Axonal TAU sorting in SH-SY5Y-derived neurons requires the C-terminus of TAU but is independent of Ankyrin G — Source link

The Slow Axonal Transport of the Microtubule-Associated Protein Tau and the Transport Rates of Different Isoforms and Mutants in Neurons Herpes simplex virus-mediated expression of the axonal protein tau in human model neurons (NT2-N cells). Sorting mechanisms of tau and MAP2 in neurons: suppressed axonal transit of MAP2 and locally regulated microtubule binding. Subcellular localization of tau mRNA in differentiating neuronal cell culture: Implications for neuronal polarity High molecular weight tau proteins and acquisition of neuronal polarity in peripheral nervous system Abstract Somatodendritic missorting of the axonal protein TAU is a hallmark of Alzheimer’s disease and related tauopathies. Cultured rodent primary neurons and iPSC-derived neurons are used for studying mechanisms of neuronal polarity, including TAU trafficking. However, these models are expensive, time-consuming and/or require the sacrification of animals. In this study, we evaluated four differentiation procedures to generate mature neuron cultures from human SH-SY5Y neuroblastoma cells, in comparison to mouse primary neurons, and tested their TAU sorting capacity. We show that SH-SY5Y-derived neurons, differentiated with sequential RA/BDNF treatment, are suitable for investigating axonal TAU sorting. These human neurons show pronounced neuronal polarity, axodendritic outgrowth, expression of the neuronal maturation markers TAU and MAP2, and, importantly, efficient axonal sorting of endogenous and transfected human wild type TAU, similar to primary neurons. We demonstrate that axonal TAU enrichment requires the presence of the C-terminal half, as a C-terminus-lacking construct (N-term-TAU HA ) is not axonally enriched in both neuronal cell models. Moreover, SH-SY5Y-derived neurons do not show formation of a classical axon initial segment (AIS), indicated by the lack of Ankyrin G (ANKG) and tripartite motif-containing protein 46 (TRIM46) at the proximal axon, which suggests that successful axonal TAU sorting is independent of classical AIS formation. Taken together, our results suggest i) that SH-SY5Y-derived neurons are a valuable human neuronal cell model for studying TAU sorting, which is readily accessible at low cost and without animal need, and that ii) the mechanisms of axonal TAU targeting require the TAU C-terminal half but are independent of ANKG or TRIM46 enrichment at the proximal axon.

Understanding the mechanisms of axonal TAU sorting in healthy neurons is a critical step towards the elucidation of pathological processes that accompany TAU missorting. Several mechanisms are thought to drive axonal TAU sorting 33,34 , including active anterograde TAU trafficking, selective axonal TAU-MT interactions, or active retrograde retention of axonal TAU. In this context, the axon initial segment (AIS), a key player for developing and maintaining neuronal polarity with Ankyrin G (ANKG) as a master organizer of the molecular structure [35][36][37] , may be relevant, as polarized MT filaments at the AIS are involved in motor protein-mediated soma-to-axontrafficking 35,38 , presumably including TAU transit 33,34 . The tripartite motif containingprotein 46 (TRIM46) may also be involved into anterograde TAU sorting by regulating the MT architecture at the AIS 39 . Moreover, recent studies found direct evidence for a retrograde TAU diffusion barrier within the AIS that is compromised by MT disassembly 40,41 . Notably, little is yet known about the intrinsic components of TAU, i.e. functional domains, interaction motifs, posttranslational modifications and relevant TAU interactions, although some recent studies in rodent primary neurons assumed major roles for phosphorylation sites within the Proline-rich region and for the very Nterminal domain of TAU, but only a minor role for the four MT-binding repeat domains 42,43 .
Which neuronal cell models are available to address these key aspects of TAU sorting and missorting? Two commonly used cellular systems are rodent primary neuron cultures [40][41][42][43] and recently also human induced pluripotent stem cell (iPSC)-derived neurons [44][45][46][47][48] . Both systems have considerable limitations. Rodent primary neuronal cultures require the use and sacrification of animals, and rodent TAU differs in several aspects from human TAU, e.g. regarding the isoform expression pattern 49 . In addition, the human cellular environment, i.e. the entire sorting machinery and interaction partners may differ from that in humans. Overexpressed human TAU often appears uniformly distributed in rodent primary neurons 50 . IPSC-derived neurons overcome many of these issues, but their generation is expensive, time-consuming and results in mixed cultures with variable homogeneity and differentiation efficiency [51][52][53] . The neuroblastoma cell line SH-SY5Y is human-derived 54 and cells can be differentiated with various substances into polarized neuronal cells (SH-SY5Y-derived neurons) at low cost 55 . SH-SY5Y-derived neurons express common neuronal maturation marker proteins [56][57][58][59][60][61] , including all six major human TAU isoforms 62,63 , show proper axonal TAU localization, and mimic the phosphorylation state of TAU in the adult human brain 64- 67 . In brief, SH-SY5Y-derived neurons unite the benefits of a cell linei.e. robust maintenance, fast growth rates, the saving of animal resources and simple accessibility for stable genetic engineering 68,69 with the higher scientific relevance of primary and iPSC derived neurons. In this study, we evaluated four differentiation procedures of SH-SY5Y cells for generating neuronal cells and tested their suitability for the investigation of TAU sorting. Mouse primary forebrain neurons were used as a reference neuronal cell model. We show that RA/BDNF-treated SH-SY5Y-derived neurons achieve pronounced cellular polarity, axonal outgrowth, strong upregulation of the neuronal markers TAU and MT-associated protein 2 (MAP2), and, importantly, efficient axonal sorting of endogenous TAU and overexpressed human TAU (0N3R-TAU HA ). In contrast, TAU without the C-terminal half (N-term-TAU HA ) is not axonally enriched in SH-SY5Y-derived and mouse primary neurons. Of note, SH-SY5Y-derived neurons do not show enrichment of ANKG and TRIM46 at the proximal axon despite successful axonal TAU targeting, suggesting an ANKG-and TRIM46-independent mechanism of axonal TAU enrichment. In brief, our results give evidence that i) SH-SY5Y-derived neurons show successful axonal sorting of endogenous and transfected human TAU, ii) the N-terminus of TAU is not sufficient for proper axonal sorting, and ii) the axonal targeting of endogenous and 0N3R-TAU HA is independent of ANKG, a master organizer of the AIS molecular structure, and TRIM46, a AIS-located MTorganizing protein.
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in

Molecular biology
For experiments with endogenous and transfected TAU, an expression vector, which harbored the cDNA encoding for the tdTomato fluorescent protein, was used as volume marker control. For experiments with recombinant TAU constructs, N-terminally tagged 0N3R-TAU HA and N-term-TAU HA constructs were cloned into the same, CMV promoter-driven expression vector (see Suppl. Sequence 1). For this, the intended TAU cDNA sequences were amplified by PCR using primer pairs with restriction sites (AgeI, BamHI) corresponding to the multiple cloning site of the backbone vector. Additionally, the forward primer contained a human influenza hemagglutinin (HA) tag sequence overhang with start codon (5'-ATGTACCCATACGATGTTCCAGATTACGCT-3'), which was fused to the 5' site of the TAU sequence. The PCR product was analyzed via gel electrophoresis, and correctly sized inserts were cut out and purified with NucleoSpin™ Gel & PCR Clean up Kit (Macherey & Nagel). Vector and insert were digested (30 min, 37 °C) according to manufacturer's protocols (enzymes by New England Biolabs), either purified directly (insert) or from agarose gel (vector) and ligated (16-18 h, 20-22 °C) with T4 DNA ligase (NEB) according to the manufacturer's protocol.
For plasmid amplification, plasmids were transformed into chemocompetent bacteria (E. coli Top10™, Thermo Fisher Scientific, TFS), which were generated like previously described 70 . Bacteria were spread onto pre-warmed agar plates with antibiotics (16-18 h, 37 °C), colonies were picked and grown in nutrient medium with antibiotics. For plasmid isolation and purification from bacterial suspensions, the PureYield™ MidiPrep Kit (ProMega) and NucleoBond Xtra™ Midi/Maxi Kit (Macherey & Nagel) were used according to the manufacturer's protocols. Plasmid concentration and purity were checked spectrophotometrically (NanoDrop™ 100, TFS) and isolates were long-term stored at -20 °C.
Differentiation. Differentiation protocols were started 24 h after seeding. For RA-based differentiation, fresh SMM with 10 µM retinoic acid (RA, Sigma-Aldrich) was added at day 0 (d0) and replaced every 2-3 days up to d7. Differentiation with RA & the brain-derived neurotrophic factor (BDNF, Peprotech) was adapted from former protocols 56 . Briefly, after RA treatment up to d7 as described above, cells were washed once with DPBS, and SHM-10 medium without FBS (referred to as 'SHM-0') containing 10 ng/ml BDNF was added and once replaced until d14. Differentiation protocols for Phorbol-12-myristate-13acetate (TPA)-and RA/TPA-based differentiation were adapted from former protocols 60,71 . In brief, fresh SHM-10 with 81 µM TPA (Sigma-Aldrich) was added and replaced every 2-3 days until d9 for TPAbased differentiation. For RA & TPA-treatment, fresh SHM-10 with 10 µM RA was added, replaced after 2-3 days up to d4 or d5, whereupon cells were washed once with PBS and cultivated in fresh SHM-10 with 81 µM TPA until d9 with one medium change in between.
Two alternative long-term differentiation procedures were tested for RA, RA/TPA and TPA treatments. Transfection SH-SY5Y-derived neurons were transfected using the polymer-based PolyJet™ DNA transfection reagent (#SL100688, SignaGen) at d5 of differentiation. Transfection was performed according to the manufacturer's protocol with the following modifications: Before transfection, conditioned medium was collected from each well and stored at 37 °C with 5 % CO2. For one well of a 24-well plate, 0.33 µg of plasmid DNA and 1 µl PolyJet™ reagent were separately prepared with 25 µl DMEM (#P04-03500, TFS), then mixed, incubated and added dropwise to the cultures. Medium was changed to previously collected, conditioned medium, supplemented with fresh doses of RA or TPA, 3-16 h after transfection. Transfection efficiency ranged from 5-10 %. For experiments with recombinant TAU, tdTomato-and TAU HA construct-encoding plasmids were co-transfected in SH-SY5Y-derived neurons. For this, DNA of both plasmids was incubated commonly with PolyJet™. The plasmid DNA was mixed with a ratio of three (TAU HA plasmid) to one part (tdTomato plasmid). The total DNA amount per well was not altered, i.e. 0.33 µg was used for one well of a 24-well plate.
Maintenance. Four days after seeding, the medium was doubled by adding neuronal maintenance medium (NMM), i.e. NPM without FBS. Cytosine arabinoside (Ara-C, #C1768, Sigma-Aldrich) was added to an end concentration of 0.5 -1.0 µg/ml to impair survival of glial, endothelial and other proliferating cells.
Transfection. Mouse primary forebrain neurons were transfected at day 6 in vitro (div6). Transfection was performed as described previously 73 . In brief, 0.25 µg of plasmid DNA and 0.375 µl Lipofectamine™ 2000 (#743517, Invitrogen) were mixed for one well of a 24-well plate, incubated for 30 minutes and added dropwise to the cultures. Medium was changed to previously collected, conditioned medium 60 minutes after transfection. For experiments with recombinant TAU, tdTomato-and TAU HA constructencoding plasmids were co-transfected in mouse primary neurons. For this, the DNA of both plasmids was mixed with a ratio of three parts (TAU HA DNA) to one part (tdTomato DNA) before addition of Lipofectamine™ 2000. The total DNA amount per well was not altered, i.e. 0.25 µg per 24-well (1 µg per 6-well) was used for one well of a 24-well plate.
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted September 24, 2020. ; https://doi.org/10.1101/2020.06.26.173526 doi: bioRxiv preprint

Fixation & Immunofluorescence
Fixation and immunofluorescence of cell cultures was done as described previously 73 with slight modifications. In brief, cells were fixed for 1 h with 3.7 % formaldehyde (in PBS), permeabilized and blocked with 5 % bovine serum albumin (BSA, Sigma-Aldrich) and 0.1 % Triton X-100 (AppliChem) in PBS for 5 minutes, incubated with the primary antibody (diluted in PBS) either for 2-3 h at 20-22 °C or preferentially at 4 °C for 16-18 h, thoroughly washed, and incubated with the corresponding secondary antibody (diluted in PBS), which was coupled to AlexaFluor™ fluorophores, for 1-2 h at 20-22 °C. For detection of ANKG and TRIM46 levels, the fixation process was modified (30 minutes instead of 1 h) according to the sensitivity of AIS proteins to fixation procedures reported in previous studies 74, 75

Microscopy
Immunostained cells were imaged with a widefield fluorescence microscope (Axioscope 5, Zeiss), using an LED excitation lamp (Colibri 7, Zeiss), a fluorescence camera device (Axiocam 503 mono, Zeiss), objectives with 10 x, 20 x, 40 x (air-based) and 63 x (oil-based) magnification (Zeiss) and Zen imaging software (Zen Blue pro, Zeiss). Exposure time and light intensity were adjusted to avoid excessive sample bleaching and to preclude saturation of fluorescent signal detection. For the analysis of protein expression levels, all images were taken with identical settings regarding laser intensity and exposure time to ensure statistical comparability. For documentation during cell cultivation, images were taken with a brightfield light microscope (DM IL LED, Leica) using objectives with 10x and 20 x (air-based) magnification (Leica) and Las X imaging software (Leica).

Data analysis
Differentiation Efficiency. The efficiency of SH-SY5Y cell differentiation was defined as relative number of SH-SY5Y-derived neurons compared to undifferentiated and non-neuronal cells. SH-SY5Y cells were considered as differentiated if they fulfilled the following morphological criteria: i) roundish shaped soma, ii) changed nuclear morphology, i.e. smaller and brighter nuclear signal, iii) axonal outgrowth exceeds soma diameter at least twofold and iv) determined axon length exceeds 50 µm. Further analysis was performed only with cells reaching these criteria.
Axonal outgrowth. Axonal outgrowth of SH-SY5Y-derived neurons was determined using the Fiji/ImageJ software (NIH). The processes were traced manually, and axon length was determined as pixel number, transformed to absolute distances with respect to the original image resolution. The measured section reached from the soma-to-axon-border to the most distal axonal part, which was still detectable. Included cells fulfilled the described differentiation criteria and the axon was distinctive from processes of adjacent cells.

Endogenous protein expression.
For the analysis of total endogenous expression of the neuronal maturation markers TAU and MAP2, SH-SY5Y-derived neurons were analyzed with Fiji/ImageJ software. Entire cells were encircled manually, and the mean fluorescence intensity (MFIneuron) of this marked region of interest (ROI) was determined. To exclude background fluorescence, MFI of an empty area next to the neuron was determined in the same manner (MFInoise) and subtracted from the MFIneuron. Further, MFI was determined in absence of primary antibodies (MFIunspecific), and also subtracted from MFIneuron to eliminate impact of unspecific binding of the secondary antibody. Protein levels of differentiated cells were normalized to that of undifferentiated cells. To ensure comparability of protein levels, all cell cultures of one experiment were cultivated on the same plate, immunostained and mounted in parallel with identical antibody solutions and imaged with identical microscope settings.
Axonal enrichment of endogenous TAU. By using Fiji/ImageJ, a random patch within the soma, which had no overlap with the nuclear signal, was set as first ROI for each analyzed SH-SY5Y-derived neuron.
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted September 24, 2020. ; https://doi.org/10.1101/2020.06.26.173526 doi: bioRxiv preprint An axonal segment of around 5 -10 µm in 50 -75 µm distance to the soma was chosen as second ROI. TAU signals were measured as MFI for both somatic and axonal ROIs (MFIsoma, MFIaxon). To exclude background fluorescence, MFI of empty areas next to the soma and axon were determined in the same manner (MFInoiseSoma, MFInoiseAxon) and subtracted from the MFIsoma/MFIaxon. The ratio of MFIaxon and MFIsoma was calculated (axon-to-soma-ratio). The axon-to-soma-ratio of the volume marker tdTomato was determined using identical ROIs and identical calculation steps. The axon-to-soma-ratio of TAU was normalized to this of tdTomato to achieve the axonal enrichment factor of TAU (AEFTAU). Only tdTomato-positive SH-SY5Y-derived neurons were included for analysis.
Axonal enrichment of TAU HA constructs. ROI selection and calculation of the axonal enrichment factor of TAU HA constructs (AEFTAU_HA) were done as described above for endogenous TAU with a few deviations. Instead of TAU signals, the intensity of anti-HA signal was measured. Only cells that were successfully co-transfected with tdTomato and TAU HA were included for analysis. Further, to overcome signal of unspecific anti-HA antibody binding, MFIs of somatic and axonal ROIs from untransfected cells (MFIunspecificSoma, MFIunspecificAxon) were measured and averaged for each experiment and subtracted from all MFIs of TAU HA from transfected cells.
Plot profiles of the proximal axon. The expression profiles of ANKG, TRIM46 and MAP2 proteins were determined at the proximal axon of SH-SY5Y-derived and mouse primary neurons. By using the Fiji/ImageJ software, the first 75 µm of the axon, starting with the soma-to-axon-border, were traced with a segmented line (thickness: three pixels), and a plot profile was generated. To exclude influence of background fluorescence, a plot profile of the region right next to the axon was measured and subtracted from the original value. Further, a plot profile was determined in absence of primary antibodies and subtracted from the sample values to eliminate impact of unspecific binding of the secondary antibody.
Protein enrichment at the proximal axon. For the calculation of the protein enrichment at the proximal axon, the section of the mean plot profiles, where the fluorescent signal reached at least 50 % of the maximum value, was determined in mouse primary neurons. All values within this section were averaged, separately for each individual experiment. In order to achieve the enrichment factor of ANKG andTRIM46 at the proximal axon, this average value was normalized against a baseline value of each experiment. This baseline value was acquired by averaging the most distal values of the plot profile. Here, the number of included values was equal to the value number of the previously determined section at the proximal axon. The identical sections used for mouse primary neurons were then applied to the corresponding plot profiles of SH-SY5Y-derived neurons to determine the protein enrichment of ANKG and TRIM46 in the same manner.
Statistical analysis. All statistical calculations were performed using PRISM™ analysis software (V8.3, GraphPad Inc.). For comparisons between one group of samples and a constant number (e.g. 1.0) or between two groups of samples, a two-sided student's t-test was used for the determination of p-values. If the standard deviation differed significantly among the two groups (p ≥ 0.05), the t-test was conducted with Welch's correction. When multiple groups were tested for significant differences, a Brown-Forsythetest was applied to check if the standard deviation (SD) differed significantly (p ≥ 0.05) among the tested groups. If this was the case, a Brown-Forsythe and Welch's ANOVA test with Dunnett's T3 correction for multiple comparisons was conducted. If the SD was not significantly different (p < 0.05), an ordinary one-way ANOVA test with Tukey's correction for multiple comparisons was conducted. The chosen statistical tests are mentioned for all experiments in the corresponding figure caption. The Gaussian distribution of all data sets was confirmed by a Shapiro-Wilk normality test prior to statistical analyses. Significance levels are symbolized with asterisks if p < 0.05 (*), p < 0.01 (**) or p < 0.001 (***), p < 0.0001 (****).
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted September 24, 2020. ; https://doi.org/10.1101/2020.06.26.173526 doi: bioRxiv preprint

RA/BDNF treatment produces homogeneous cultures of highly polarized neurons
We first aimed to investigate whether SH-SY5Y-derived neurons are suitable for studying TAU sorting. For this, naïve human SH-SY5Y neuroblastoma cells (Fig. 1A) were differentiated using four distinct procedures (Fig. 1B). Cells were either treated only with retinoic acid (RA) until day 7 of differentiation (d7), with RA until d7 followed by brain-derived neurotrophic factor (BDNF) in serum-free conditions until d14, with the phorbol ester TPA until d9 or with RA until d4 followed by TPA until d9. To compare the amount of successfully differentiated neurons, cells were classified as differentiated when they fulfilled several criteria, e.g. roundish shaped somata and axonal processes of at least 50 µm length (see methods for details). While treatment with RA (52.5 % ± 11.8 % neurons), TPA (39.7 % ± 10.2 %), and RA/TPA (47.4 % ± 5.4 %) generated mixed cultures of non-neuronal and neuronal cells, the sequential RA/BDNF treatment resulted in high numbers of SH-SY5Y-derived neurons (74.1 % ± 11.4 %) (Fig. 1C&D).
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted September 24, 2020. ; https://doi.org/10.1101/2020.06.26.173526 doi: bioRxiv preprint Taken together, SH-SY5Y-derived neurons, which were treated with RA/BDNF, showed the highest differentiation efficiency and pronounced neuronal maturation, indicated by axonal outgrowth and the highest expression levels of TAU and MAP2 protein.

Endogenous TAU is efficiently sorted in SH-SY5Y-derived neurons upon RA/BDNF treatment
To further evaluate the suitability of SH-SY5Y-derived neurons for TAU sorting studies, the efficiency of endogenous TAU sorting was determined. For this, the cells were transfected with tdTomato-expressing plasmids at day 5 of differentiation (d5), and the axonal enrichment factor of TAU (AEFendoTAU) was defined as axon-to-soma-ratio of TAU, normalized to the axon-to-soma-ratio of the randomly distributing volume marker tdTomato (see methods for details).
Since the results of the comparative analyses revealed RA/BDNF-treated SH-SY5Yderived neurons as the most suitable system in many respects, i.e. we observed high numbers of neurons, pronounced neuronal polarity, efficient sorting of endogenous TAU, and unchanged viability up to nine days after transfection, only those neurons were used for all further experiments.

SH-SY5Y-derived neurons show robust sorting of overexpressed 0N3R-TAU HA
Efficient axonal sorting of transfected and overexpressed TAU constructs is an oftenfaced problem in primary neuron models 50 . Therefore, we checked whether SH-SY5Yderived neurons achieve an endogenous-like sorting efficiency of transfected human wild type TAU. For this, we co-transfected SH-SY5Y-derived neurons with tdTomato and the shortest TAU isoform 0N3R, fused to an N-terminal HA tag (0N3R-TAU HA ) (Fig. 4B), at d5 of differentiation. The subcellular distribution was again quantified at d14 (Fig. 4A&C). The same experiments were done in mouse primary forebrain neurons, with co-transfection at div6 and quantification at div9 (Fig. 4D).
Strikingly, the axonal enrichment of the overexpressed 0N3R-TAU HA construct (AEFTAU_HA = 7.4 ± 1.3) was similar to that of endogenous TAU in RA/BDNF-treated SH-SY5Y-derived neurons at d14 (~ 94 % of AEFendoTAU = 7.9 ± 1.2) (Fig. 4A&C). In mouse primary neurons, 0N3R-TAU HA was also axonally sorted (AEFTAU_HA = 7.1 ± 3.2), but less efficient than the endogenous mouse TAU (~ 49 % of AEFendoTAU = 14.5 ± 0.6) (Fig. 4D, Suppl.Fig. 2). These results demonstrate that SH-SY5Y-derived neurons are able to sort overexpressed TAU with similar efficiency as endogenous . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted September 24, 2020. ; https://doi.org/10.1101/2020.06.26.173526 doi: bioRxiv preprint TAU. This makes SH-SY5Y-derived neurons suitable for addressing key questions of TAU sorting behavior, such as the role of TAU domains, modifications, or interactions within the proximal axon.

TAU HA lacking the C-terminus shows no axonal sorting in SH-SY5Y-derived and mouse primary neurons
In order to demonstrate this suitability, we next evaluated the sorting behavior of a truncated TAU HA version. As former data suggested an important role of the N-terminus for axonal sorting 43 , we used a construct which lacks the entire C-terminal half (N-term-TAU HA ) (Fig. 4B). The sorting efficiency of N-term-TAU HA was analyzed in SH-SY5Yderived and mouse primary neurons. Interestingly, N-term-TAU HA was axonally enriched neither in SH-SY5Y-derived neurons (AEFTAU_HA = 0.97 ± 0.03) nor in mouse primary neurons (AEFTAU_HA = 0.67 ± 0.05) (Fig. 4A,C,D, Suppl.Fig. 2). Moreover, Nterm-TAU HA tended to accumulate in the soma, indicated by HA-positive puncta in some neurons (Fig. 4A, Suppl.Fig. 2). These results, which are consistent in both used neuronal model systems, strongly suggest that the N-terminal half of TAU is either not involved in the process of axonal sorting or at least not sufficient for its successful implementation.

ANKG and TRIM46 are not enriched at the AIS region in SH-SY5Y-derived neurons despite efficient axonal TAU sorting
Our data illustrate the pronounced neuronal polarity of SH-SY5Y-derived neurons, including the efficient axonal targeting of endogenous and transfected TAU. Since the axon initial segment (AIS) is essential for the development and maintenance of neuronal polarity 35,37 , we evaluated the enrichment of two AIS proteins, Ankyrin G (ANKG) and tripartite motif-containing protein 46 (TRIM46), in these cells (Fig. 5). ANKG is a master organizer of the AIS architecture as it recruits AIS-located scaffold proteins, ion channels or cytoskeletal elements 35,37,38 . TRIM46 is at MT-organizing protein that was shown to enrich at the proximal axon prior to ANKG and to promote the polarized distribution of TAU 39,77 .
In mouse primary neurons, we observed robust expression of TRIM46 and ANKG at the proximal axon, already at div9 (Fig. 5D-K) with a strong enrichment of both proteins at the proximal axon (Fig. 5M&N). In addition to TAU (Fig. 4D, Suppl. Fig. 2), also MAP2 shows a polarized subcellular distribution, as it is tightly restricted to the somatodendritic compartment (Fig. 5L). This is in line with previous studies on neuronal polarity that suggest the ANKG-dependent AIS formation to be necessary for somatic MAP2 restriction 35,39,78 . The somatic retention of MAP2 is less efficient in SH-SY5Yderived neurons, indicated by the more than 3-fold increased distance of the halfmaximal intensity (SH-SY5Y: 17.7 µm distal of soma-to-axon-border; primary: 5.8 µm), but not completely absent (Fig. 5I&L). Strikingly, almost no detectable levels of ANKG were observed at the proximal axon of SH-SY5Y-derived neurons (Fig. 5A,C,G), suggesting the absence of a classical AIS. This is supported by the random-like distribution of ANKG across the axon (Fig. 5M). Further, only low TRIM46 levels were . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted September 24, 2020. ; https://doi.org/10.1101/2020.06.26.173526 doi: bioRxiv preprint seen across the axon (Fig. 5B,C,H) without significant enrichment at the proximal part (Fig. 5N). This means that SH-SY5Y-derived neurons can efficiently sort endogenous TAU and overexpressed 0N3R-TAU HA without detectable enrichment of ANKG or TRIM46 at the proximal axon. Thus, the process of axonal TAU sorting in neurons may be independent of ANKG and TRIM46 accumulation at the AIS.
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted September 24, 2020. ; https://doi.org/10.1101/2020.06.26.173526 doi: bioRxiv preprint

Discussion
Mislocalization of the axonal TAU protein is a hallmark of Alzheimer's disease and related tauopathies. In this study, we first aimed to establish a suitable neuronal cell model, derived from SH-SY5Y cells, for studying axonal TAU sorting. We then used these SH-SY5Y-derived neurons and mouse primary forebrain neurons i) to evaluate the subcellular sorting of a C-terminus-lacking version of TAU, and ii) to test whether the efficient axonal TAU sorting in SH-SY5Y-derived neurons depends on the enrichment of ANKG and TRIM46 at the proximal axon.

RA/BDNF-treated SH-SY5Y-derived neurons are suitable for TAU sorting research
In the first step, we evaluated whether SH-SY5Y-derived neurons are a suitable model system for studying axonal TAU sorting. We found that sequential treatment with RA and BDNF generated the highest number of neurons with the most pronounced axonal outgrowth, high expression levels of TAU and MAP2 and efficient axonal sorting of TAU. As seen in former studies 56,79,80 , only the administration of BDNF allowed the use of serum-free media, which explains the pure and thereby stable neuronal cultures. Culture purity, neuronal polarity and TAU sorting efficiency were inferior in SH-SY5Yderived neurons treated with RA, TPA or TPA. Based on these findings, all further experiments were performed with RA/BDNF-treated SH-SY5Y-derived neurons.
An often-faced problem of rodent primary cultures is the inefficient sorting of transfected and overexpressed TAU constructs 42,50,81 . This effect was also observed in our study, where axonal enrichment of 0N3R-TAU HA was roughly 50 % less than for endogenous TAU in mouse primary neurons. In contrast, the axonal enrichment 0N3R-TAU HA reached endogenous-like levels (~ 94 % of endogenous TAU) in SH-SY5Yderived neurons. One explanation for this striking difference might be that SH-SY5Yderived neurons tolerate the overexpression of 0N3R-TAU HA and tdTomato for at least nine days. Successfully transfected primary neurons largely died after more than three days of overexpression (data not shown). The longer expression time may facilitate the axonal targeting of 0N3R-TAU HA as the plasmid DNA may become degraded after several days. This results in a reduction of protein overload and allows more efficient sorting of the exogenous protein. Another cause might be species-specific differences of the TAU sorting machinery, which affects the ability to sort human TAU isoforms in mouse primary neurons 49 . Further, mouse primary neurons might express a different set of isoforms that is more efficiently sorted than 0N3R-TAU. However, 0N3R-TAU was the most efficiently sorted isoform in former studies 41,81 , and 0N3R-TAU is known to be the predominant isoform in early developmental stages 3,82 , probably including div9 neurons derived from E13.5 embryonic mice. For SH-SY5Y-derived neurons, the predominant expression of 0N3R-TAU was reported 62 .
In brief, SH-SY5Y-derived neurons share properties of often-used neuronal models for TAU sorting as rodent primary neurons [40][41][42][43] or iPSC-derived neurons [44][45][46][47][48] . But these cells overcome several of their limitations, as they are human-derived, cultivatable at low cost and without the animal need, and as a neuroblastoma cell line readily . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted September 24, 2020. ; https://doi.org/10.1101/2020.06.26.173526 doi: bioRxiv preprint accessible for genetic engineering, also within the TAU-encoding MAPT gene 68,69 . Moreover, transfected TAU HA is sorted similar to endogenous TAU, which is difficult to achieve in rodent primary cultures 42,50,81 .

Axonal TAU sorting requires the C-terminal half of TAU
In the next step, we used SH-SY5Y-derived and mouse primary neurons to analyze whether TAU is still efficiently sorted when it lacks the C-terminal half (N-term-TAU HA ). No axonal enrichment of N-term-TAU HA was seen in both used neuronal cell models. Moreover, N-term-TAU HA tends to accumulate within the soma of transfected neurons. This may hint at the increased accumulation tendency in the absence of the C-terminus or enhanced cellular degradation resulting from protein quality control mechanisms. In the latter case, N-term-TAU HA could be enriched in subcellular compartments, such as lysosomes or stress granules. This somatic localization of N-term-TAU HA is interesting since recent studies claimed a major role for the very N-terminal TAU domain 43 and the P-rich domain 42 in mediating axonal TAU sorting. As both domains are present in N-term-TAU HA , they are either not involved in TAU sorting or at least not sufficient to maintain the intracellular sorting process without the C-terminus. Notably, the Cterminal repeat domains of TAU are critical for the MT-binding affinity of the entire protein 13,14 . As there is evidence that MTs and MT architecture play a role in the TAU sorting process 34,39,41 , this could explain the necessity of the C-terminus during axonal sorting. However, recent findings question that the C-terminal repeat domains have a strong impact on efficient TAU sorting 42 .
Thus, more effort is required to shed light on the TAU-intrinsic properties that ensure efficient TAU sorting. In this context, SH-SY5Y-derived neurons are valuable to conduct further and more comprehensive studies, e.g. by comparative analyses of differently truncated or modified TAU constructs.

Axonal TAU sorting is independent of ANKG and TRIM46 enrichment at the proximal axon
The formation of the AIS at the proximal part of the axon is critical for the development and maintenance of neuronal polarity, and for the polarized distribution of proteins 35,37 . There is strong evidence that the AIS and the integrity of its architecture play a major role in regulating the anterograde and retrograde trafficking of TAU 33,34,40,41 .
Therefore, we checked for the formation of an AIS in SH-SY5Y-derived neurons. We first measured the protein levels of Ankyrin G (ANKG), a master organizer of the AIS architecture [35][36][37] , at the proximal axon of SH-SY5Y-derived and mouse primary neurons. Strikingly, no ANKG was detected within the putative AIS region of SH-SY5Yderived neurons, in contrast to the mouse primary neurons. The robust axonal sorting of endogenous and transfected TAU does apparently not require ANKG enrichment at the proximal axon or detectable ANKG expression at all although ANKG is responsible for the recruitment of major AIS structural components 35 . The tracing of MAP2 signal revealed invasion of MAP2 into the proximal axon. A lack of MAP2 polarity is reported for ANKG-deficient cultured neurons 38,83 . However, the low MAP2 levels more distal . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted September 24, 2020. ; https://doi.org/10.1101/2020.06.26.173526 doi: bioRxiv preprint than 50 µm clearly prove a still existing, albeit reduced, ANKG-independent somatic MAP2 retention in these cells (see Figs. 2&5).
Recent studies revealed the tripartite motif-containing protein 46 (TRIM46), an axonal MT-organizing protein 39,84 , as a major player for the initial development of neuronal polarity 39 . TRIM46 accumulates at the proximal axon prior to ANKG and forms highly polarized MT fascicles 38,39 . TRIM46-deficiency leads to TAU missorting in cultured primary neurons 77 . Therefore, we investigated the TRIM46 levels at the proximal axon of SH-SY5Y-derived neurons. While the mouse primary neurons showed a clearly defined TRIM46 accumulation at the AIS, there were only small amounts of TRIM46 protein along the axon without a significant enrichment at the putative AIS region.
The absence of a classical AIS, indicated by the lack of ANKG and TRIM46 enrichment, together with the predominant 0N3R isoform expression 62 might hint at the incomplete maturity of SH-SY5Y-derived neurons 35 . However, SH-SY5Y-derived neurons express several neuronal maturation marker proteins [56][57][58][59][60][61][62] , and there are neuron subtypes known to lack ANKG-dependent AIS formation as well 85 . The spatial separation of endogenous TAU and MAP2 in SH-SY5Y-derived neurons is slightly less efficient compared to mouse primary neurons, which could be a direct consequence of the ANKG and TRIM46 absence, as both were shown to promote somatic retention of MAP2 83 and axonal sorting of TAU directly 39,77 .
As ANKG and TRIM46 are absent and thus do not mediate axonal sorting of TAU in SH-SY5Y-derived neurons, the question arises, which remaining factors mediate the robust axonal targeting of TAU seen in these neurons. The end-binding protein 3 (EB3) was recently shown to be critical for the AIS integrity 86,87 . EB3 accumulates and binds along the MT network at the AIS 87,88 , and links it to ANKG. This EB3-ANKG interplay is necessary for the proper assembly and maintenance of the AIS by regulating the MT architecture 86,87 . As EB3 also binds to the plus end of growing axons, it could directly promote anterograde trafficking of TAU and other axonal proteins via a piggyback-like mechanism 89,90 . Interestingly, TAU and EB3 were recently shown to co-localize at the AIS of iPSC-derived cortical neurons 48 . A central role in AIS assembly and maintenance was also shown for the MT-binding protein MTCL-1 91 .
The relevance of TRIM46 and other MT-binding proteins in maintaining neuronal polarity 39,86,87,91 strongly suggest that the organization and regulation of the MT network at the AIS is one key factor for its structural and functional integrity, which includes the correct intracellular sorting of MAP2, TAU and other compartment-specific proteins 38 . However, our results question the necessity of TRIM46 for the development of neuronal polarity and polarized Tau distribution. Interestingly, TAU and EB3 co-localize at the AIS of iPSC-derived cortical neurons 48 . It is still unclear whether this colocalization is relevant for TAU trafficking or due to the MT affinity of both proteins. Taken together, the role of MT-organizing proteins in the context of AIS integrity and axonal TAU sorting remains inconclusive and needs further elaboration, with SH-SY5Y-derived neurons as a potential tool.
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted September 24, 2020. ; https://doi.org/10.1101/2020.06.26.173526 doi: bioRxiv preprint

Conclusion
In this study, we demonstrate that SH-SY5Y-derived human neurons, generated with RA/BDNF treatment, are a suitable tool for investigating TAU sorting mechanisms in human cells. These neuronal cells show pronounced neuronal polarity, neuronal marker expression, efficient sorting of endogenous and, importantly, of transfected TAU and TAU constructs similar to mouse primary neurons. By using SH-SY5Yderived neurons, we show that the N-terminal half of TAU is not sufficient for successful axonal sorting. Further, axonal TAU targeting is independent of classical AIS formation, as ANKG and TRIM46 enrichment at the proximal axon is absent in SH-SY5Y derived neurons, but TAU is still efficiently sorted into the axon. Taken together, our data show i) that the C-terminal half of TAU is required for efficient axonal sorting, and ii) that subcellular sorting of TAU might be largely independent of classical AIS formation, which questions the claimed role of ANKG, TRIM46 and other MT-organizing proteins in developing and maintaining neuronal polarity.
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in  An ordinary one-way ANOVA with Tukey's correction for multiple comparisons was performed to determine significance levels between all four treatments (C), and all four treatments and undifferentiated cells (D&E). Significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001.
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted September 24, 2020. ; https://doi.org/10.1101/2020.06.26.173526 doi: bioRxiv preprint Figure 3: Axonal enrichment of endogenous TAU in SH-SY5Y-derived neurons. SH-SY5Y-derived neurons were transfected with tdTomato and immunostained with a polyclonal anti-TAU (K9JA) antibody. A: Endogenous TAU (green) and transfected tdTomato as volume marker (red) are shown in single channels and merged. Note the strong axonal sorting of TAU compared to tdTomato in SH-SY5Yderived neurons after RA/BDNF treatment (arrowheads), which leads to yellowish somatic staining (arrows) and mainly green axonal TAU staining in merged images. Axonal sections (small dashed boxes) are shown with higher magnification (large dashed boxes) in merged images. Scale bar: 20 µm, scale bar within dashed box: 3 µm. B: Quantification of the axonal enrichment of endogenous TAU (AEFendoTAU), normalized to intracellular tdTomato distribution in neuronal cells (see methods for details). Quantification was done for three to five independent experiments with 20-30 cells per experiment. Black dots represent independent experiments, grey bars indicate the arithmetic mean of all experiments, and error bars show SD. An ordinary one-way ANOVA with Tukey's correction for multiple comparisons was performed to determine significance levels between all conditions. Significance level: *** p < 0.001.
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in  Merge . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted September 24, 2020. ; https://doi.org/10.1101/2020.06.26.173526 doi: bioRxiv preprint three independent experiments with 20-55 cells per experiment (C), or for three to four independent experiments with 16-46 cells per experiment (D). The quantification of AEFendoTAU in SH-SY5Y-derived neurons is included for comparison (striped bar, see Fig. 3A). Black dots represent independent experiments, grey bars indicate the arithmetic mean of all experiments, and error bars show SD. An ordinary one-way ANOVA with Tukey's correction for multiple comparisons was performed to determine significance levels between endogenous TAU and the TAU HA constructs. Significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001, ns: p ≥ 0.05.
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted September 24, 2020. ; https://doi.org/10.1101/2020.06.26.173526 doi: bioRxiv preprint CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted September 24, 2020. ; https://doi.org/10.1101/2020.06.26.173526 doi: bioRxiv preprint (G&J), ANKG (H&K), and MAP2 (I&L) signals at the proximal axon of SH-SY5Y-derived (G-I) and mouse primary neurons (J-L). The colored lines indicate the arithmetic mean of three to four independent experiments with 28-36 traced cells per experiment (see methods for details). The grey area shows SD. G,H&J,K: The green and red dotted lines mark the corridor where the ANKG/TRIM46 signal are above 50 % of the maximum value (determined for the signals of mouse primary neurons in J&K). I&L: The MAP2 plot profiles were normalized to 1 prior to analysis. The blue dotted lines indicate the distance where the MAP2 signal reaches 50 % of the maximum value. M&N: Enrichment of ANKG (M) and TRIM46 (N) at the proximal axon (i.e. within the marked corridor), normalized to the baseline (see methods for details). Equal d. = equal distribution (reference value = 1). Quantification was done for three to four independent experiments with 28-36 cells per experiment. Black dots represent independent experiments, colored bars indicate the arithmetic mean of all experiments, and error bars show SD. A two-sided t-test was performed to determine significance levels between the cell models and 1. A two-sided t-test with Welch's correction was performed to determine significance levels between both cell models. Significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001.
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted September 24, 2020. ; https://doi.org/10.1101/2020.06.26.173526 doi: bioRxiv preprint