either due to a diminished capacity of GVDs to sequester p-tau Ser²⁶² or due to breakdown of GVDs with subsequent release of their components into the cytosol. Together these events lead to the cytoplasmic neuronal compartment filling up with tau phosphorylated at multiple sites available for further aggregation into NFTs, which in turn may result in neuronal degeneration and associated cognitive decline.

There is presently no evidence for an involvement of MARK1 in any of these pathological events. A previous investigation of MARK expression in human brain tissues reported increased levels of MARK1 in AD in tau tangle-containing neurons [20], in contrast with current and previous [13] findings, and this discrepancy could be due to less specific antibodies used in that study. Another report described an association between active MARK and neurofibrillary tangles but did not further describe the antibody characteristics [49].

In the present study we demonstrate that phosphorylated MARK4 and to some extent MARK3 are highly localized to GVD expressing neurons and that this expression is strongly correlated with phosphorylation of tau at Ser²⁶². This may represent a cellular defense mechanism to remove activated MARK and p-tau Ser²⁶² from the cytosol, thereby reducing the phosphorylating effect on tau Ser²⁶² that appears to be a critical step for subsequent neurodegeneration in AD.