SH-SY5Y cells were cultured in DMEM/F12, GlutaMAX [Thermo Fisher Scientific (TFS)] supplemented with 10% fetal bovine serum (FBS, Biochrom AG), and 1 × antibiotic/antimycotic solution (TFS) in a humidified incubator at 37°C, 5% CO₂.

Primary mouse neurons were isolated and cultured as described before (Zempel and Mandelkow, 2017) with slight modifications. In brief, brains of FVB/N mouse embryos were dissected at embryonic day 13.5. Brainstem and meninges were removed and the whole cortex was digested with 1 × trypsin (PAN-Biotech). The cell suspension was diluted in prewarmed (37°C) neuronal plating medium [Neurobasal media (TFS), 1% FBS, 1 × antibiotic/antimycotic solution (TFS), 1 × NS21 (PAN-Biotech)] seeded onto coated plates, and cultivated in a humidified incubator at 37°C, 5% CO₂. Four days after plating, the medium was doubled with neuronal maintenance media [Neurobasal media (TFS), 1 × antibiotic/antimycotic solution (TFS), 1 × NS21 (PAN-Biotech)], and cells were treated with 0.5 μg/ml AraC (Sigma-Aldrich).

SH-SY5Y cells were co-transfected with expression plasmids containing Dendra2c-TAU (TAU^D2) isoforms and tdTomato-N1-EB3 by using Lipofectamine2000 (TFS) according to the manufacturer’s protocol. Control cells were transfected with tdTomato-N1-EB3 and an empty Dendra2c plasmid. Only cells showing both Dendra2c and tdTomato signal were used for analysis. Two days after transfection, cells were transferred into a live-cell imaging chamber (ALA Scientific) and EB3 comets of a single cell were imaged for 120 s (1 frame per second) with a Leica DMi8 microscope (Leica). The original movie file was processed by ImageJ software (Schindelin et al., 2012; Schneider et al., 2012) as follows: An average minimum projection was calculated from all frames and subsequently subtracted from all. Afterwards, a threshold was set and microtubule dynamics were analyzed by ImageJ plugin TrackMate (Tinevez et al., 2017), using LoG detector with an estimated blob size of 2 px. The following parameters were examined: cell area (μm²), microtubule number/μm², microtubule run length (μm), microtubule stability (s), and microtubule growth rate (μm/s). Data were filtered for quality ≥1.5, track start >0 s, and track end <120 s. All experiments were conducted in three replicates and at least five cells were analyzed per condition. Cell area was measured from at least 30 cells. Shapiro–Wilk test was performed to test for normal distribution of data; afterwards, statistical analysis was performed by either one-way ANOVA with correction for multiple comparisons (Tukey’s test) for normally distributed data or Kruskal–Wallis test with Dunn’s correction for multiple comparison of non-parametric data using GraphPad Prism software (v8.0.1). The statistical test performed is highlighted in the figure legends for each graph.

SH-SY5Y cells were lysed with RIPA buffer [50 mM HEPES pH 7.6, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% sodium dodecyl sulfate, 0.5% sodium deoxycholate, 1 × cOmplete ULTRA EDTA-free protease inhibitor cocktail (Merck)]. Samples were diluted with SDS buffer and separated on 10% polyacrylamide gels. Proteins were transferred to PVDF membranes and blocked in 5% non-fat dry milk in TBS-T. Membranes were incubated with the primary antibody overnight at 4°C, washed with TBS-T, and incubated with the corresponding secondary HRP-coupled antibody for one hour at room temperature. The used antibodies are listed in Table 1. Luminescent signals were detected with ChemiDoc XRS + system (Bio-Rad).

To analyze TAU axonal enrichment, primary mouse neurons were transfected as described before with Lipofectamine2000 (TFS) (Zempel et al., 2017b). In brief, neurons were co-transfected with the corresponding N-terminally HA-tagged TAU (TAU^HA) isoform and tdTomato. After two days, neurons were fixed and stained for endogenous mouse Tau with a pan-TAU antibody, or HA for exogenous TAU^HA and MAP2 as described before (Zempel and Mandelkow, 2017), using antibodies listed in Table 1. After imaging on a fluorescence microscope (Axioscope 5, Zeiss), axonal sorting was analyzed by measuring mean fluorescence intensities (MFI) of tdTomato and TAU in the axon and soma. Axons were identified by the absence of MAP2 and the presence of TAU signal, branching pattern of ∼90°, a length of above 300 μm, and constant diameter. Axonal enrichment factor (AEF) was calculated as follows: AEF = (MFI_(TAU,Axon)/MFI_(TAU,Soma))/(MFI_(tdTom,Axon)/ MFI_(tdTom,Soma)). The experiment was performed in three replicates and at least 10 cells were analyzed per condition. For the analysis of TAU expression, neurons were transfected with TAU^HA isoforms, fixed, and stained for TAU by using a pan-TAU antibody and HA-tagged TAU by using an anti-HA antibody (see Table 1 for details). Expression was analyzed by measuring the MFI of pan-TAU in the soma and axon of untransfected and transfected cells.

Shapiro–Wilk test was performed to test for normal distribution of data; afterwards, statistical analysis was performed by either one-way ANOVA with correction for multiple comparisons (Tukey’s test) for normally distributed data or Kruskal–Wallis test with Dunn’s correction for multiple comparison of non-parametric data using GraphPad Prism software (v8.0.1). The statistical test performed is highlighted in the figure legends for each graph.

The antibodies used for this study are listed in Table 1.

TAU is considered an axonal protein in mature neurons (Binder et al., 1985). However, previous results indicate a difference in axonal enrichment of the six human TAU isoforms that were overexpressed as Dendra2c-tagged fusion proteins in primary rodent neurons (Zempel et al., 2017a). To further investigate TAU isoform axonal sorting and rule out the influence of a big protein tag (such as Dendra2c with approx. 26 kDa) on cellular localization, all six human TAU isoforms were fused to an HA-tag (TAU^HA) and co-transfected for two days with a volume marker (tdTomato) into primary mouse neurons aged for 7–9 days (DIV). One benefit of the HA-tag is the relatively small size, which leads to a size shift of the TAU isoforms of only 1.1 kDa (Figure 1A). Control neurons were transfected only with tdTomato as a volume marker, assuming an unbiased axodendritic distribution after two days of expression. To quantify axonal targeting of the different versions of TAU, we normalized the axonal presence of TAU against the unbiased distribution of tdTomato, expressed as the AEF (see Methods for details). To distinguish axons and dendrites, neurons were fixed and stained for endogenous mouse TAU or transfected human TAU^HA using an anti-HA antibody and the dendritic marker MAP2. Axons were identified by the absence of MAP2, the presence of TAU or HA signal, and further morphological criteria (see Methods for details; Figure 1B). Differences in axonal sorting of TAU were observed for the different TAU^HA isoforms (Figures 1C,D). For endogenous TAU, which mostly consists of 0N3R in prenatal rodent brains (McMillan et al., 2008; Bullmann et al., 2009), a strong axonal localization is visible from the immunofluorescence images, indicated by a strong TAU signal in the axon and only low levels in the soma and dendrites (AEF of ∼10, Figures 1C,D). Strong axonal enrichment was also observed in neurons expressing 0N3R-TAU^HA. In contrast, neurons expressing 2N4R-TAU^HA show lower levels of axonal TAU but still strong axonal targeting compared to tdTomato (Figure 1C). To compare the different TAU^HA isoforms quantitatively, axonal enrichment of TAU was calculated from soma-to-axon ratio of TAU fluorescence intensity and normalized to soma-to-axon ratio of the tdTomato signal (Figure 1D). Axonal enrichment of endogenous TAU was approximately 10-fold higher compared with tdTomato, while the human TAU^HA isoforms showed approximately 4- to 6-fold enrichment (Figure 1D). From all six transfected isoforms, 0N3R-TAU^HA showed the strongest axonal enrichment, which reaches approximately 65% of endogenous TAU. A relatively efficient (but still ∼40% less than endogenous TAU) axonal sorting could also be observed for 1N3R-TAU^HA. Of note, both 2N-TAU^HA isoforms showed the weakest axonal enrichment. Differences in TAU expression levels were assessed by measuring mean fluorescence intensities of the pan-TAU signal in the soma and axon (MFI) of neurons expressing only endogenous TAU (endo. TAU) or neurons co-expressing endogenous TAU with TAU^HA (Figure 1E). Comparison of expression rates of endogenous TAU with neurons expressing endogenous and TAU^HA revealed only slight overexpression rates for the different TAU isoforms (up to 0.5 times), both in the soma and axon, which were not significantly higher than in control cells (Figure 1E). This might indicate that either primary neurons downregulate endogenous TAU expression to prevent overexpression of TAU or that the amount of exogenously expressed TAU is small compared with endogenous TAU. To rule out that differential expression and overexpression of TAU isoforms influences cellular sorting mechanisms, we correlated somatic TAU expression levels of endogenous TAU and TAU^HA with the axonal enrichment of TAU (Figures 1F–H). No significant correlation was observed for any of the experimental groups, indicating that the expression levels likely do not impact TAU sorting. Taken together, our results implicate that the efficiency of axonal sorting of TAU is independent of TAU expression levels. Furthermore, the subcellular distribution of TAU seems to be isoform dependent and is influenced by the number of N-terminal inserts, since 2N4R- and 2N3R-TAU^HA showed the lowest axonal enrichment. No significant difference was observed between the other 3R- and 4R-expressing neurons, hinting towards a repeat-independent sorting mechanism of the isoforms.

One main function of TAU is its role in microtubule stability and spacing (Weingarten et al., 1975; Witman et al., 1976). The influence of individual TAU isoforms on microtubule dynamics, however, has not been addressed in living cells. To investigate TAU isoform-specific effects on microtubules, SH-SY5Y neuroblastoma cells were transfected with TAU^D2 isoforms. The expression levels of TAU^D2 were confirmed by Western blot analysis (Figure 2A). Of note, only low levels of endogenous TAU (∼55 kDa) were observed in control and transfected cells, which most likely corresponds to 0N3R-TAU that is mainly expressed by undifferentiated SH-SY5Y cells (Figures 1A, 2A, long exposure) (reviewed in Bell and Zempel, 2020). Exogenously expressed TAU isoforms show a shift in size of approx. 26 kDa due to the fused D2-tag (Figure 2A). TAU^D2 overexpression rate was assessed by comparing the amount of endogenous and exogenous TAU and considering a transfection efficiency of ∼15–20% in SH-SY5Y cells, which resulted in a ∼50-fold overexpression. Since differentiated SH-SY5Y cells show an increase of about 12-fold of endogenous TAU expression, overexpression of D2-tagged TAU isoforms was normalized to expression in differentiated SH-SY5Y cells, which results in approx. four times higher final expression of exogenously expressed TAU.

Since microtubules are essential components of the cytoskeleton, initial analysis of TAU isoform-specific effects focused on changes in cell size, which may hint at cytoskeletal alterations. Cell area was significantly (∼20%) smaller in cells overexpressing 4R-TAU^D2 isoforms compared with control cells (Figure 2B). In addition, cells expressing 2N3R- and 0N3R-TAU^D2 had a significantly greater cell area than 4R-expressing cells (∼20%), comparable with untransfected cells, while 1N3R only showed a trend towards increased cell size compared with the 4R-TAU^D2 isoform-expressing cells (∼15% bigger cell size). To further investigate the role of TAU isoforms in microtubule dynamics, SH-SY5Y cells were co-transfected with the corresponding TAU^D2 isoform and tdTomato-N1-EB3 (Figures 2C–G). EB3 binds the growing microtubule plus-ends and can be used in live-cell imaging to track microtubule assembly (Stepanova et al., 2003). Expression of TAU^D2 was confirmed via the green fluorescence of the D2-tag in live-cell imaging. EB3 comets were recorded for 2 min (1 fps), images were processed as described in Figure 2C, and microtubule dynamics were analyzed afterward using TrackMate plugin for ImageJ (Tinevez et al., 2017). The overall number of microtubules was quantified and normalized to the corresponding cell area (Figure 2D). Microtubule count was slightly increased compared with the control cells (5.5 MTs/μm²) in cells expressing 2N4R- and 0N4R-TAU^D2 (7.2 and 7.3 MTs/μm², respectively). In contrast, 3R-TAU^D2 isoforms showed no difference in microtubule counts compared with control cells, despite a slight trend towards a reduced number of microtubules for 0N3R. Comparing 3R- and 4R-TAU^D2 isoforms showed that microtubule numbers were significantly decreased for 0N3R compared to 4R-TAU^D2-expressing cells (Figure 2D). To further investigate how microtubule counts might be affected by TAU^D2 isoforms, microtubule stability, total microtubule run length, and microtubule growth rate (Figures 2E–G) were analyzed. We observed several tendencies that, however, did not reach statistical significance: 1N- and 0N-TAU^D2 isoforms seem to cause a slight decrease in microtubule stability independent of the presence of the second N-terminal repeat (Figure 2E), and 2N4R-TAU^D2 seems to increase microtubule run length (Figure 2F). Microtubule growth rate was slightly, but not significantly, increased for individual TAU isoform-overexpressing cells compared with control cells (Figure 2G), but it was significantly increased (p < 0.0143, unpaired t-test) when we compared all TAU isoform-expressing cells to control cells. All in all, the results indicate that TAU isoforms have a significant impact on cell size and show slight differences in microtubule counts and impact on microtubule growth rate, without significant differential effects on microtubule stability and run length in undifferentiated SH-SY5Y cells.