EHS
EHS

Stress induced TDP-43 mobility loss independent of stress granules


Establishment of a model system for live cell single-molecule tracking of TDP-43

To study TDP-43 aggregation, its mobility in different cellular compartments and its role in stress granules on a single-molecule level, we engineered dual-transgenic TDP-43-Halo and G3BP1-SNAP cell lines (see Methods). As the attachment of bright, membrane permeable and photo-stable fluorophores is a key requisite for single-molecule studies, the commonly used HaloTag12 and SNAP-Tag13 systems were used in this study. To ensure that the fusion of the HaloTag (33 kDa) to TDP-43 does not alter the protein’s localization and function, the HaloTag was fused to either the N- or the C-terminus of TDP-43, named HaloTDP-43 and TDP-43Halo, respectively (Fig. 1a, Supplementary Fig. 1a). To spatially assign stress granules during image analysis we used G3BP1SNAP as a stress granule marker.

Fig. 1: Generation and quality control of TDP-43Halo cell line.
figure 1

a Schematic overview of the TDP-43Halo and G3BP1SNAP constructs. b Spinning disk confocal images of the TDP-43Halo cell line under unstressed conditions (red: TDP-43-TMR, green: G3BP1-SiR, scale bar 20 µm). c Spinning disk confocal images of immune-labeled naïve H4 cells as well as TDP-43Halo cells under unstressed conditions (cyan: anti-G3BP1-Alexa532, magenta: anti-TDP-43-Alexa647, scale bar 20 µm). d Spinning disk confocal images of the TDP-43Halo cell line under 30 min and 60 min sodium arsenite treatment (red: TDP-43-TMR, green: G3BP1-SiR, scale bar 20 µm and 2 µm). e Spinning disk confocal images of the immunostained TDP-43Halo cell line and naïve H4 cells under 60 min sodium arsenite treatment (magenta: anti-TDP-43-Alexa647, cyan: anti-G3BP1-Alexa532, scale bar 20 µm and 2 µm). f Western Blot overview of the TDP-43Halo cell line and naïve H4 cells stained with anti-vinculin, anti-TDP-43 or anti-G3BP1 antibodies showing proper expression of the transgenic constructs. The experiment was performed in triplicates. g Quantification of the overexpression for the TDP-43Halo cell line compared to endogenous TDP-43 in naïve H4 cells shows a 1.5–2x overexpression of TDP-43Halo, n = 5 independent preparations, data are displayed as the mean +/− STD. Source data are provided as a Source Data file.

We first ensured proper functionality of the tagged TDP-43 and G3BP1 constructs in the transgenic cell lines by Sir-Halo and TMR-SNAP staining, as well as immunostaining of endogenous proteins, and compared observed localizations as a function of stress duration to that in naïve H4 cells (Fig. 1b, c). In contrast to a diffusive cytoplasmic G3BP1 signal under unstressed conditions, G3BP1 positive stress granules containing TDP43Halo could be observed after 30 min and 60 min of sodium arsenite stress (Fig. 1d), which was also confirmed by immunostaining (Fig. 1e).

We then quantified the total of TDP-43Halo and G3BP1SNAP overexpressed proteins by Western blot analysis (Fig. 1f). Densitometric analysis showed an over-expression of 1.5–2 fold for the TDP-43Halo construct compared to endogenous TDP-43 in naïve H4 cells (Fig. 1g). We focused on C-terminally tagged TDP-43Halo, since it allows for the visualization of both, full-length and fragmented TDP-43 species (Fig. 1f). Complementary data for the N-terminally tagged TDP-43 construct (HaloTDP-43) are given in the supplementary materials (Supplementary Figs. 1 and 2).

C-terminally tagged TDP-43 showed the formation of a prominent 35 kDa TDP-43 fragment and an increased cytoplasmic localization, which was not seen for the N-terminally tagged TDP-43. Fragmentation of TDP-43 leads to the disruption or complete abolishment of the nuclear localization sequence (NLS)14,15. A disrupted or missing NLS leads to a subsequent accumulation of TDP-43 fragments in the cytoplasm, and could explain the observed slightly higher cytoplasmic level of TDP-43Halo as compared to the naïve H4 cells (Fig. 1c).

Taken together, the attachment of either Tag the respective protein, did not alter the formation of stress-induced phase-separated compartments or their interaction of G3BP1 and TDP-43 with latter. This establishes the transgenic cell lines as a model system for studying stress-induced dynamical changes of TDP-43 mobility using single-molecule imaging of Halo-tagged TDP-43 within different cellular compartments.

Single-molecule tracking monitors the region-specific TDP-43 mobility

To monitor TDP-43 in a region-specific manner, we established a single molecule tracking pipeline, using photoactivatable Janelia Fluor 646 for labeling of Halo-tagged TDP-4316 and TMR-labeling for G3BP1SNAP (Fig. 2a). The usage of a photoactivatable dye enabled labeling at high concentrations (nM range) and single-molecule tracking concentrations were achieved by continuously activating only a small subset of labeled TDP-43Halo per frame. Thus, a high number of frames could be recorded while continuously controlling the frame-wise emitter density17,18.

Fig. 2: Overview of the single molecule tracking experiments and diffusion analysis.
figure 2

a Labeling and illumination scheme. TDP-43Halo and G3BP1SNAP are labeled with PA-JF646-HaloTag ligand (HTL) and TMR-SNAP-Tag ligand (STL) for 1 h at 37 °C and 5% CO2 24 h before imaging. TDP-43Halo is illuminated continuously with 640 nm and the PA-JF646-dye is continuously activated with 405 nm, the TMR-dye is illuminated with 532 nm and imaged every 200 frames. b Overview of the region assignment and region-specific track assignment (cytoplasm/tracking region: white outline and green tracks; nucleus: blue outline and tracks; stress granules: red outline and tracks). Zoom-ins and exemplary tracks for all regions are shown and show both, mobile TDP-43Halo as well as more immobile molecules. Scale bar 5 µm. ce Shuttling of TDP-43Halo in and out of stress granules. The c and d. Number of shuttling events significantly decreases with increasing stress duration (c: not normalized to SG size, values are displayed as median +/− interquartile range, p-value = < 0.0001, d: normalized to SG size, values are displayed as mean +/− STD, p-value = 0.0006). The ratio of TDP-43Halo tracks going in and out of stress granules stays constant over all stress durations (e), values are displayed as median +/− interquartile range. Statistical test: Two-tailed Mann–Whitney test. The number of analyzed cells per condition (n number) is given in Supplementary Table 1, the experiments cells were examined in independent experiments. Source data are provided as a Source Data file.

TDP-43Halo-PA-JF646 molecules were imaged for 120 min under unstressed or stressed conditions (0.5 mM sodium arsenite) using continuous 405 nm activation. In addition, every 200 frames (i.e. every 1.34 s) the G3BP1SNAP-TMR channel was recorded. The movies were grouped into 20 min time slots and tracking and diffusion analysis was performed (see Methods). Figure 2b shows the general experimental workflow for region assignment and tracking analysis. Fluorescence from the G3BP1SNAP constructs was used in a control channel to assign cellular regions for analysis. The cellular outline (white line) and the nuclear outline (blue line) were drawn manually (see Methods). The stress granules (red line) were assigned by an intensity threshold of the G3BP1SNAP signal and tracked dynamically over the whole movie. In each frame, single TDP-43Halo molecules were localized and their position was linked through successive frames of the movie, yielding single-molecule tracks (Fig. 2b). Tracks crossing from one region to another (e.g. from the cytoplasm to stress granules) were split at the region border and the parts of the tracks are assigned to the respective region (Fig. 2b).

TDP-43Halo dynamic stress granule shuttling decreases with stress duration

In a first step, we assessed the stress-dependent shuttling of TDP-43Halo molecules in and out of stress granules (Fig. 2c–e). Figure 2c shows a significant decrease of the number of shuttling events with increasing stress duration. This effect is still visible when the number of shuttling events are normalized to the stress granule size (Fig. 2d). Such a decrease in the number of shuttling events can indicate either an increased interaction strength of TDP-43Halo inside of stress granules, or an overall decreased stress granule dynamics, potentially due to solidification of the granule11,19. Interestingly, the ratio between TDP-43Halo molecules entering or leaving the stress granules stayed constant over the whole stress period (Fig. 2e). Taken together, these observations can be explained by a decrease in stress granule dynamics with increasing stress duration and might hint towards a gain of solid-like properties in stress granules.

Sodium arsenite stress reduces TDP-43 mobility

To investigate the mobility changes of TDP-43 under sodium arsenite stress, single-molecule tracking analysis was employed (Fig. 3). To biochemically validate the nature of TDP-43 slow-down, additionally, the solubility of TDP-43Halo and endogenous TDP-43 was assessed using a solubility assay and subsequent Western Blotting.

Fig. 3: Sodium arsenite stress leads to a reduction of TDP-43Halo mobility observed in single-molecule tracking.
figure 3

a Jump distance analysis and 3-exponential fitting results giving three different diffusion coefficients and fractions, slow (D1/F1), medium (D2/F2) and fast (D3/F3), in a stress-time-course experiment of TDP-43Halo cells. Stress duration increased from unstressed (magenta) to 120 min (green) of 0.5 mM sodium arsenite stress. Data are presented as mean values overlayed with the corresponding data points displaying the movie wise spread of data. b Stress time course of the effective diffusion coefficient Deff for the TDP-43Halo construct (whole cell, mean + STD, two-tailed Welch’s t-test, p-value = <0.0001). Source data are provided as a Source Data file. c. Solubility assay of the TDP-43 construct. Solubility was assessed under unstressed conditions and different stress time-points. An increasing insoluble TDP-43 fraction was observed with increasing stress duration (unstressed, 40 min, 80 min and 120 min of 0.5 mM sodium arsenite treatment, anti-Vinculin, anti-TDP-43, n = 3). d. Analysis of the different diffusion constants (slow/D1/magenta, medium/D2/light rose, fast/D3/green) and the respective fraction within the stress time-course experiment (whole cell). For all experiments, the data are presented as mean values +/− STD and the standard deviations were calculated from the movie-wise distribution of the plotted value and statistical significance was assessed with a multiple unpaired t-test with Welch’s correction. P-value ranges: <0.0001: ****, 0.0002: ***, 0.0021: **, 0.032: *, 0.123: ns. The number of analyzed cells per condition (n number) is given in Supplementary Table 1, the experiments cells were examined in independent experiments.). Source data are provided as a Source Data file.

From the single molecule tracks, displacement histograms and cumulative displacement histograms are computed from all jumps within the tracks subjected to analysis (Fig. 3a). Diffusion coefficients and the respective fractions were computed by fitting the cumulative displacement histograms with a multi-exponential fit function20,21 (see Methods). For the extraction of the apparent diffusion coefficients, a three-exponential fit function was used, since it fitted best (compared to mono- and double exponential) to the cumulative jump-distance histogram (Supplementary Fig. 3). This results in three diffusion regimes (D1/slow, D2/medium, D3/fast). The three fractions (F1/slow, F2/medium, F3/fast) indicate the proportion of TDP-43Halo molecules in each diffusion regime. An overview of the fitting parameters and fit quality is given in supplementary Fig. 3. Three diffusion coefficients allow to account not only for the assessment of a bound and mobile fraction but also for the aspect of anomalous diffusion22,23. To reduce the bias towards slower moving and bound molecules, only the first five jumps of every tracked single-molecule time-trace were used in the data analysis23, however, similar results were obtained when all jumps were considered (Supplementary Fig. 4). For simplicity the effective diffusion coefficient Deff was computed (see Methods), representing the weighted average of all diffusion coefficients and respective fractions.

Figure 3a depicts the data obtained from detected TDP-43Halo tracks, recorded under unstressed and different stress conditions. In order to obtain a first quantitative comparison about TDP-43Halo mobility as a function of stress duration, the effective diffusion coefficient Deff of TDP-43Halo was determined for the whole cell (Fig. 3b). After 120 min of sodium arsenite stress, TDP-43Halo showed a significant reduction in mobility as compared to the unstressed condition (unstressed: Deff = 4.94 µm2/s, standard deviation: 0.84 µm2/s, stressed: Deff = 1.82 µm2/s, standard deviation: 0.45 µm2/s, statistical test: Welch’s t-test), suggesting oligomerization or aggregation of TDP-43Halo or localization to small compartments that restrict mobility (Fig. 3b). As shown in Fig. 3c, an increase of the insoluble TDP-43Halo fraction was observed with increasing stress, starting between 40 and 80 min after stress onset, although the band is rather faint as compared to the soluble fraction. Under unstressed conditions, however, all TDP-43 species were found in the soluble fraction. When comparing these data to the single molecule tracking data, it is likely that the reduced mobility of TDP-43Halo with increasing stress is partly caused by the formation of insoluble TDP-43 aggregates. The reduction of Deff is very pronounced, however, the bands depicting the insoluble TDP-43 fraction are rather faint. This indicates an aggregate-independent mechanism of TDP-43 mobility reduction, since the small amount of insoluble TDP-43 detected by the Western blot alone, cannot explain the observed strong reduction in TDP-43Halo mobility.

To get more detailed insight into the contribution of different TDP-43Halo species (fast, medium, slow) on the mobility reduction, the different diffusion states and the respective fractions were analyzed (Fig. 3d).

Figure 3d shows, that TDP-43Halo mobility reduction is caused by a decrease of the fast diffusion coefficient (D3) accompanied also by a reduction of the fraction of fast TDP-43Halo species (F3) resulting in an increase of both slow fractions, F2 and F1. Together, this suggests a general slow-down of TDP-43Halo with increasing stress duration, caused by mobility reduction and the formation of less-mobile TDP-43 species.

To get further biochemical insights into TDP-43 oligomerization and aggregation size-exclusion chromatography combined with TDP43 dot blotting was performed (Supplementary Fig. 5). After 120 min of sodium arsenite stress, a clear shift of TDP43 species eluting at earlier fractions (from 44 ml onwards) compared to late eluting TDP43 species (84 ml onwards) in the non-stressed condition was observed (Supplementary Fig. 5). This indicates higher oligomeric or aggregated species after 120 min of sodium arsenite stress and thus supporting the theory of TDP-43 oligomerization and aggregation with increasing sodium arsenite stress.

Several control measurements were conducted. As depicted in Supplementary Figs. 6a and 6b, no reduction of the effective diffusion Deff coefficient was observed for the C- and N-terminally tagged TDP-43 constructs during 120 min of measurement time under unstressed conditions. These results verify that the tracking environment or other external factors do not slow down TDP-43Halo. Also, the mobility of the HaloTag alone was assessed under stressed conditions. In this case, no mobility reduction was seen during 120 min of sodium arsenite stress (Supplementary Fig. 6c), thus excluding an unspecific stress-related mobility reduction.

Moreover, we also observed a similar effect of TDP-43 mobility reduction for the N-terminally tagged TDP-43 construct (Supplementary Fig. 7) ensuring that the observed decrease in TDP-43 mobility is independent of the HaloTag position.

Table 1 gives an overview of the effective diffusion coefficient Deff of the HaloTDP-43 and TDP-43Halo constructs at different time-points (unstressed, 60 and 120 min of sodium-arsenite stress, whole cell). The comparison shows, that HaloTDP-43 generally displays a lower mobility than the TDP-43Halo construct (as expected due to the lack of fragments in the case of HaloTDP-43) and that also for HaloTDP-43 a stress-related slow-down of mobility is observed.

Table 1 Comparison of the effective diffusion coefficient Deff between TDP-43Halo and HaloTDP-43 constructs at different stress durations (unstressed, 60 min and 120 min 0.5 mM incubation with sodium arsenite, error given as the standard deviation (STD))

In addition, we assessed TDP-43Halo and HaloTag mobility under 0.4 M D-Sorbitol, an oxidative and osmotic stressor (supplementary fig. 8)24. Sorbitol stress leads to an immediate, significant reduction of TDP-43Halo mobility in the whole cell and all cellular regions. Also for the HaloTag alone, we observed a decreased mobility throughout the whole cell. An immediate and comparable decrease of TDP-43Halo and HaloTag mobility under Sorbitol stress strongly argues for a general effect caused by the osmotic stressor. It was previously shown that sorbitol stress leads to cell shrinkage and an overall reduced mobility due to crowding effects25,26, which is in agreement with our data.

Sodium arsenite is an oxidative stressor and was shown to damage mitochondria and cause ATP depletion27. Furthermore ATP depletion can lead to a loss in mobility as reported for several different proteins while other proteins seem to be unaffected26,28,29,30. For this reason, we assessed intracellular ATP-concentrations under unstressed and sodium arsenite stress conditions (60 and 120 min) and found an insignificant reduction of ATP levels after sodium arsenite stress (Supplementary Fig. 9), which is most likely insufficient to solely explain the observed strong reduction in TDP-43Halo mobility.

To investigate the role of ALS-causing TDP-43 mutations on TDP-43 mobility, we engineered TDP-43Halo constructs bearing either one mutation in the alpha-helical structure (M337V)14,31 and another mutation in the Glycine-/Serine-rich domain (A382T)14,32 (Supplementary Figs. 10, 11). Single-molecule tracking did not show any significant alterations in the course of mobility of mutant TDP-43 as compared to the wild-type constructs (Supplementary Figs. 12, 13), suggesting that the selected mutants do not have an additional effect on TDP-43 mobility with increasing stress. Before stress application and at low stress conditions we observed a faster mobility for familial mutants M337V and A382T (Supplementary Figs. 12, 13). Together, these results suggest that the pathological effect of the A382T and M337V mutations may not be based on an overall faster aggregation of mutated TDP-43 species.

Longer stress leads to less efficient recovery of slow TDP-43 species

To find out whether and to which extent the reduced mobility of TDP-43Halo can be reversed, we again employed our single-molecule tracking analysis and bulk solubility assessment. Reversibility of TDP-43 slow-down was assessed by stressing TDP-43Halo cells for 60 min and 120 min with sodium arsenite and subsequent single-molecule tracking until 4 h after stress removal (Fig. 4a–f; statistical evaluation Supplementary Fig. 14). While short stress exposure (1 h) seems to allow almost complete restoration of TDP-43Halo mobility after 4 h of recovery, longer stress exposure (2 h) initiated processes hindering complete amelioration of TDP-43 mobility reduction (Fig. 4a, b and Supplementary Fig. 14a).

Fig. 4: Stress and recovery experiment.
figure 4

a, c, e Stress and recovery time courses of the effective diffusion coefficient Deff and the diffusion coefficients D1 (slow), D2 (medium) and D3 (fast) and their respective amplitudes (F1, F2, F3) plotted for the whole cell for 1 h stress duration and up to 4 h of recovery (green: D3/F3, orange: D2/F2, red: D1/F1). Stress and recovery start points are marked by arrows. Source data are provided as a Source Data file. b, d, f Stress and recovery time courses of the effective diffusion coefficient Deff and the diffusion coefficients D1 (slow), D2 (medium) and D3 (fast) and their respective amplitudes (F1, F2, F3) plotted for the whole cell for 2 h stress duration and up to 4 h of recovery (green: D3/F3, orange: D2/F2, red: D1/F1). Stress start points are marked by arrows and the recovery period is highlighted in gray. The number of analyzed cells per condition (n number) is given in Supplementary Table 1, the experiments cells were examined in independent experiments. Source data are provided as a Source Data file. g Solubility assay of the TDP-43Halo wild-type after different stress and recovery durations (1: Stress 2 h, Recovery 4 h, 2: Stress 2 h, Recovery 2 h, 3: Stress 1 h, Recovery 4 h, 4: Stress 1 h, Recovery 1 h). Antibodies: anti-vinculin, anti-TDP-43, n = 3. For all experiments, the data are presented as mean values +/− STD and the standard deviations were calculated from the movie-wise distribution of the plotted value.

To get more insight, we again turned to the detailed analysis of the diffusion constants and respective fractions (D1/F1/slow, D2/F2/medium, D3/F3/fast) to test whether a reduced mobility is caused by a general slow-down of TDP-43Halo or a shift in the respective fractions. The fast diffusion coefficient D3 was significantly reduced after 2 h of stress and both recovery conditions as compared to the unstressed condition (supplementary Fig. 14b). For the 1 h stress condition, D3 was significantly decreased after 120 min of recovery. After 240 min of recovery, D3 adapted comparable values for both stress conditions, although the decrease for the 1 h stress condition was no longer significant with respect to the unstressed condition.

The fast fraction F3 was significantly reduced after 120 min of recovery (Supplementary Fig. 14b) and a longer stress duration led to a lower fast fraction than a shorter stress duration (1 h stress: F3 = 0.37, 2 h stress: F3 = 0.24). While 4 h after stress, F3 showed complete recovery for cells stressed for 1 h, for longer stress duration F3 remained reduced even after 4 h of recovery. This again underlines the interplay between two different effects that lead to a reduced TDP-43 mobility: (1) a decrease in mobility, and (2) an increase of immobile TDP-43Halo species.

To further biochemically characterize the nature of the reduced mobility after recovery, the previously described biochemical solubility assay was performed after either 1 h or 2 h of sodium arsenite stress and 2 h or 4 h of recovery, respectively (Fig. 4g). After longer stress exposure, TDP-43Halo showed a significant insoluble fraction, even after 4 h of recovery, while after short stress almost no insoluble TDP-43 was detected (Fig. 4g). The region-specific analysis of the recovery data is shown in supplementary fig. 15 and region-wise recovery is comparable to TDP-43Halo recovery observed in the whole cell. Note, after 1 h and 2 h of sodium arsenite stress, stress granules could be only assigned until 100 min and 120 min of recovery, respectively (Supplementary Fig. 15c, f), since they dissolved afterwards. Exemplary time-lapsed movies of G3BP1SNAP covering 120 min of sodium arsenite stress and covering 4 h of recovery are shown in supplementary movies 1 and 2, respectively.

Together, single-molecule tracking data and biochemical characterization indicate that TDP-43Halo is capable to recover from short stress insults while longer stress leads to persistent, insoluble TDP43 aggregates.

TDP-43 mobility is reduced in stress granules, nucleus and cytoplasm

To study TDP-43Halo mobility in different subcellular compartments and to identify the contribution of different subcellular fractions of TDP-43Halo to the overall mobility reduction seen in the whole cell, diffusion data were analyzed in a region-specific manner (Fig. 5).

Fig. 5: Region- and diffusion class specific analysis of stress-induced TDP43 mobility reduction.
figure 5

Stress time course of the effective diffusion coefficient Deff for the TDP-43Halo construct shown for the cytoplasm (a), the nucleus (b) and stress granules (c) (mean + STD, two-tailed Welch’s t-test, cytoplasm p-value = <0.0001, nucleus p-value = <0.0001, stress granules p-value = 0.0010). The number of analyzed cells per condition (n number) is given in Supplementary Table 1, the experiments cells were examined in independent experiments. Source data are provided as a Source Data file.

TDP-43Halo showed the highest mobility in the cytoplasm (Fig. 5a) under unstressed conditions with an effective diffusion coefficient of 5.40 µm2/s (standard deviation: 1.07 µm2/s) and sodium arsenite stress lead to a significant and continuous reduction of the effective diffusion coefficient to 2.23 µm2/s (standard deviation: 0.59 µm2/s) after 120 min of sodium arsenite stress. Since the HaloTag was attached to the C-terminus of TDP-43 and TDP-43 can be fragmented N-terminally33, the construct visualizes both, full-length as well as fragmented TDP-43Halo. A high mobility in the cytoplasm might therefore reflect the presence of full-length and fragmented TDP-43. Under unstressed conditions, TDP-43Halo mobility in the nucleus was in general slower as compared to the cytoplasm (Fig. 5b, Deff = 4.20 µm2/s, standard deviation: 1.15 µm2/s), indicating an often bound or confined state of TDP-43 in the nucleus11,34. Notably, also in the nucleus Deff reduced continuously with stress duration reaching the lowest value after 120 min with Deff = 1.26 µm2/s (standard deviation: 0.50 µm2/s).

Stress granules are often discussed as the site of TDP-43 aggregation within the cell8,35,36. To test this hypothesis, we measured stress-dependent mobility of TDP-43Halo within stress granules (Fig. 5c). First stress granules could be assigned after 20 min of sodium arsenite stress. Although we observed a substantially lower TDP-43Halo mobility in the stress granules first assigned at 20 min (Deff = 2.05 µm2/s, standard deviation: 0.60 µm2/s), compared to the unstressed value in the cytoplasm (Deff = 4.20 µm2/s, standard deviation: 1.15 µm2/s), prolonged stress only led to a modest further reduction of mobility within stress granules reaching at 120 min a value of Deff = 1.45 µm2/s (standard deviation: 0.49  µm2/s). In fact, Deff in the cytosol after 120 min was comparable to that in stress granules during early stress. Thus, the decrease in mobility in the cytoplasm and nucleus between 40 min and 120 min after stress induction was far more dramatic than observed in SGs.

Taken together, we found region-specific effects of stress-induced TDP-43Halo slow down within different cellular regions. Within stress granules TDP-43Halo showed, already after short stress durations, an expected, strong reduction of mobility as compared to TDP-43Halo in the cytoplasm under unstressed conditions, accompanied by a further, moderate decrease in mobility with prolonged stress. Surprisingly, we also found pronounced decreased TDP-43 mobility in the nucleus and the cytoplasm upon sodium arsenite stress, suggesting that aggregation or oligomerization also occurs outside of stress granules.

Super-resolution imaging reveals inhomogeneous spatial distribution and mobility of TDP-43

Previous studies indicated, that stress granules are not necessarily homogenous phase-separated compartments but can exhibit regions of higher density termed ‘core’ that are surrounded by a less dense ‘shell’18,37. Using super-resolution microscopy, it was reported that stress granule components like G3BP1 or poly(A)-RNA localize in a distinct substructure within stress granules37,38. Thus, we were interested if TDP-43 exhibits a similar substructure within stress granules at different stress time points.

To obtain such high-resolution sub-compartmental spatial information immunolabelled TDP-43 and G3BP1 were imaged using stimulated emission depletion (STED) microscopy under unstressed and different stress durations (30 min, 60 min and 120 min of 0.5 mM sodium arsenite) in naive H4 cells (see Methods). TDP-43 shows an inhomogeneous distribution with denser regions (higher intensity) and less dense regions (lower intensity) within stress granules (Supplementary Fig. 16). This supports the idea of an inhomogeneous distribution of TDP-43 within G3BP1-positive stress granules. In contrast G3BP1 appears more homogenously distributed throughout the stress granules. This could in part be attributed to a more saturated fluorescence signal due to the high G3BP1 density within stress granules.

The results of the STED measurements which were obtained in fixed cells indicate an inhomogeneous distribution of TDP-43 within different cellular compartments and in particular stress granules. Next, we wanted to determine whether such inhomogeneities were correlated with local mobility changes and therefore turned again to live-cell single-molecule imaging.

Single-molecule tracking analysis showed a slow-down of TDP-43Halo movement with increasing stress duration (Figs. 3 and 5). A reduced mobility can be caused by several processes, e.g., by localization to a confined compartment, by interaction with other protein complexes, by pathological aggregation or by physiological interactions. Also solubility assessment of TDP-43 under different stress condition (Fig. 3c), confirmed an insoluble fraction with increasing stress exposure. To study the spatial distribution of the different TDP-43 mobility regimes, we performed TALM analysis (tracking and localization microscopy)18,39,40 and investigated local TDP-43Halo diffusivities (diffusion mapping, DM) as a means to obtain a super-resolved diffusivity map20,41.

In TALM analysis, the fitted position of every detected spot is marked and thereby a super-resolved image can be created from the tracking data. A more frequent localization of TDP-43 at a given location, indicates binding to cellular structures or aggregation events. Frame-to-frame position jumps of localized molecules can be converted into average diffusion coefficients for each pixel in the image20, indicating regions of local high or low mobility. The combination of these two methods allows for a correlation between binding hotspots or regions with increased TDP-43 localization and local mobility patterns. The visualization of such patterns can help to elucidate the origin of the observed TDP-43 slow-down in the cytoplasm and stress granules.

Figure 6a shows an exemplary G3BP1 image and the respective TDP-43Halo TALM and DM images recorded under unstressed and stressed conditions. The G3BP1 image was used to assign the cellular regions, in particular the location of stress granules (see Methods). To clearly separate stress granules from the stress-granule-free cytoplasm, a conservative approach in stress granule detection was chosen, resulting in a slightly overestimated stress granule area (see Methods). For the diffusion mapping (DM) images, higher mobility is depicted with a blue and lower mobility with a red color scheme (Fig. 6a). Note, in order to increase sensitivity, mobility greater than 6 µm2/s is depicted in dark blue.

Fig. 6: TALM (tracking and localization microscopy) and displacement mapping (DM) analysis of TDP-43 movement in H4 cells.
figure 6

a Overview of representative images obtained with TALM and DM analysis for TDP-43Halo imaged under unstressed and stressed (80–100 min and 100–120 min sodium arsenite treatment) conditions (color map: blue = fast movement, red = slow movement). Bottom: Color bar for the DM analysis. Blue color depicts faster movement and red color depicts slow movement. The maximum allowed jump distance for the analysis was 5 pixels. For better visualization, mobility >6 µm2/s is displayed in dark blue. b Crops of TALM and DM images of regions within the cytoplasm and stress granules. Local binding hotspots (TALM, white spots) correlate with a reduced TDP-43 mobility (DM, red regions) in both regions. An overview of the number of analyzed cells per condition giving similar results is given in Supplementary Table 1.

Under unstressed conditions, TDP-43Halo showed numerous binding hotspots or clusters in the nucleus and stress granules and regions of increased TDP-43 localization, termed localization patches, in the cytoplasm, that all correlated with local low TDP-43Halo mobility (red).

Binding hotspots and localization patches appeared as bright spots (or regions) within TALM images, originating from frequent TDP-43 localizations from one, but more typical from several TDP-43 molecules (Supplementary Fig. 17).

Under unstressed conditions, diffusivity mapping clearly showed that despite the binding hotspots and localization patches, TDP-43 diffusion was strongly dominated by high TDP-43Halo mobility (blue dominated DM image) (Fig. 6a, upper panel). Such a mobility pattern of TDP-43 could be explained by the physiological shuttling between the nucleus and cytoplasm34.

For longer stress duration (80–100 min, 100–120 min), diffusivity mapping showed a shift towards medium and low TDP-43Halo displacements throughout the whole cell. This fits well to the observed TDP-43Halo mobility reduction in all cellular compartments obtained with single-molecule tracking (Fig. 3). To get more detailed insight into the TDP-43Halo behavior in different regions, Fig. 6b shows exemplary cropped areas from the cytoplasm and stress granules. TDP-43Halo showed distinct binding hotspots within stress granules under both stress conditions. These observations can be explained by localized ‘binding hotspots or clusters’ within stress granules as previously observed for G3BP1 and IMP118 and are consistent with the inhomogeneous stress granule structure observed with STED microscopy (supplementary fig. 16). The binding hotspots in stress granules correlate with a low TDP-43Halo mobility in DM (red spots, Fig. 6b). These binding hotspots were caused by the repeated localization of TDP-43 to these regions since an overlay between a TALM image and track start points from all molecules localized in one movie showed that several TDP-43Halo tracks are originating from these binding hotspots detected within stress granules (Supplementary Fig. 17a). The same is observed for the cytoplasm, however, here a more dispersed distribution of track start points can be seen. In addition, Supplementary Fig. 17b depicts a kymograph of a stress granule region, showing repeated binding of several TDP-43Halo molecules during the measurement period.

Based on the unexpected finding that also in the cytoplasm reduced mobility of TDP-43Halo was observed with increasing stress, the spatial distribution of TDP-43Halo in the cytoplasm was of special interest. Figure 6b shows an inhomogeneous distribution of TDP-43Halo localizations within the cytoplasm (TALM). Most notably, areas with frequent localizations correlate with an observed reduced TDP-43Halo mobility (DM). We refer to these areas as cytoplasmic patches which were assigned using a mobility threshold as described in the Methods part. We observed that the cytoplasmic area, covered by these patches, is significantly increasing with stress duration (Fig. 7a), indicating stress-related patch formation. Furthermore, we performed an anisotropy analysis of TDP-43Halo diffusivity in the cytoplasm, stress granules and cytoplasmic patches as described previously17, to assess anomalous diffusion and potential trapping of TDP-43 in these regions (See Methods, Fig. 7b). TDP-43Halo showed an anisotropy value of around 1 inside the cytoplasm (excluding the patches) at unstressed and all stress conditions, indicating pure Brownian motion. Inside stress granules TDP-43Halo exhibited an anisotropy value of around 1.3, implying some confinement and trapping of TDP-43 inside these granules. Interestingly, TDP-43Halo showed the highest anisotropy values in a range between 1.7 and 2.8 inside the cytoplasmic patches, indicating strong anomalous diffusion and trapping of TDP-43Halo. Note, a similar trend was also found when analyzing regions of identical sizes (Supplementary Fig. 18), ruling out a that the observed differences in anisotropies are caused by the different sizes of the respective regions. These differences in TDP-43 diffusion within stress granules, cytoplasmic patches and the rest of the cytoplasm also are evident in the computed angular plots of TDP-43 motion in the respective regions (Supplementary Fig. 19). Furthermore, diffusion analysis showed a very low effective diffusion coefficient Deff in the range of 0.49–0.34 µm2/s inside of the cytoplasmic patches (Fig. 7c) further indicating a strong confinement, oligomerization or aggregation inside these patches.

Fig. 7: Quantification of cytoplasmic TDP-43 patches and TDP-43 localization hotspots.
figure 7

a Analysis of the cytoplasmic area covered by the patches at unstressed and different stress conditions, showing a significant increase of patch-covered cytoplasmic area with stress (stat. test: two-tailed Mann−Whitney, p-value = <0.0001, data are displayed as the median +/− the interquartile ranges). b Anisotropy analysis of TDP-43Halo molecules located in the cytoplasm (outside of cytoplasmic patches), stress granules and cytoplasmic patches at unstressed and different stress conditions. Values are displayed as the mean +/− the STD from 50 resamplings performed with 50% of the data. c Diffusivity analysis of TDP-43Halo molecules located in the cytoplasm (outside of cytoplasmic patches), in stress granules and in cytoplasmic patches at unstressed and different stress conditions. The data are presented as mean values +/− STD and the standard deviations were calculated from the movie-wise distribution of diffusion coefficients. d Analysis of TDP-43Halo localization cluster density inside of stress granules and the cytoplasm (stat. test: two-tailed Mann–Whitney, p-value = <0.0001). Data are displayed as the median +/− the interquartile ranges). For all displayed data (ae) the number of analyzed cells per condition (n number) is given in Supplementary Table 1, the experiments cells were examined in independent experiments. Source data are provided as a Source Data file.

To further exclude unspecific stress effects, we assessed stress-related patch formation of the HaloTag alone (Supplementary Fig. 20). The HaloTag alone showed a similar patch-covered area as compared to TDP-43Halo under unstressed conditions. However, and most importantly, the patch-covered area did not increase with increasing stress duration for the HaloTag alone, excluding a general mechanism of stress-related patch formation.

In addition to the analysis of diffusion in cytoplasmic patches, we further characterized the TDP-43 localization hotspots or clusters observed in the TALM images (Fig. 6) using DBSCAN cluster analysis (see Methods). Figure 7d shows that the density of TDP-43 localization clusters found within stress granules was increasing with sodium arsenite stress, whereas the observed cluster density in the cytoplasm (including the cytoplasmic patches) was independent of stress. Moreover, the observed stress-induced cluster density in stress granules reached a level of around 0.76 cluster per µm2, which is significantly higher than the cluster density observed in the cytoplasm (0.1 cluster per µm2). Importantly, the detections per cluster and the mean cluster area were not different between the cytoplasm and stress granules (Supplementary Fig. 21). Binding hotspots inside stress granules exhibited on average around 350 localizations per cluster. With an average track length of TDP-43Halo inside of stress granules of 10 frames, this corresponds to ~35 TDP-43 binding events per cluster. The mean cluster area was determined to be ~0.015 µm2, corresponding to a 2D circle with a diameter of ~138 nm. This finding is consistent with observations made by Niewidok et al. determining a size of ~150–200 nm for stress granule binding regions for the stress granule proteins G3BP1 and IMP118. Also Jain et al. determined the size of stress granule cores to be around 200 nm37.

In addition, TALM and DM analysis showed TDP-43 localization accompanied by a reduced TDP-43Halo mobility to a vesicle-like compartment (Examples are shown in Fig. 8). Interestingly, TDP-43 was not equally distributed throughout the whole vesicle, but was localized more often at the outside than at the center of the vesicle and single TDP-43Halo tracks show TDP-43 movement along the vesicle outline, seemingly avoiding the interior (Fig. 8a). TDP-43Halo mobility was strongly decreased in the vesicles (Deff = 0.522 µm2/s, standard deviation: 0.170 µm2/s), as compared to TDP-43 mobility inside of stress granules (Deff = 1.27–1.05 µm2/s, Fig. 4c). One explanation for this behavior is the diffusion of a TDP-43 aggregates anchored to the vesicular membrane.

Fig. 8: Tracking and localization microscopy (TALM) and displacement mapping (DM) analysis of vesicular structures.
figure 8

a DM (left) and TALM (middle) images showing exemplary vesicular structures observed within the cytoplasm. Selected TDP-43Halo tracks overlayed on top of the TALM image (right) show confined movement along the outline of the exemplary vesicles. Scale bar 1 µm, n = 9. TDP-43Halo mobility is displayed in a range of 0 µm2/s (red) to 6 µm2/s (blue). b Diffusivity analysis of TDP-43Halo molecules localized within the vesicular structure show a strongly reduced TDP-43Halo mobility as compared to the mobility observed within stress granules (n = 9 independent cells). Number of analyzed cells per condition for TDP-43 within stress granules is listed in Supplementary Table 1. For all experiments, the data are presented as mean values +/− STD and the standard deviations were calculated from the movie-wise distribution of the plotted value. Source data are provided as a Source Data file.

Taken together, TALM and DM analysis are valuable tools for the visualization and better interpretation of single-molecule tracking data. With these tools, we were able to show patches of increased TDP-43Halo localization in the cytoplasm that additionally correlate with a reduced mobility. The area covered by these patches increased with stress duration and showed a very low TDP-43Halo mobility. Diffusion of TDP-43Halo inside these patches is highly anomalous, indicating potential trapping of TDP-43Halo to these regions. Together, the results using the described methods suggest that TDP-43Halo is capable to oligomerize or aggregate in the cytoplasm independent of stress granules as a result of stress.



Source link

EHS
Back to top button