Mobility and subcellular localization of endogenous, gene-edited Tau differs from that of over-expressed human wild-type and P301L mutant Tau

Alzheimer’s disease (AD) and a subset of frontotemporal dementia termed FTLD-Tau are characterized by a massive, yet incompletely characterized and understood redistribution of Tau. To establish a framework for understanding this pathology, we used the genome-editing tool TALEN and generated Tau-mEOS2 knock-in mice to determine the mobility and subcellular localization of endogenous Tau in hippocampal cultures. We analysed Tau in axons, dendrites and spines at three stages of maturation using live-cell imaging, photo-conversion and FRAP assays. Tau-mEOS2 cultures were compared with those over-expressing EGFP-tagged forms of human wild-type (hWT-Tau) and P301L mutant Tau (hP301L-Tau), modelling Tau accumulation in AD and FTLD-Tau, respectively. In developing neurons, Tau-mEOS2 followed a proximo-distal gradient in axons and a subcellular distribution similar to that of endogenous Tau in neurons obtained from wild-type mice, which were abolished, when either hWT-Tau or hP301L-Tau was over-expressed. For the three conditions, FRAP analysis revealed a similar mobility in dendrites compared with axons; however, Tau-mEOS2 was less mobile than hWT-Tau and hP301L-Tau and the mobile fraction was smaller, possibly reflecting less efficient microtubule binding of Tau when over-expressed. Together, our study presents Tau-mEOS2 mice as a novel tool for the study of Tau in a physiological and a pathological context.

different from human WT Tau-expressing mice where NFT formation is delayed 17 . These animal studies have been complemented by other investigations in cell lines which have shown, for example, that transient expression of WT or disease-linked mutations of Tau (R406W, P301L, Δ N296) had a dramatic effect on the microtubule cytoskeleton as observed by immunofluorescence microscopy 18 . Additional insight has been gained in the roundworm C. elegans where the Tau homologue Ptl-1 has been shown to regulate structural neuronal integrity 19,20 .
In vivo, Tau has been reported to be an axonal protein under physiological conditions, although a role has also been demonstrated in the dendritic compartment where levels are considerably lower 21,22 . In neuronal cultures, Tau has been shown to be ubiquitously expressed, with an enrichment being reported in the axonal compartment 23 . In tauopathies, Tau accumulates in the somatodendritic domain. Aβ incubation and synaptic activation have both been demonstrated to cause the accumulation of Tau in spines24. On the other hand, P301L Tau has been shown to accumulate in an Aβ -dependent manner in spines in vivo 21 . Similarly, over-expression of P301L Tau was also found to cause increased spine localization 25 . Together, these findings demonstrate that Tau localization to spines is tightly regulated.
In order to determine how Tau accumulation impairs neuronal function, it is important to understand how the protein is distributed in neurons, how it is localized, and the nature of its mobility characteristics. A few studies have assessed the transport of Tau under both physiological and pathological conditions, leading to the claim that, at least in axons, Tau is transported anterogradely through a combination of free diffusion and transient association with microtubules, hopping from microtubule to microtubule ('kiss and hop'), with a reported 40 ms dwell time 26 , and diffusing along the microtubule lattice 27 . Some studies assessing Tau transport have also applied newer microscopy methods such as FRAP (Fluorescence Recovery After Photobleaching) or sptPALM (single particle tracking Photo-Activated Localization Microscopy) 28 to over-expression systems 29 . Here, we generated a novel mouse strain using the genome editing tool TALEN to equip endogenous Tau (with its multiple isoforms) with an mEOS2 tag that could be used to track Tau and conduct photo-conversion. We then assessed the behaviour of the tagged Tau and compared it to that of over-expressed human WT (hWT-Tau) and P301L mutant Tau (hP301L-Tau).

Results
Tau-mEOS2 knock-in mice generated with the genome-editing tool TALEN express all three major central nervous system Tau isoforms. The protein Tau is encoded by the MAPT gene, and alternative splicing generates multiple isoforms 3 . To visualize all forms of endogenous Tau in vivo, we established a novel transgenic mouse line by introducing a photo-convertible mEOS2 tag in-frame into the carboxy-terminus of the MAPT gene, using the genome-editing tool TALEN 30 (Fig. 1a). Three TALEN pairs predicted to yield a high targeting efficiency and to have low off-target effects were designed and assembled using the TALEN toolkit provided by the Zhang Laboratory (Addgene Kit # 1000000019). A donor construct comprising the mEOS2 cassette flanked by 800 bp long arms both upstream and downstream of the genomic sequence containing the MAPT stop codon was designed (Fig. 1b). The TALEN reaction was established in N2A cells by evaluating three TALEN pairs for genome-editing activity using the T7 endonuclease I assay (data not shown). The pair with the highest activity was used for in vitro transcription, and the thereby generated TALEN mRNAs were combined with the donor construct and injected into pronuclei of C57Bl/6 single cell mouse embryos by classical transgenesis 31 , followed by transfer into pseudo-pregnant foster mice. To identify animals among the offspring (F0) that had undergone an mEOS2 targeting event, flanking PCR genotyping was performed, using two primer pairs (P1/P2 and P3/ P4), with one primer binding to the mEOS2 sequence (P3, P4), and the other flanking primer binding to regions outside the donor construct sequence, one upstream of the left arm (P1) and the other downstream of the right arm (P4) (Fig. 1c). Of a total of 120 offspring, two mice (#90 and #120) were identified as positive. To ensure that the recombination had not induced INDEL (insertion or deletion) mutations, we PCR-amplified the top 5 predicted off-target sites (using the above software) and conducted Sanger sequencing; this confirmed the absence of mutations (data not shown).
We next established two homozygous Tau-mEOS2 mouse lines (TauEOS90 and TauEOS120) employing a sequential breeding procedure. We performed an initial analysis that included western blotting and the establishment of primary neuronal cultures with both strains, and found no overt differences between them. Subsequently, we only used the TauEOS90 strain, which is referred to as Tau-mEOS2 from here on. Western blot analysis using a pan-Tau antibody (DAKO Tau) and including WT and Tau knock-out samples 32 as controls confirmed that the size of Tau-mEOS2 was about 75 kDa. As predicted, this is approximately 25 kDa larger than endogenous Tau, indicating correct targeting of the mEOS2 cassette (Fig. 1d). Tau is an unusual protein in that its longest human brain isoform contains 80 serine and threonine residues and 5 tyrosine residues, many of which are phosphorylated under physiological conditions. Owing to its basic amino acid content, its heterogeneous phosphorylation and the existence of multiple isoforms, the protein runs like a smear, and the individual isoforms can only be resolved by treating the extract with protein phosphatase 3,33 . By incubation with lambda phosphatase, the three murine central nervous system isoforms were revealed with a 0N:1N:2N ratio similar to that in WT mice, indicating that alternative splicing is maintained in the Tau-mEOS2 mice; the ratio in WT mice was 81:8.5:10.5, and the ratio in the Tau-mEOS2 mice was 70.5:8:21.5 (Fig. 1e). By titrating Tau-mEOS2 against WT extracts, we also found that the level of Tau expression was approximately fifteen-fold less than that of WT (Fig. 1d). To determine whether this was a reflection of reduced transcript levels, we used qRT-PCR and found an almost three-fold reduction in the levels of Tau-mEOS2 compared to WT Tau, suggesting that the tag interferes with either the transcription or turnover of Tau-mEOS2 mRNA (Fig. 1f). The Tau-mEOS2 mice, upon gross analysis, did not display any impairments. Together, these findings demonstrate that the Tau-mEOS2 strain is overtly normal, despite significantly reduced Tau levels, presenting it as a tool for the analysis of Tau mobility and subcellular localization in response to both internal and external stimuli. Over-expression of WT and P301L mutant Tau reverses the axo-dendritic gradient and abolishes the proximo-distal gradient in the axons of Tau-mEOS2 hippocampal cultures. Tau is conventionally treated as an axonal protein; however, we and others have previously demonstrated that, even under Flanking PCR using two primer pairs (P1/P2, P3/P4) to identify mice that have integrated the donor construct by homologous recombination (HR) (Tau-mEOS2 mice) identifies two positive founders (#90 and 120) in 120 animals obtained by microinjection. (d) Western blot analysis shows that the size of Tau-mEOS2 is about 75 kDa, which is approximately 25 kDa larger than endogenous Tau as predicted, indicating correct targeting of mEOS2 into the MAPT locus. Wild-type (WT) and Tau knock-out (KO) samples were included as controls. (e) Dephosphorylation reveals three murine central nervous system isoforms in Tau-mEOS2 mice with a 2N:1N:0N ratio similar to that in WT mice. (f) qRT-PCR analysis reveals almost three-fold reduced levels of Tau-mEOS2 compared to WT Tau (twotailed unpaired Students t-test, * * P < 0.01, n = 3 per group). physiological conditions, Tau is also present in the dendrites and in spines, albeit at lower levels 21 . Tau has been further localized to neuronal nuclei 34 ; it has also been found in nuclear fractions, in isolated intact nuclei and as speckles in the nuclei of SH-SY5Y but not N2a neuroblastoma cells 35 . These analyses used Tau-specific antibodies and, as far as conventional immunohistochemistry is concerned, required antigen retrieval methods, unless Tau was over-expressed as is the case for Tau transgenic mice that model the Tau pathology in AD and FTD 36 . What these data illustrate is the complexity in revealing the distribution of Tau, some of which may reflect nonspecific binding of the different anti-Tau antibodies that have been employed.
To clarify whether Tau is expressed at significant levels in dendrites and nuclei, we established primary hippocampal cultures from homozygous Tau-mEOS2 mice and used confocal imaging to visualize the mEOS2-tagged Tau in live neurons. We found that Tau-mEOS2 showed a strong axonal expression in mature (DIV18) neurons, with a gradient towards enrichment in the growth cone of distal axons (Fig. 2a). Tau-mEOS2 was further present in the somatodendritic domain (Fig. 2b), with very low levels in dendritic spines (Fig. 2c). Within the limit of detection, we did not visualize EOS-tagged endogenous Tau in the nucleus, suggesting that the level is either very low and/or that previously reported detection represented background reactivity (Fig. 2d). Interestingly, as for the staining of WT neurons with anti-Tau antibodies, Tau was found to be expressed throughout the cell, with an enrichment in the axon. This shows that, even with a reduced level of Tau, the neuron is capable of maintaining an axo-dendritic gradient. This is important for treatment strategies aimed at reducing Tau levels 37 .
To demonstrate that the endogenously expressed mEOS2-tagged Tau in the Tau-mEOS2 mice can be photo-converted, we assessed Tau transport in the axon. Time-lapse imaging revealed the bidirectional transport of Tau at about ~0.25 μ m/s out of the region of interest (Fig. 2e). This value is lower than that previously obtained for over-expressed CFP-hTau40 in retinal ganglion cell neurons (0.4~0.6 μ m/s) indicating that the transport of Tau may be affected by expression levels and the experimental system 38 .
We were next interested in determining whether the axo-dendritic gradient is impaired when Tau is over-expressed, as is routinely done to study Tau. For this, we used hippocampal cultures at two stages of maturation: DIV5, when neurons are immature, and DIV18, when the neurons have developed mature spines. We over-expressed EGFP-tagged forms of human WT (hWT-Tau) and human P301L mutant Tau (hP301L-Tau), one modelling the form of Tau that accumulates in AD and the other the form of Tau that accumulates in FTD. Importantly, as for the Tau-mEOS2 mice, the EGFP tag was added to the carboxy-terminus of the two over-expression constructs. However, the fact that an exogenous promoter rather than the endogenous promoter was used to drive expression of hWT and hP301L Tau could potentially result in a different spatial specificity, as it has been shown in transfection studies that the 3′ UTR of the Tau-encoding MAPT gene has a role in axonal targeting of the mRNA 39 . By DIV5, Tau was already found to be present in all major neuronal compartments (soma, axon and dendrite) in Tau-mEOS2 neurons (Fig. 3a,b). Moreover, it was axonally enriched, displaying a proximo-distal gradient, similar to that found in hippocampal cultures from WT C57Bl/6 mice that were stained with the anti-Tau antibody Tau5 and a fluorescent secondary antibody (Fig. 3c). By establishing line profiles through an axon and a dendrite running in parallel, the intensity of Tau in the axon was found to be approximately 2.5-fold higher than that in the dendrite both when Tau-mEOS2 cultures were captured and when WT cultures were stained with the Tau5 antibody (Fig. 3b' ,c'). More generally, dendritic staining was found to be uniformly weak, whereas that in axons was generally more variable. Upon over-expression of either hTau-EGFP or hP301LTau-EGFP in WT neurons, both forms of Tau were found to accumulate throughout the neuronal cytoplasm, although the fluorescence intensity for the axon was much lower than that for the somatodendritic domain (Fig. 3d,e). This was already evident at DIV5 and was maintained with neuronal maturation (Fig. 3a-e, lower panel). We also established line profiles for all experimental conditions to reveal the fluorescent Tau signal starting at a dendrite and proceeding through the soma and along the axon to its tip. This not only demonstrated that the axo-dendritic gradient was inverted in the over-expression systems, but also revealed that the proximodistal gradient of Tau was abolished within the axon (Fig. 3f). Together these results demonstrate that, despite a 15-fold lower expression level, Tau-mEOS2 shows a distribution similar to that of endogenous Tau, which is distorted when Tau is over-expressed in its WT form or carrying a mutation (P301L) found in human tauopathy.
Spine localization of over-expressed P301L Tau is increased compared with that of over-expressed WT Tau or endogenous mEOS2-tagged Tau. Tau localization in the dendrite has previously been found to extend to spines; however, levels of endogenous Tau were reported to be low, unless Tau was over-expressed, causing it to enter spines in a phosphorylation-dependent manner 25,40 . To assess spine localization of Tau, we co-expressed EGFP-tagged Lifeact to visualize F-actin 41 , as a spine marker, due to its co-localization with RFP-tagged PSD95 in DIV18 WT neurons (Fig. 4a).
Using live-cell imaging to assess hWT-Tau and hP301L-Tau over-expressing neurons at three stages of neuronal maturation (DIV5, 12 and 18), we found that although both forms of Tau were mainly excluded from filopodia within the limit of detection, they were targeted to dendritic spines, with a lower intensity of spinal compared to dendritic Tau, and a more pronounced spine localization of P301L Tau compared to WT human Tau (Fig. 4b-d), as shown by applying a higher gain to expose spine localization (Fig. 4e). Endogenous Tau was also targeted to spines, albeit at much lower levels (Fig. 4f). Unambiguously identifying spine localization is easier in hWT-Tau and hP301L-Tau over-expressing neurons than in Tau-mEOS2 cultures, as in the latter only a fraction of neurons is fluorescently tagged, allowing a better discrimination against a background of non-fluorescing neurons. We next performed spine counts and found that these did not statistically differ between hTau-mEOS2 and hWT-Tau over-expressing neurons, but were significantly reduced in hP301L-Tau over-expressing neurons (Fig. 4g). Hoover and colleagues previously reported no reductions in their P301L-Tau cultures, and whereas we observed approximately 95 spines/100 μ m in WT cultures, they reported around 40 spines/100 μ m. However, the two studies differ in several aspects: they transfected 0N4R and not 2N4R Tau, and they transfected at DIV7-10   and analysed the neurons at DIV21 25,40 , whereas we transfected the neurons only at DIV18 and analysed them at DIV19/20. Together, our findings reveal that even under physiological conditions, Tau is targeted to dendritic spines, and that over-expression of P301L mutant Tau causes a decrease in spine density.

FRAP analysis reveals a similar mobility of Tau in dendrites and axons, but in both compartments,
Tau-mEOS2 is less mobile than hWT-Tau and hP301L-Tau. We next used the FRAP assay to address the dynamics of Tau in DIV18-21 hippocampal cultures under over-expression conditions compared to the lower expression encountered in the Tau-mEOS2 mice, and analysed the recovery profiles to determine the mobility and mobile fraction in the different neuronal compartments. It has to be considered that trafficking of molecules into the spine is unidirectional (from the dendritic shaft through the spine neck into the spine head), whereas in the axon and dendrite, Tau can move into the bleached area from both ends. The bleached spine is also much smaller than the bleached area of either the axon or dendrite, which generally results in a faster recovery. Therefore, the spine FRAP data were anaysed separately from those obtained for axons and dendrites.
We first determined the dynamics of hWT-Tau, hP301L-Tau and Tau-mEOS2 in the axon and dendrite (Fig. 5a,b), and determined the mobility by calculating the diffusion constant t 1/2 (half time, the time required for the fluorescence to reach 50% recovery, with more mobile molecules displaying a smaller t 1/2 ) and the mobile fraction (Fig. 5c). We found in both the dendrite and the axon that the mobility of Tau-mEOS2 was lower than that of either hWT-Tau or hP301L-Tau. We further found when either hWT or hP301L Tau was over-expressed, that the mobile fraction of Tau was between 70 and 80% of the total pool, whereas for Tau-mEOS2, this fraction  Tau-mEOS2 compared to over-expressed hWT-and hP301L-Tau in the dendrite and axon. In both compartments, the t 1/2 of hTau-mEOS2 is higher than for either hWT-Tau or hP301L-Tau, and the mobile fraction is much smaller. (One-way ANOVA, * * * P < 0.005; N > 15).
was reduced by a factor of two, possibly reflecting less efficient binding of Tau to microtubules when Tau is over-expressed. Interestingly, for any experimental condition, we did not observe a difference in the mobility or mobile fraction of Tau between axons and dendrites.
We next determined the dynamics of hWT-Tau and hP301L-Tau in the spine (the Tau-mEOS2 signal was too weak to be included in the analysis) (Fig. 6a,b), and determined the diffusion constant t 1/2 and the mobile fraction in this subcellular compartment (Fig. 6c). Because F-actin is a critical Tau-interacting protein and is enriched in spines, we specifically determined the mobility of F-actin (Lifeact-RFP) depending on the form of Tau being expressed. This revealed that although F-actin was less mobile than tau in spines, its dynamics was not affected by the form of Tau being over-expressed (which was also the case for dendrites and axons). Interestingly, there was a smaller mobile fraction of hP301L-Tau in the spines (55%) compared to that of hWT-Tau (80%). One explanation may be that more hP301L-Tau is tethered with F-actin in spines compared to hWT-Tau, which reduces its turnover rate in spines. We also found that whereas the mobility of hP301L-Tau in the axon and dendrite was higher than that of hWT-Tau (see above), it did not differ significantly in the spine.

Conclusions and outlook.
In our study, we present the Tau-mEOS2 mouse strain as an interesting system in which to investigate transport in neurons, under conditions of both physiologically and pathologically phosphorylated Tau. Regarding the significantly reduced Tau levels in Tau-mEOS2 compared with WT mice it remains to be seen whether these can be increased by generating mice in which the mEOS2 tag is added to the amino-terminus of Tau. In the Tau-mEOS2 mice described above, we observed a Tau gradient that differed from that of over-expression systems. A possible explanation is provided by a recent study that used photo-conversion of human WT Tau tagged with photo-convertible Dendra2 followed by over-expression in relatively young cortical cultures to reveal an anterograde flow across the axon initial segment into the axon, but no flow back into the somato-dendritic compartment, explaining axonal enrichment of Tau 29 . When Tau was hyperphosphorylated using the phosphatase inhibitor okadaic acid, the retrograde barrier was found to break down functionally, with Tau now flowing into the somato-dendritic domain where it is known to accumulate in AD 29 . Because okadaic acid can have multiple effects that go beyond Tau hyperphosphorylation, the study further used a '4KXGE' Tau mutant mimicking constitutive phosphorylation, which again showed a functional break-down of the retrograde barrier and a relocalization of this form of Tau into the somatodendritic domain. This could potentially explain why over-expressed Tau accumulates in the somatodendritic domain, whereas endogenous Tau is more localized to the axon. However, there are also alternative explanations, as the over-expression of Tau could lead to a saturation effect, increasing the pool of Tau that is not captured by microtubules and hence starts accumulating in the somatodendritic domain.
In conclusion, with the Tau-mEOS2 mice we have developed a versatile tool that will allow for the monitoring of endogenous Tau in response to a range of stimuli. For example, neuronal activity has been found to drive physiological levels of Tau into the spine in a process in an incompletely understood process that is impaired by by the peptide Aβ that drives AD 24 . Here, the Tau-mEOS2 mice may be a tool to understand this process in more detail, using a system that avoids over-expression.

Materials and Methods
Ethics statement. All experimental procedures in this study were conducted under the guidelines of the Australian Code of Practice for the Care and Use of Animals for Experimental Purposes and were approved by the University of Queensland Animal Ethics Committee (QBI/412/14/NHMRC; QBI/027/12/NHMRC). Mice were maintained on a 12-hour light/dark cycle and housed in a PC2 facility with ad libitum access to food and water.
Generation of Tau-mEOS2 mice using TALEN gene editing. Mice with an EOS cassette inserted into the carboxy-terminus of the open reading frame of the tau-encoding MAPT gene (Tau-mEOS2 mice) were generated using the TALEN gene editing tool 30 . Three TALEN pairs that target a sequence near the stop codon of exon 12 of MAPT were designed using the online tool "TAL Effector Nucleotide Targeter 2.0" (https://tale-nt. cac.cornell.edu/node/add/talen). Three TALEN pairs that were predicted to yield a high targeting efficiency and low off-target effects were selected and assembled using the TALEN toolkit (Addgene Kit # 1000000019) and the PCR-golden gate ligation method 42 . A donor construct comprising the mEOS2 cassette flanked by 800 bp long arms both upstream and downstream of the genomic sequence containing the MAPT stop codon was PCR-amplified and subcloned into the pGEM-T easy vector (Promega) using the NEB Gibson assembly kit. TALEN activity was evaluated in N2A cells using the T7E1 assay. The pair that gave the highest activity was then used for in vitro transcription, using the Ambion megascript T7 kit. TALEN mRNAs (10 ng/μ l each) were combined with the donor construct (15ng/μ l) and injected into pronuclei of C57BL/6 single cell mouse embryos followed by transfer into pseudo-pregnant foster mice.
The following primer pairs were used to identify offspring (F0) animals with a correct in-frame mEOS2 targeting event.
Live cell imaging, FRAP assay and photo-conversion. Primary neurons were transfected using Lipofectamine 2000 and used for live-cell imaging 24 to 48 h post-transfection. The following constructs were used for transfection: hTau-EGFP (human full-length 2N4R Tau with carboxy-terminally tagged EGFP), hP301L-EGFP (FTD P301L mutant 2N4R Tau with carboxy-terminally tagged EGFP), Lifeact-RFP (pmTagRFP-T-Lifeact-7, Addgene Plasmid #54586), and Lifeact-EGFP (kindly provided by Dr. Fred Meunier). Photo-conversion was achieved using the 405 nm UV laser on the LSM710 confocal microscope (Zeiss). For the FRAP assay, DIV 18-21-transfected neurons were analysed using a Zeiss LSM 710 confocal microscope equipped with a 5% CO 2 , 37 °C humidified cage. Time-lapse images of the GFP and RFP channels were captured with a Zeiss Plan-Apochromat 63x Oil DIC (N.A. = 1.4) objective at 2 s intervals using a 4× optical zoom. Regions of interest (ROIs) of the same size or length in the spine, axon or dendrite were photo-bleached by scanning with a 405 nm and 488 nm laser, respectively, combined at the maximum power, at a scanning speed of 3.5 μ s/pixel which was repeated five times. Images were analysed using Image J and Prism 6. For each ROI, the total intensity was measured and background intensities were subtracted. A dendritic or axonal region away from the FRAP ROIs was used to compensate for the fluorescence loss during image acquisition. Data from different cells were normalized to pre-bleaching as 100% and immediately post-bleaching as 0%. The following formula was used for the fitting of FRAP curves: % Recovery = (1 − e −Kt ) • 100%, with K being the recovery rate constant, and (t1/2) the half-time = ln(2)/k (half-time being the time required for fluorescence to reach 50% recovery). Data are presented throughout as mean ± SEM.