Differential preference in protein co-aggregation

We next investigated if the proteins validated above formed the same or different aggregates by analyzing the pairwise co-aggregation of these proteins under heat stress. We co-transfected plasmids expressing each pair of the aggregation-prone proteins into RPE-1 cells to examine their propensity to form co-aggregates (coaggs). It turned out that not all proteins aggregated into the same inclusions, but each coalesced with its selective partners at different levels of efficiency (Fig. 2a–h, Supplementary Fig. ^3a). For example, TDP43 co-aggregated with G3BP1, SEC16A and NUP88 with sequentially decreasing levels but did not aggregate with RPA40 (Fig. 2a–d, h). Again, we observed heterogeneity in the degree of coaggs formation among cells. For example, the majority of cells co-expressing TDP43-mNG and Halo-SEC16A contained no or low numbers of TDP43/SEC16A coaggs, whereas some cells exhibited much more extensive co-aggregation (Fig. 2b). Such heterogeneity was partially attributed to the different expression levels of TDP43 and SEC16A after transient transfection, as there was weak positive correlation between co-aggregation levels and the integrated intensities of TDP43 and SEC16A (Supplementary Fig. ^{2c, d}). Nevertheless, TDP43/SEC16A co-aggregation was not solely determined by expression levels, as some cells with below-median expression levels formed coaggs and vice versa (Supplementary Fig. ^{2c, d}).

Open in a separate window

Fig. 2

Selective co-aggregation among aggregate-forming proteins.

a Representative images showing that, when co-transfected into RPE-1 cells, TDP43 co-aggregated with G3BP1 in stress granule (SG). In “Zoom out”, the yellow box indicates the region shown in zoom-in views, and green dashes demarcate the nucleus. Scale bars except in “Zoom Out” represent 1 μm, and represent 5 μm in “Zoom Out”. b Representative images for SEC16A inclusions enriching TDP43 mildly (<10%) and more extensively (≥ 10%) are shown. Arrowheads: co-aggregates (coaggs). Scale bars except in “Zoom Out” represent 1 μm, and represent 5 μm in “Zoom Out”. c Representative images showing TDP43 co-aggregation with NUP88 in the cytoplasm. Scale bars except in “Zoom Out” represent 1 μm, and represent 5 μm in “Zoom Out”. d Representative images showing that TDP43 rarely co-aggregated with RPA40. Arrows: TDP43 aggregates; triangles: RPA40 aggregates. Scale bars except in “Zoom Out” represent 1 μm, and represent 5 μm in “Zoom Out”. e Representative images showing that G3BP1 rarely co-aggregated with SEC16A. Arrows: G3BP1 aggregates; triangles: SEC16A aggregates. Scale bars except in “Zoom Out” represent 1 μm, and represent 5 μm in “Zoom Out”. f Representative images showing that G3BP1 rarely co-aggregated with NUP88. Scale bars except in “Zoom Out” represent 1 μm, and represent 5 μm in “Zoom Out”. g Representative images showing that NUP88 rarely co-aggregated with SEC16A. Arrows: NUP88 aggregates; triangles: SEC16A aggregates. Scale bars except in “Zoom Out” represent 1 μm, and represent 5 μm in “Zoom Out”. h Co-aggregation matrix of TDP43, G3BP1 and newly identified aggregation-prone proteins SEC16A, RPA40 and NUP88. The number in each colored square is the mean number of coaggs per cell ± standard deviation (s.d.) derived from analysis of the experiments as in (a–g) and Supplementary Fig. ^3a. n=45 cells examined over 3 independent experiments. i TDP43/SEC16A, TDP43/G3BP1 coaggs (i.e. TDP43-containing SGs) and SEC16A inclusions without TDP43 in the same cell, which are marked by arrowheads, arrows and triangles, respectively. Representative images from 3 independent experiments are shown. Scale bars except in “Zoom Out” represent 1 μm, and represent 5 μm in “Zoom Out”.

Notably, G3BP1, SEC16A and NUP88 rarely co-aggregated with each other even though they all showed some degree of co-aggregation with TDP43 (Fig. 2e–h). We further confirmed that the coaggs of TDP43/G3BP1 (i.e. TDP43-containing SGs) and TDP43/SEC16A are orthogonal types of aggregates by simultaneously labeling TDP43, G3BP1 and SEC16A with different fluorescent tags. Even though TDP43/G3BP1 and TDP43/SEC16A coaggs could co-exist in the same cells and some appeared adjacent to each other, they did not fuse or colocalize (Fig. 2i). Importantly, RNA fluorescence in situ hybridization (FISH) revealed that TDP43/G3BP1 coaggs contained abundant poly-adenine (polyA) RNAs, as previously reported⁵⁹, whereas TDP43/SEC16A coaggs were devoid of polyA RNA (Supplementary Fig. ^4a).

To examine whether TDP43 and SEC16A directly interact in co-aggregates, we employed a system of dimerization-dependent FP (ddFP) consisting of a pair of nonfluorescent FP-derived monomers that can stably fluoresce only when two proteins tagged by each part of the ddFP heterodimerize⁶⁰. When TDP43 and SEC16A were respectively tagged with mEGFP-ddRFP-B and ddRFP-A-Halo, TDP43/SEC16A coaggs emitted ddRFP fluorescence (Supplementary Fig. ^3b). As a control, G3BP1 tagged with ddRFP-B-EGFP was coexpressed with ddRFP-A-Halo-SEC16A. Although some SEC16A aggregates and G3BP1 aggregates (i.e. SGs) appeared adjacent to each other, no ddRFP fluorescence was detected (Supplementary Fig. ^3b).

Timelapse imaging revealed that TDP43 partitions dynamically between SGs and TDP43/SEC16A coaggs. In about half of the cells that formed SGs under heat stress, SGs were observed to undergo spontaneous dissolution to below 10% of the initial volume. In these cells, TDP43/SEC16A coaggs gradually emerged as SGs were disappearing (Supplementary Fig. ^{4d, e}, Supplementary Movie ¹), and exhibited a significantly higher increase of TDP43 intensity in coaggs with SEC16A when compared with cells that maintained SGs (Supplementary Fig. ^4f). We confirmed that such dynamic TDP43 partition was not due to G3BP1 overexpression in the above experiment, as the same phenomenon was observed in cells where the endogenous G3BP1 was tagged in-frame with mScarlet-I through CRISPR/Cas-mediated genome editing without ectopic G3BP1 expression (Supplementary Fig. ^4g). In addition, to test if the partition of TDP43 can also be modulated by the level of SGs, we overexpressed G3BP1 or used RNA interference (RNAi) to inhibit G3BP1/2 expression, which respectively stimulated or suppressed SG assembly (Supplementary Fig. ^{4h, i})^61,62. Correspondingly, the percentage of SEC16A inclusions accumulating TDP43 was respectively reduced or increased (Supplementary Fig. ^4j). Furthermore, TDP43^5FL, an RNA-binding-deficient mutant with reduced recruitment to SGs¹⁵, displayed increased co-aggregation with SEC16A when compared with wild-type (WT) TDP43 (Supplementary Fig. ^{4b, c}). Together, these results suggest that TDP43 co-aggregation with SEC16A is modulated by the capability of SGs to recruit TDP43.

TDP43/SEC16A coaggs are formed by TDP43 coalescence with ERES under proteotoxic stress

Whereas TDP43/G3BP1 and TDP43/NUP88 coaggs had been reported previously^15,19, TDP43/SEC16A coaggs represents a new class of TDP43-containing aggregates, which we further investigated. Immunofluorescence staining revealed the co-aggregation of endogenous untagged TDP43 and SEC16A under heat stress or combined with MG132 (Supplementary Fig. ^{3d, e}). TDP43/SEC16A coaggs could also be observed in different cell types including HEK293T (Supplementary Fig. ^3f) and motor neurons (elaborated further below). Apart from heating, TDP43/SEC16A coaggs could be induced by other proteotoxic stresses such as treatment with MG132 or the autophagy inhibitor Bafilomycin A1 (Baf) (Fig. 3a, b). Acute treatment by a sublethal concentration (0.5mM) of sodium arsenite (As) for 1hr induced TDP43-containing SGs as reported previously (Supplementary Fig. ^5a)^15,63,64, although TDP43 level in As-induced SGs was lower than in heat-induced SGs (Supplementary Fig. ^{5a, b}, Fig. 2a). Moreover, the acute arsenite treatment did not induce TDP43/SEC16A coaggs (Fig. 3a, b). Recently, it was reported that in certain types of cells treated with elevated concentration of sodium chloride (NaCl) for 4hr, ERES become enlarged dynamic (liquid-like) condensates, termed Sec bodies (SBs)^65,66, but whether SBs contain TDP43 was unknown. We thus treated RPE-1 cells with the same stress, but it did not induce TDP43/SEC16A co-aggregation (Fig. 3a, b), indicating that TDP43/SEC16A coaggs were distinct from SBs.

Open in a separate window

Fig. 3

Proteotoxic stress induces TDP43 aggregation with preformed ERES.

a Representative images showing TDP43/SEC16A coaggs in cells treated with 10μM MG132 or 100nM Bafilomycin A1 (Baf) for 12-16hr. In comparison, untreated cells, cells treated with 0.5mM sodium arsenite (As) for 1hr, or cells treated with 200mM NaCl for 4hr were shown. Cyan box: TDP43-mNG channel; magenta box: JF646-Halo-SEC16A channel. In “Zoom Out”, arrowheads indicate coaggs, yellow boxes indicate regions shown in zoom-in views, and green dashes demarcate the nuclei. Scale bars except in “Zoom Out” represent 1 μm, and represent 5 μm in “Zoom Out”. b Quantification of the experiments as in (a) and Fig. 2b showing the volume percentage (% [v/v]) of SEC16A inclusions that enrich TDP43 under the indicated conditions. n=45 cells examined over 3 independent experiments. The values of individual cells are plotted as dots, and the mean values of cells in the same experiment as horizontal segments, the median of which is elongated and thickened. Different colors denote different experiments. Asterisks and “ns” respectively stand for significant (unadjusted p-value≤0.05) and non-significant (unadjusted p-value>0.05) in two-sided t-tests between the experimental means of untreated and each other condition. c Selected frames from a timelapse recording of ERES in a cell shifted to 42°C. Imaging started at 60min after the temperature shift. Cyan: TDP43-mNG; Magenta: JF646-Halo-SEC16A. The scale bar represents 1 μm. A representative timelapse of 3 independent experiments is shown. d Quantification of the experiment in (c) showing the relative intensities of SEC16A and TDP43 over time in ERES (normalized to their respective initial values). e Representative images of SEC24C or SEC31A immunostaining in cells with TDP43/SEC16A coaggs. Arrowheads: coaggs; triangles: SEC16A inclusions without TDP43. The scale bar represents 1 μm. f, g Quantification of the experiments as in (e) showing the percentage of TDP43/SEC16A coaggs that contained SEC24C (f) or SEC31A (g), compared with that of SEC16A inclusions without TDP43 in the same cells or in cells without TDP43/SEC16A coaggs. n=60 cells examined over 3 independent experiments. The values of individual cells are plotted as dots, and the mean values of cells in the same experiment as horizontal segments, the median of which is elongated and thickened. Different colors denote different experiments. Asterisks stand for significant (p-value≤0.05) in two-sided t-tests. h Quantification of the experiments as in (e) showing that the volume percentage of mature ERES (defined as SEC16A/SEC31A inclusions) that contained TDP43 in cells with TDP43-ERES. n=60 cells examined over 3 independent experiments. The values of individual cells are plotted as dots, and the mean values of cells in the same experiment as horizontal segments, the median of which is elongated and thickened. Different colors denote different experiments. i Representative images of the ER-localization of TDP43/SEC16A coaggs (indicated by arrowheads) and SEC16A inclusions without TDP43 (indicated by triangles). The scale bar represents 1 μm. The timelapse recording can be found in Supplementary Movie ³. j Quantification of the experiments as in (i) and Supplementary Movie ³ showing the percentage of frames in which all TDP43/SEC16A coaggs or all SEC16A inclusions without TDP43 were associated with the ER. As a control, the percentage of ER-residing frames of simulated, randomly distributed particles of the same number and sizes as TDP43/SEC16A coaggs is shown. n=45 cells examined over 3 independent experiments. The values of individual cells are plotted as dots, and the mean values of cells in the same experiment as horizontal segments, the median of which is elongated and thickened. Different colors denote different experiments. The asterisk and “ns” respectively stand for significant (p-value≤0.05) and non-significant (p-value>0.05) in two-sided t-tests.

To observe how TDP43/SEC16A coaggs initiate under proteotoxic stress, we performed timelapse imaging of cells subjected to heat stress at 42°C. Rather than co-assembly with soluble SEC16A to produce coaggs de novo, TDP43 gradually aggregated in pre-existing SEC16A inclusions (Supplementary Fig. ^5c, Fig. 3c, d, Supplementary Movie ²). These SEC16A inclusions often became enlarged or clumped as TDP43 accumulated (Fig. 3c, Supplementary Fig. ^5d) and this can be observed under various stress conditions (Fig. 3a). The average size of TDP43/SEC16A coaggs, as estimated by structured illumination microscopy (SIM), was ~0.25 μm³, substantially larger than that of SEC16A inclusions without TDP43 (~0.06 μm³) but smaller than SGs containing TDP43 (~2.5 μm³) (Supplementary Fig. ^5e). Nonetheless, ~20% TDP43/SEC16A coaggs had similar size as SEC16A inclusions without TDP43, indicating that size alone was insufficient to distinguish between them (Supplementary Fig. ^5f).

We next examined whether TDP43/SEC16A coaggs contained COPII protein complex, which consists of SEC23/SEC24 subcomplex as the inner layer and SEC13/SEC31 as the outer layer, and are recruited to SEC16A inclusions to form mature ERES (Supplementary Fig. ^5c)^55,67,68. Notably, not all SEC16A inclusions colocalized with SEC24C or SEC31, but almost all TDP43/SEC16A coaggs contained both COPII proteins and the percent COPII recruitment was much higher than that of SEC16A inclusions without TDP43 (Fig. 3e–g), indicating that TDP43/SEC16A coaggs originated from mature ERES. In cells that contained TDP43/SEC16A coaggs, the volume percentage [v/v] of (mature) ERES (defined as SEC16A/SEC31A foci) that enriched TDP43 was ~30% on average (Fig. 3h). Interestingly, in these cells, SEC16A inclusions without TDP43 exhibited further reduced levels of COPII recruitment compared to those in cells free of TDP43/SEC16A coaggs (Fig. 3f, g), suggesting that TDP43/SEC16A coaggs sequestrate COPII subunits away from other SEC16A inclusions. Further supporting that TDP43/SEC16A coaggs resulted from TDP43 aggregation with ERES, timelapse imaging revealed that TDP43/SEC16A coaggs remained associated with the ER like other SEC16A inclusions (Fig. 3i, j, Supplementary Movie ³). Therefore, we hereafter refer to TDP43/SEC16A coaggs as “TDP43-ERES”, and ERES without TDP43 aggregation as “ordinary ERES”.

TDP43 proteins that accumulate in ALS-affected neurons are often subject to phosphorylation (commonly at Ser409 and Ser 410)^24,69,70, although it has also been suggested that hyperphosphorylation suppresses aggregation⁷¹. It remains unclear which forms of TDP43 cause cellular dysfunction and death and which forms are a result of cellular malfunction^72,73. We examined whether TDP43 in co-aggregation with ERES was phosphorylated at Ser409 by using an antibody specific for this species (TDP43^pSer409) and found that TDP43^pSer409 was not enriched in TDP43-ERES but rather was in SGs (Supplementary Fig. ^{5g, h}). In addition, ubiquitinated TDP43 has served as a clinical marker for ALS, although the non-ubiquitinated isoform was also shown to be enriched in aggregates^10,24. By immunostaining against ubiquitin (Ub), we found that TDP43-ERES were not enriched for ubiquitin, regardless of whether MG132 was applied to prevent degradation of ubiquitinated proteins (Supplementary Fig. ^{5i, j}). By contrast, ubiquitin was observed to enrich in SGs and the level increased dramatically after MG132 treatment (Supplementary Fig. ^{5i, j}).

Because TDP43 is an RNA-binding protein but TDP43-ERES did not contain polyA RNA (Supplementary Fig. ^4a), we suspected that newly synthesized or unfolded TDP43 were enriched in ERES. To test this, we employed a system of Halo-tagged TDP43 and different HaloTag ligand dyes to differentiate between pre-existing (mature) versus newly synthesized (new) TDP43 (Supplementary Fig. ^6a). To facilitate this analysis, we used MG132 to inhibit the proteasomal degradation of soluble TDP43⁷⁴. Although both the mature and new pools of TDP43 aggregated in ERES (Supplementary Fig. ^6b, Supplementary Movie ⁴), the ratio between the new and mature TDP43 in ERES, normalized to the new/mature ratio of TDP43 in the whole cell, was always greater than one (Supplementary Fig. ^{6c, d}), indicating that newly synthesized rather than mature TDP43 was the favored substrate for aggregation in ERES. Such selectivity for nascent TDP43 was not attributed to different dye properties, as the preference was not abolished by switching the use of dyes (Supplementary Fig. ^6e). Interestingly, the normalized new/mature ratio gradually decreased during the lifetime of TDP43-ERES (Supplementary Fig. ^{6c, d}), suggesting that nascent TDP43 formed the seed for aggregation, to which pre-existing TDP43 was added during aggregate growth. In support of this possibility, cycloheximide (CHX), an inhibitor of protein synthesis, strongly suppressed TDP43 aggregation in ERES (Supplementary Fig. ^{6f, g}). In ALS-affected neurons, TDP43 aggregates contain not only full length TDP43 but also truncated variants, such as 35kDa and 25 kDa C-terminal (C’)-fragments^75,76. Notably, Halo-TDP43 used in pulse-chase experiment was tagged at the N terminus (N’) whereas TDP43-mNG used in previous experiments was tagged at the C terminus, but both formed TDP43-ERES, suggesting that co-aggregation with ERES did not require N’- or C’-truncation.

TDP43-ERES are non-dynamic (solid-like) aggregates distinct from SGs or ordinary ERES

In contrast to their highly dynamic nature in SGs or ERES without TDP43^15,54, molecules of both TDP43 and SEC16A are immobile in TDP43-ERES. In fluorescence recovery after photobleaching (FRAP) experiments, FP-tagged TDP43 and SEC16A in TDP43-ERES were partially photobleached, but the bleached and unbleached regions hardly mixed over time (Fig. 4a, b, Supplementary Movie ⁵). Quantification of the FRAP data yielded the mobile fractions of TDP43 and SEC16A molecules in coaggs to be roughly 15% and 10%, respectively (Fig. 4c). By contrast, TDP43 and G3BP1 in SGs rapidly and fully mixed after partial photobleaching (Supplementary Fig. ^{7a, b}, Supplementary Movie ⁶) with mobile fractions of each protein to be above 60% (Fig. 4c).

Open in a separate window

Fig. 4

TDP43-ERES are non-dynamic (solid-like) aggregates.

a Selected frames from a representative timelapse recording of a TDP43-ERES coagg after partial photobleaching. Insets show individual channels of the yellow boxed region. Each scale bar represents 1 μm. b Quantification of the experiment in (A) showing the relative intensities of TDP43 (cyan lines) and SEC16A (magenta lines) in bleached (solid lines) and unbleached regions (dashed lines) over time. c Quantification of partial-FRAP experiments as in (a) and Supplementary Fig. ^7a comparing the mobile fractions of proteins in TDP43-ERES versus SGs. n=15 aggregates examined over 3 independent experiments. The values of individual aggregates are plotted as dots, and the mean values of aggregates in the same experiment as horizontal segments, the median of which is elongated and thickened. Different colors denote different experiments. Asterisks and “ns” respectively stand for significant (p-value≤0.05) and non-significant (p-value>0.05) in two-sided t-tests. d Selected frames from a representative timelapse recording of a TDP43-ERES after photobleaching in its entirety. The yellow box marks the position of the coagg immediately before photobleaching. The scale bar represents 1 μm. e Quantification of the experiment in (d) showing the relative intensities of TDP43 (cyan line) and SEC16A (magenta line) over time. f Quantification of full-FRAP experiments as in (d) and Supplementary Fig. ^7c comparing the exchangeable fraction of proteins in TDP43-ERES versus SGs. n=15 aggregates examined over 3 independent experiments. The values of individual aggregates are plotted as dots, and the mean values of aggregates in the same experiment as horizontal segments, the median of which is elongated and thickened. Different colors denote different experiments. Asterisks and “ns” respectively stand for significant (p-value≤0.05) and non-significant (p-value>0.05) in two-sided t-tests. g Representative images showing TDP43-ERES (indicated by arrowheads) and SEC16A inclusions without TDP43 (indicated by triangles) in a cell before and after 3.5% [v/v] 1,6-hexanediol (Hex) treatment for 15min. Insets show individual TDP43-mCherry (cyan) and mEGFP-SEC16A (magenta) channels. The scale bar represents 1 μm. h Quantification of the experiments as in (g) comparing the after/before ratios of TDP43-ERES and SEC16A inclusions without TDP43. n=50 cells examined over 5 independent experiments. The median values of cells in the same experiments are plotted as dots. The asterisk stands for significant (p-value≤0.05) in two-sided t-test.

The non-dynamic (solid-like) nature of TDP43-ERES was also reflected by a lack of exchange with the cytosol: when co-aggregated TDP43 and SEC16A were photobleached entirely, the fluorescence recovery for either protein was below 20% over 2.5min (Fig. 4d–f, Supplementary Movie ⁷). On the other hand, TDP43 and G3BP1 in SGs rapidly exchanged with the cytosol (Supplementary Fig. ^7c–e, Supplementary Movie ⁸) with more than 55% exchangeable with the cytosol (Fig. 4f).

Lastly, TDP43-ERES were resistant to 1,6-hexanediol (Hex) (Fig. 4g, h), which disrupts weakly aggregated macromolecules⁷⁷. In fact, Hex treatment led to a~30% increase in the number of TDP43-ERES due to the formation of new TDP43/SEC16A coaggs in some cells (Fig. 4h, Supplementary Movie ⁹). By contrast, after Hex treatment for 15min, ~50% of SEC16A inclusions without TDP43 regardless of size and nearly all SGs dissolved (Fig. 4g, h, Supplementary Fig. ^7f–h), consistent with previous reports^65,66,78. Together, the above findings demonstrated that TDP43-ERES were non-dynamic (solid-like) aggregates distinct from SGs and ordinary ERES.

TDP43 aggregation with ERES impairs ER-to-Golgi transport

As ERES is critical for vesicular transport from ER to Golgi, we examined whether TDP43-ERES aggregation compromises its function. Retention Using Selective Hooks (RUSH) is a method in which secretory substrates, such as ManII or TNFα, fused with streptavidin-binding protein (SBP), can be artificially retained in the ER by interaction with the streptavidin anchor. Addition of biotin then releases the retained substrates to enable quantitative analysis of protein trafficking (Supplementary Fig. ^8a)⁷⁹. We introduced the RUSH system into cells expressing TDP43-mNG and Halo-SEC16A and assayed ER-to-Golgi transport of ManII-SBP-mCherry (RUSH-ManII) under heat stress. As expected, in the absence of biotin, RUSH-ManII was retained in the ER and not delivered to Golgi, regardless of whether TDP43 aggregated in ERES (Supplementary Fig. ^8b). After incubation with biotin for 1hr, RUSH-ManII appeared in Golgi in most cells without TDP43-ERES and was no longer seen in the ER (Fig. 5a, b, Supplementary Fig. ^7c). By contrast, in cells with extensive TDP43-ERES aggregation (> 10% SEC16A inclusions with TDP43), a considerable fraction of RUSH-ManII remained in the ER and outside Golgi (Fig. 5a, b). Furthermore, overexpression of TDP43-mNG but not mNG alone significantly increased the fraction of cells that displayed ER export defect (Fig. 5c), supporting that accumulated TDP43 impaired ER-to-Golgi transport.

Open in a separate window

Fig. 5

TDP43-ERES aggregation impairs ER-to-Golgi transport.

a Representative images of RUSH assays in cells without or with TDP43 co-aggregation with ERES. After biotin addition for 1hr, the subcellular localization of ManII-SBP-mCherry (RUSH-ManII) was imaged with GM130 (Golgi marker) or calreticulin (CRT – ER marker) visualized by using immunofluorescence staining. The inset shows a contrast-adjusted view; arrowheads indicate TDP43-ERES; arrows point to fragmented Golgi; triangles point to RUSH-ManII retention in the ER; green dashes demarcate the nucleus. The 3D projections (view from top) of these cells are displayed. Each scale bar represents 1 μm. b Quantification of the experiments as in (a) and Supplementary Fig. ^8c showing the ratios between RUSH-ManII intensity outside and inside Golgi. Cells were categorized based on their volume percentage of TDP43-ERES (none: 0%; mild: 0-10%; severe: ≥ 10%). n=45 cells examined over 3 independent experiments. The values of individual cells are plotted as dots, and the mean values of cells in the same experiment as horizontal segments, the median of which is elongated and thickened. Different colors denote different experiments. The asterisk and “ns” respectively stand for significant (p-value≤0.05) and non-significant (p-value>0.05) in two-sided t-tests. c Quantification of cells transfected with RUSH-ManII and Halo-SEC16A only or additionally with mNG or TDP43-mNG showing the outside/inside Golgi ratios of RUSH-ManII intensity. n=45 cells examined over 3 independent experiments. The values of individual cells are plotted as dots, and the mean values of cells in the same experiment as horizontal segments, the median of which is elongated and thickened. Different colors denote different experiments. The asterisk and “ns” respectively stand for significant (p-value≤0.05) and non-significant (p-value>0.05) in two-sided t-tests. d Representative images of TDP43-ERES trapping RUSH-ManII after biotin addition for 1hr. A representative slice of the z-stack is displayed. The scale bar represents 1 μm. e Selected frames of a representative timelapse recording of an ordinary ERES (in a cell without TDP43-ERES) during RUSH-TNFα transport from ER to Golgi. Biotin was added at 0”. Arrows indicate the budding of a tubular transport intermediate. The scale bar represents 1 μm. f Quantification of relative RUSH-TNFα intensity over time (normalized to the max intensity), compared to the relative intensity of SEC16A in the ERES tracked in (e). CC: cross-correlation between RUSH-TNFα intensity and SEC16A intensity. g Selected frames of a representative timelapse recording of a TDP43-ERES during RUSH-TNFα assay. The scale bar represents 1 μm. h Quantification of RUSH-TNFα and SEC16A intensities over time (normalized to their respective max intensities) in the TDP43-ERES tracked in (g). i Quantification of the percentage of TDP43-ERES and ERES without TDP43 in either the same or different cells that retained RUSH-TNFα in the experiments as in (e–h) and Supplementary Fig. ^{9d, e}. The criterion for cargo retention is CC≥0.5. n=9 cells examined over 3 independent experiments. Cells without biotin supplement were included as a control. n=6 cells examined over 3 independent experiments. Each dot represents one cell, and different colors denote different experiments. The asterisk and “ns” respectively stand for significant (p-value≤0.05) and non-significant (p-value>0.05) in two-sided t-tests.

Notably, RUSH-ManII was retained in TDP43-ERES after biotin addition for 1hr (Fig. 5d), suggesting that TDP43 aggregation caused ERES dysfunction to trap secretory cargos recruited to these abnormal sites. To directly test this hypothesis, we took advantage of another RUSH reporter, TNFα-SBP-mCherry (RUSH-TNFα), which is pre-enriched in ERES prior to biotin supplement⁵⁶. After adding biotin, ordinary ERES that already accumulated RUSH-TNFα efficiently exported this substrate via characteristic tubular transport intermediates/vessels (Fig. 5e), consistent with previous observations⁵⁶. As such, the intensity of RUSH-TNFα in ordinary ERES decreased dramatically after cargo release while SEC16A intensity remained relatively stable (Fig. 5f, Supplementary Movie ¹⁰). By contrast, TDP43-ERES, although similarly recruited RUSH-TNFα (Supplementary Fig. ^{9a, b}), failed to form tubular transport intermediates to release this cargo, such that the intensity of RUSH-TNFα remained stable and/or correlated with SEC16A intensity over time - the cross-correlation (CC) between RUSH-TNFα and SEC16A intensities was close to 1 (Fig. 5g, h, Supplementary Movie ¹¹). We hence calculated the CC between RUSH-TNFα and SEC16A intensities for TDP43-ERES and ordinary ERES in our timelapse recordings and defined CC≥0.5 as the benchmark for cargo retention. Based on this criterion, the percentage of TDP43-ERES that retained RUSH-TNFα was significantly higher than that of ordinary ERES (Fig. 5i). Surprisingly, in cells with TDP43-ERES, a lower percentage of SEC16A inclusions without TDP43 enriched RUSH-TNFα (Supplementary Fig. ^{9a, b}) and were more likely to fail in cargo release (Supplementary Fig. ^{9d, e}, Fig. 5i, Supplementary Movie ¹²), suggesting that TDP43-ERES had dominant effects over ordinary ERES in the same cells.

Artificially induced TDP43 aggregation at ERES inhibits ER-to-Golgi transport

To further investigate whether TDP43-ERES aggregation (but not cytosolic TDP43, overexpressed SEC16A, or any pleiotropic effects of the heat stress) was the cause of ERES dysfunction, we employed a method to directly aggregate TDP43 in ERES. When TDP43 and SEC16A were respectively fused with three copies of FKBP and FKBP-rapamycin binding domain bearing T2098L mutation (FRB*), the rapamycin analog AP21967 (AP), which dimerizes FKBP and FRB*, directly induced TDP43/SEC16A co-aggregation (Supplementary Fig. ^9f). As a control, cells expressing TDP43 and SEC16A without FKBP or FRB* tagging were not responsive to AP (Supplementary Fig. ^9g, Fig. 6a, b). Furthermore, like the TDP43-ERES induced by proteotoxic stresses, AP-induced TDP43/SEC16A coaggs contained COPII proteins (Fig. 6c), and 3xFKBP-TDP43-mNG and 3xFRB*-Halo-SEC16A molecules inside were immobile as revealed by FRAP (Fig. 6d).

Open in a separate window

Fig. 6

Artificially induced TDP43-ERES retain secretory cargo.

a Representative images of cells cultured in biotin-free media expressing 3xFKBP-TDP43-mNG, 3xFRB*-Halo-SEC16A and RUSH-TNFα without or with AP21967 (AP) treatment for 6-10hr. Arrowheads and triangles indicate TDP43-ERES and ordinary ERES, respectively. In “Zoom out”, yellow boxes mark the regions displayed in zoom-in views, and cyan dashes demarcate the nuclei. Scale bars except in “Zoom Out” represent 1 μm, and represent 5 μm in “Zoom Out”. b Quantification of the experiments as in (a) and Supplementary Fig. ⁹g showing the volume percentage of SEC16A inclusions that enriched TDP43 under the indicated conditions. n=45 cells examined over 3 independent experiments. The values of individual cells are plotted as dots, and the mean values of cells in the same experiment as horizontal segments, the median of which is elongated and thickened. Different colors denote different experiments. The asterisk and “ns” respectively stand for significant (p-value≤0.05) and non-significant (p-value>0.05) in two-sided t-tests. c SEC24C or SEC31A immunostaining in cells with AP-induced TDP43/SEC16A coaggs. Scale bars except in “Zoom Out” represent 1 μm, and represent 5 μm in “Zoom Out”. Representative images of 3 independent experiments are shown. d Selected frames from a timelapse recording of an AP-induced TDP43-ERES after photobleaching. Insets show individual channels. 3xFKBP-TDP43-mNG was entirely photobleached whereas 3xFRB*-(JF646)-Halo-SEC16A was partially bleached. The scale bar represents 0.5 μm. A representative timelapse of 3 independent experiments are shown. e Selected frames of a representative timelapse recording of an AP-induced TDP43-ERES during RUSH-TNFα assay. The scale bar represents 1 μm. f Quantification of RUSH-TNFα and SEC16A intensities over time (normalized to their respective max intensities) in the TDP43-ERES tracked in (e). g Quantification of the percentage of AP-induced TDP43(-mNG)-ERES, mNG-ERES and ERES without TDP43 (in cells without AP treatment) that retained RUSH-TNFα in the experiments as in (e) and Supplementary Fig. ^9h–k. n=9 cells examined over 3 independent experiments. Each dot represents one cell, and different colors denote different experiments. The asterisk and “ns” respectively stand for significant (p-value≤0.05) and non-significant (p-value>0.05) in two-sided t-tests. h Working model of TDP43 aggregation at ERES causing retention of transport cargos. TDP43-ERES also exert a dominant effect over ordinary ERES in the same cells possibly through sequestration of essential transport factors. The ER-to-Golgi transport defect caused by TDP43-ERES co-aggregation may further induce ER stress.

Next, we examined whether AP-induced TDP43 aggregation at ERES affected cargo export from the ER. The AP-induced TDP43-ERES still enriched RUSH-TNFα (Fig. 6a), but this cargo was not exported after biotin addition (Fig. 6e, f, Supplementary Movie ¹³). By contrast, in the absence of AP, cargo release from ERES was unaffected, suggesting that FKBP or FRB tagging alone did not compromise ERES function (Supplementary Fig. ^{9h, i}, Fig. 6g). As an additional control, AP-induced recruitment of 3xFKBP-mNG to ERES but did not cause ERES to retain RUSH-TNFα (Supplementary Fig. ^{9j, k}, Fig. 6g). These results suggest that TDP43 aggregation in ERES is sufficient to impair ER-to-Golgi transport (Fig. 6h).

Mammalian cell culture and plasmid transfection

Cell culture was performed according to the American Type Culture Collection (ATCC) guidelines with minor modifications. Briefly, RPE-1 cells were maintained in DMEM:F12 (Thermo 11320082)+10% [v/v] FBS (Thermo 16140071) and passaged using 0.05% [w/v] trypsin solution (Thermo 25300120). HEK293T cells were cultured in DMEM (Thermo 11965118)+10% [v/v] FBS and passaged similarly. For high-resolution microscope imaging, cells were seeded on glass-bottom dishes (Iwaki 3930-035 or 3931-035) pre-coated with 10μg/mL fibronectin (Sigma 10838039001) overnight at 4°C. Plasmid transfections were performed using Lipofectamine 3000 (Thermo L3000015) according to the manufacturer’s protocol. The transfection amounts of different plasmids were optimized (Supplementary Data ²). Plasmids generated in this study and their maps will be submitted to Addgene and also available on request. Cells were analyzed by microscopy imaging, Western Blot (WB) or other means 24–48hr after plasmid transfection.

Aggregate induction and inhibitor/chemical treatments

To examine protein aggregation under different conditions, cells were either untreated, mock treated or heat stressed at 42°C, or treated with different chemicals for 12–16hr. MG132 (Sigma M7449), inhibitor of aspartyl proteases and cysteine proteases that are also found in proteasomes, was used at 10 μΜ final concentration. The autophagy inhibitor Bafilomycin A1 (Baf) (Sigma SML1661) was used at 100nM. Sodium arsenite (Santa Cruz SC-250986) was used at 0.5mM for 1hr. NaCl (Axil BIO-1111) was used at 200mM for 4hr to replicate the condition used for Sec body (SB) induction⁶⁶. The protein translation inhibitor cycloheximide (CHX) (Sigma C1988) was used at 20μg/mL for 6hr. 1,6-Hexanediol (Hex) (Sigma 88571) was used at 3.5% [v/v] and cells were imaged immediately by timelapse recording.

HaloTag (Halo) and Immunofluorescence (IF) staining

Halo staining was performed according to the dye manufacturer’s protocol with minor modifications. Briefly, Janelia Fluor 549 (JF549) or JF646 ligand (Promega GA1110 or GA1120) was added to media to a final concentration of 15–50nM and incubated for 15-30min or indicated time. Then cells were washed once using fresh media (unless indicated otherwise), and either imaged directly or fixed using 4% [w/v] paraformaldehyde (PFA) (EMS 15710) diluted in phosphate buffered saline (PBS) (Thermo 14190250) for 20min at room temperature (RT). For IF staining of proteins except from calreticulin (CRT), live cells were washed once using fresh media and fixed using 4% [w/v] PFA. After three PBS washes, cells were permeabilized in PBS+0.1% [v/v] Triton X-100 (Thermo 85111) for 10min and blocked in PBS+0.1% [v/v] Tween 20 (Sigma P1379) (PBST)+1% [w/v] BSA (Sigma A2153). Cells were incubated with primary antibodies overnight at 4°C in PBST+1% [w/v] BSA. Then after four washes with PBS, cells were incubated with fluorophore-conjugated secondary antibodies in the same diluent for 1hr at RT. Again, cells were washed four times using PBS before microscopy imaging. For IF staining of CRT, the following modifications were made: cells were instead fixed using 2% [w/v] PFA+1.5% [w/v] glutaraldehyde (GA, Sigma 5882) at 37°C; after PBS washes to remove fixatives, cells were incubated for 10min at RT with 0.1% [w/v] sodium borohydride (NaBH₄, Sigma 452882) in PBS to reduce GA; primary antibody incubation was reduced to 2hr at 4°C. Antibodies used in this study are listed in Supplementary Data ².

RNA fluorescence in situ hybridization (FISH)

RNA FISH was performed according to a published protocol⁵⁹ with minor modifications. Briefly, cells were fixed and permeabilized as in IF, and then washed thrice in PBS and twice in SSC wash buffer (2x SSC buffer + 15% [v/v] formamide), in which the 2x SSC buffer (300mM NaCl, 30mM sodium citrate, pH = 7.0) was diluted from 20x SSC (Sigma S6639) using UltraPure DNase/RNase-Free Distilled Water (Thermo 10977015). Then Alexa Fluor 647 (AF647)-conjugated oligo(dT)₃₀ purchased form IDT was diluted to a final concentration of 1ng/μL in FISH buffer, which consists of SSC wash buffer + 10% [w/v] dextran sulfate (Sigma D8906)+2mM vanadyl-ribonucleoside complex (NEB S1402)+0.02% [w/v] BSA, + 1mg/mL tRNA (Sigma 10109541001), and used to stain cells at 37°C overnight. After incubation, cells were washed twice in SSC wash buffer and twice in PBS, re-fixed in 4% [w/v] PFA for 10min at RT, and washed thrice in PBS before imaging.

Microscopy imaging

Wide field images were taken on a EVOS M5000 microscope using a 40x air objective. Confocal images were taken on Yokogawa CSU-W1 spinning disk microscope (installed on Nikon Eclipse Ti-E inverted microscope body) with a CFI Apochromat λ 100x oil objective (NA=1.4), Cargille Type 37 immersion oil (Cargille Labs 16237-16) and Photometrics Prime 95B Scientific CMOS cameras. Z-stacks were taken using a Piezo stage (Piano Z from Physik Instrumente) every 0.1 μm for a total thickness of 5 or 10 μm except for Supplementary Movie ⁹, in which the step size was 0.5 μm. To enable Structured Illumination Microscopy (SIM), the Live-SR super-resolution module (Roper Scientific) was engaged except for Figs. 4, ​,5e,5e, g, 6d, e, Supplementary Fig. ^{7a, b}, Supplementary Fig. ^{9a–d, h, i}. For live cell imaging, the environment chamber was stabilized to 37°C and CO₂ was provided by LCI FC-5N CO₂ mixer (Live Cell Instrument) and calibrated to maintain normal pH in cell culture media. For imaging heat stressed cells, the on-stage incubator (Live Cell Instrument CU-501) was heated to and stabilized at 42°C. To avoid drifting, cells were pre-treated in a 42°C CO₂ incubator for 15min before they were transferred to the microscope and allowed 30min to further stabilize before imaging starts.

Correction of chromatic aberration, image analysis and statistics

Chromatic aberration of confocal images along the z direction was calibrated using 0.1 μm Tetraspec Microspheres (Thermo T7279) and corrected in ImageJ/Fiji¹²⁵ using a custom Jython script. 3D reconstruction and analysis of z-stacks were performed in Imaris whereas single z-slice images were analyzed using custom ImageJ Jython scripts. Measurements made in Imaris and ImageJ were analyzed using custom R scripts.

Quantification of co-aggregates

For quantification of the (volume) percentage of SEC16A inclusions with TDP43 aggregation, TDP43/SEC16A coaggs and SEC16A inclusions without TDP43 were respectively identified in Imaris using its Labkit machine learning plugin¹²⁶. For other quantifications, inclusions of two different proteins, e.g. SEC16A and Hsp104^DWB, were respectively identified in Imaris by manual thresholding based on local contrast, and the mutual overlap volumes between pairs of aggregates were measured. A SEC16A inclusion is categorized as in colocalization with Hsp104^DWB if the overlap volume is no less than 80% of the former’s volume.

Fluorescence recovery after photobleaching (FRAP) assay

Aggregates were imaged for 11 frames before photobleaching by high power laser in part or in full. Then the aggregates were imaged continuously for at least 2.5min and quantified in ImageJ using a custom Jython script.

RNA inhibition (RNAi)

TriFECTa RNAi Kits including predesigned dicer-substrate small interference RNAs (DsiRNAs) were purchased from IDT. The sequences of DsiRNAs are listed in Supplementary Data ². Each DsiRNA was co-transfected at a final concentration of 10nM with plasmids using Lipofectamine 3000 according to the manufacturer’s protocol. Briefly, DsiRNAs were immediately added after mixing Lipofectamine 3000 and P3000-bound plasmids.

Generation of in-frame knock-in cell lines by CRISPR/Cas

mScarlet-I-G3BP1 in-frame tagging was conducted using Alt-R CRISPR/Cas system (IDT) according to the vendor’s protocol with minor modifications. Briefly, a crRNA targeting the N-terminus of G3BP1 (synthesized at IDT, sequence available in Supplementary Data ²) was annealed with an equimolar amount of tracrRNA (IDT 1072534) to form the guide RNA (gRNA) duplex. Then the gRNA was mixed with SpCas9 (IDT 1081059) to form the gRNA/Cas9 complex, which in turn was co-transfected with a mScarlet-I donor (synthesized at IDT, sequence available in Supplementary Data ²) into RPE-1 by the Neon Electroporation system (Thermo MPK1025). Cells expressing mScarlet-I-G3BP1 were sorted twice by fluorescence-activated cell sorting (FACS) on a Sony SH800S. PCR genotyping and WB showed that cells were heterozygous for mScarlet-I-G3BP1 tagging.

Western blot (WB)

Cells cultured in each 35mm dish were dissolved in 200-300μL RIPA buffer (Thermo 89900)+2mM PMSF (Sigma P7626)+3μL Protease Inhibitor Cocktail (PIC, Sigma P8340) with regular swirling for 10min. The protein concentrations of cell lysates were measured using BCA assay (Thermo 23225) and equal amounts of proteins were loaded onto 4-15 % polyacrylamide gradient gels (Bio-Rad 4561084) and fractionated by SDS-polyacrylamide gel electrophoresis (SDS-PAGE). Then proteins were electroblotted onto nitrocellulose membranes (BIO-RAD 1704158). After blocking in Odyssey Blocking Buffer (PBS, LI-COR 927), membranes were incubated sequentially with primary and secondary antibodies in Odyssey Blocking Buffer mixed with an equal volume of PBS+0.2% [v/v] Tween 20. After each incubation, membranes were washed four times in PBST. Tween 20 was removed by rinsing in PBS before detecting the fluorescence of secondary antibodies using a LI-COR Odyssey CLx Scanner. The fluorescence of protein bands was quantified in ImageStudioLite (Bio-Rad).

Aggregate purification

HEK293T cells cultured in two T225 flasks (Thermo 159934) were respectively transfected with plasmids carrying Hsp104^DWB-mEGFP-FLAG and, as a control, Hsp104^DWB-mEGFP. 24hr after transfection, both flasks were transferred to a 42°C CO₂ incubator for 12hr to exacerbate protein aggregation. Then cells were gently washed once in PBS and harvested by washing off using PBS. The pellets from centrifugation at 120g for 5min was washed again using PBS and frozen in liquid nitrogen for future use. When needed, the frozen pellets were thawed at 4°C, each lysed by incubation in 1000μL RIPA buffer + 2mM PMSF+60μL PIC+112.5 Kunitz Unit/mL DNase I (RNase-free, NEB M0303)+2.5mM MgCl₂ (Sigma 5985-OP) with occasional pipette aspiration, and a brief centrifugation at 150g for 5min was conducted to remove cell debris. A small aliquot of the supernatant was used for BCA assay of protein concentration whereas the rest was loaded onto sucrose gradient layers of 20% and 50% [w/v] sucrose, each dissolved in 1500μL RIPA buffer. Centrifugation at 60,000g for 30min on an ultracentrifuge (Thermo Sorvall WX100) enriched Hsp104^DWB-mEGFP(-FLAG)-labeled aggregates at the interface between 20 and 50% [w/v] sucrose. This layer was extracted through needle and syringe to yield ~750μL of sample, which in turn was mixed with 200μL of pre-equilibrated anti-FLAG M2 resin (Sigma A2220), rotated at 4°C for 1hr, and washed three times with RIPA buffer to immunoprecipitate Hsp104^DWB-mEGFP-FLAG-labeled aggregates. In the end, aggregates were eluted in urea elution buffer (8M urea, 20mM Tris pH=7.4, 100mM NaCl). At each step, the presence of Hsp104^DWB-mEGFP-FLAG-labeled aggregates was checked by microscopy, and aliquots were saved for analysis by SDS-PAGE and Coomassie Brilliant Blue staining or WB.

Liquid chromatography-tandem mass spectrometry (LC-MS/MS)

Proteins were reduced by 10mM TCEP for 20min at RT and alkylated with 55mM 2-chloroacetamide (CAA) in the dark for 30min. Then, the proteins were digested with 2μg endoproteinase LysC for 3hr followed by 2μg trypsin at 37°C overnight, and the digestion was terminated by adding trifluoroacetic acid (TFA) to a final concentration of 1% [v/v]. Subsequently, the peptides were desalted using C18 HLB cartridges (Waters), dried by centrifugal evaporation, and resuspended in 25μl TEAB, pH = 8.5. The samples were differentially labeled with TMT10-plex isobaric tandem mass tags (Thermo) at 25°C overnight, and the reaction was quenched by addition of 30μl of 1M ammonium formate, pH = 10 before pooling all samples into a low-binding microfuge tube. The pooled samples were desalted and fractionated on a self-packed spin column filled with C18 beads (Dr Maisch) in the step gradient of 14%, 24% and 60% [v/v] acetonitrile dissolved in 10mM ammonium formate, pH=10. Each fraction was dried by centrifugal evaporation and further washed and dried twice by addition of 60% [v/v] acetonitrile + 0.1% [v/v] formic acid to remove residual ammonium formate. Then the samples were resuspended in 10μl of 2% [v/v] acetonitrile + 0.06% [v/v] TFA+0.5% [v/v] acetic acid and transferred to an autosampler plate. Online chromatography was performed using an EASY-nLC 1000 liquid chromatography system (Thermo) with a single-column setup and 0.1% [v/v] formic acid and 0.1% [v/v] formic acid + 99% [v/v] acetonitrile as mobile phases. The fractionated samples were injected and separated on an Easy-Spray reversed-phase C18 analytical column (Thermo) with 75 μm inner diameter, 50cm length and 2 μm particle size. The column was maintained at 50°C in 2–33% [v/v] acetonitrile gradient for over 55min, followed by an increase to 45% [v/v] over the next 5min, and then to 95% [v/v] over 5min. The final mixture was maintained on the column for 4min to elute all remaining peptides. The total run duration for each sample was 70min at a constant flow rate of 300 nL/min. MS data were acquired on a Q Exactive HFX mass spectrometer (Thermo) using data-dependent mode. Samples were ionized under 2.1kV and 300°C at the nanospray source. Positively charged precursor signals (i.e. MS1) were detected using an Orbitrap analyzer set to a resolution of 60,000, automatic gain control (AGC) target of 3,000,000ions, and max injection time (IT) of 50ms. Precursors with charges 2–7 of the highest ion counts in each MS1 scan were further fragmented using higher-energy collision dissociation (HCD) at 36% normalized collision energy. The fragment signals (i.e. MS2) were analyzed by the Orbitrap analyzer at a resolution of 7500, AGC of 100,000 and max IT of 100ms. Data-dependent (MS2) scans were acquired in Top40 mode. To avoid re-sampling of high abundance peptides, the precursors used for MS2 scans were excluded for 30sec.

Proteomics data analysis

Proteins were identified using Proteome Discoverer v2.3 (Thermo). Briefly, the raw mass spectra were searched against human primary protein sequences retrieved from Swiss-Prot (June 11 2019 release)¹²⁷ with carbamidomethylation on Cys and TMT10-plex on Lys and N-terminus set as fixed modifications, and deamidation of Asn and Gln, acetylation of N-terminus, and Met oxidation as dynamic modifications. Trypsin/P was set as the digestion enzyme and allowed up to three missed cleavage sites. Precursors and fragments of mass errors within 10 ppm and 0.8Da, respectively, were accepted, while reporter ions were at least 1Da apart from each other. Peptides were matched to the spectra at a false discovery rate (FDR) of 1% (strict) and 5% (relaxed) against the decoy database and quantitated using TMT10-plex method. The search results were further processed using a custom R script, in which Gene Ontology (GO) enrichment was performed using clusterProfiler^128,129. The SG/nucleolus proteome, IDR-containing proteins and LLPS substrates were retrieved from previous publications and databases^130–133.

Retention Using Selective Hooks (RUSH) assay for measuring ER-to-Golgi transport

RUSH was performed as previously described with minor modifications (Supplementary Fig. ^8a)⁷⁹. Briefly, the RPE-1 cell culture media was substituted with custom made biotin-free DMEM:F12 media (Cell Culture Technologies)+10% [v/v] FBS from seeding (1 d before transfection). Then a plasmid carrying both the ER anchor ss-Strep-KDEL and the reporter ManII (aa1-116 only)-SBP-mCherry or TNFα-SBP-mCherry was co-transfected with TDP43 and SEC16A constructs using Lipofectamine 3000. To release cargo retention, D-biotin (Thermo B20656) was added to a final concentration of 40μM. For tracking ERES in timelapse imaging, SEC16A inclusions were identified in ImageJ using ilastik machine learning plugin¹³⁴ and then tracked by Trackmate¹³⁵. The trajectories were analyzed by custom R scripts. Only SEC16A inclusions that enriched transport cargo (RUSH-TNFα) and tracked for no less than 10 frames (~15sec) were used for cross-correlation (CC) analysis.

AP21967-induced TDP43-ERES co-aggregation

3xFKBP-TDP43-mNG and 3xFRB*-Halo-SEC16A were co-transfected with the RUSH-TNFα construct. 1μM rapamycin analog AP21967 (AP, Takara 635056) was added ~18hr post-transfection to treat the cells for 6–10hr to induce co-aggregation.

Differentiation of motor neurons (MNs) from stem cells

The use of stem cells for MN differentiation was approved by the Human Biomedical Research Office, Agency for Science, Technology and Research of Singapore under protocol number 05-02-028. The human embryonic stem cells (hESCs) and induced pluripotent stem cells (hiPSCs) used in this study are: NDS00268, NDS00269 and NDS00270 (ALS patients) and BJ, GM23720 and H9 (non-ALS). BJ was obtained from National Center for Global Health (Japan), GM23720 was obtained from the NIGMS Human Genetic Cell Repository, H9 was purchased from Corning, and NDS00268, NDS00269, NDS00270 were obtained from NINDS Human Cell and Data Repository. The induction of MNs from stem cells were performed according to previous publications^136,137 and illustrated in Supplementary Fig. ^10a. In brief, stem cells were cultured on Matrigel (Corning 354230)-coated plates in StemMACS iPS-Brew XF (Miltenyi Biotec 130-104-368). The cell culture media was changed every other day and cells were passaged at a split ratio between 1: 6 to 1: 10 every 6-8 days. On the day of differentiation (D0), stem cells were digested with Accutase (Nacalai Tesque 12679-54) for 5-10min to acquire single-cell suspensions. Accutase was removed by centrifugation and PBS wash. Then cells were seeded at 70% confluency on dishes pre-coated with Matrigel and cultured in StemMACS iPS-Brew XF+5μM Y-27632 (STEMCELL Technologies 72304). On D1, the media was replaced with Neural Induction Media (NIM) comprising the Neural Media (NM)+0.5μM LDN193189 (STEMCELL Technologies 72147)+4.25μM CHIR99021 (Miltenyi Biotec 130106539), where NM is a mixture of 50% [v/v] DMEM:F12+50% [v/v] MACS Neuro Media (Miltenyi Biotec 130-093-570)+0.5x GlutaMax (Thermo 35050061)+1x NEAA (Thermo 11140050)+1x N-2 (Thermo 17502048)+1x MACS NeuroBrew-21 (Miltenyi Biotec 130-093-566). On D3, 1μM retinoic acid (RA, Sigma R2625) was supplemented to the media, which was then changed every 2-3 days until D7 or when cells became confluent. At this point, stem cells have differentiated into MN progenitors, which can be passaged at a ratio between 1: 2 and 1: 3 onto new Matrigel-coated dishes. These progenitor cells were cultured in NIM+1μM RA+5μM Y-27632 and the media was changed every day. From D11, MN differentiation media, i.e. NM+1μM RA+1 μΜ Purmorphamine (Miltenyi Biotec 130-104-465), was added and changed every other day until D17. Then MN maturation media, i.e. NM+10ng/mL GDNF (Miltenyi Biotec 130-129-547)+10ng/mL BDNF (Miltenyi Biotec 130-093-811)+200 μM L-ascorbic acid (Sigma A5960), was added and changed every other day until MNs were ready on D28. The quality of MNs was checked by morphology and IF staining of (motor) neuron markers including NeuN/RBFOX3, NFM and ChAT.

Immunofluorescence staining of post-mortem ALS patient samples

The post-mortem samples of ALS patients were acquired from the Johns Hopkins/Temple University ALS Postmortem Core in form of formalin-fixed paraffin-embedded (FFPE) tissue sections of approximately 20mm×10mm x 5 μm in dimensions. The demographic data of these patients are shown in Supplementary Data ³. For immunofluorescence staining, the samples were deparaffinized by incubating thrice in fresh Xylene (Sigma 534056) and rehydrated by incubating twice in ethanol and once in 90% and 70% [v/v] ethanol. Each incubation time is 5min. Then the samples were boiled in epitope retrieval solution (IHC World IW-1100) in an autoclave (Hirayama HV-50) using the liquid cycle (121°C 15min), washed twice with PBS, and permeabilized for 10min in PBS+0.4% [v/v] Triton X-100. After three PBS washes, the samples were blocked in serum free protein block (Agilent X090930-2) for 1hr at RT, and incubated with primary antibodies diluted in background reducing antibody diluent (Agilent S302283-2) overnight at RT. After three PBS washes, the samples were incubated with diluted secondary antibodies for 2hr at RT and washed with PBS. The nuclei were stained by incubation with 1μg/mL DAPI (Sigma MBD0015) in PBS for 30min at RT followed by three more PBS washes. The stained sections were preserved in Prolong Gold antifade mountant (Thermo P36930) with coverslips sealed by transparent nail polish (Beauty Language). For image quantification, only cells positive for anti-NeuN immunostaining were counted. The analyzes of these patient samples were performed in a double-blinded manner.

The authors would like to thank J. Shorter (UPenn) for recommending the DWB mutations for stable Hsp104-aggregate binding and generously sharing a Hsp104^DWB construct. The authors thank P. Kanchanawong and K. Paramasivam for kindly sharing FP constructs, and B.C. Low for kindly sharing HEK293T cell line. The authors thank H.T. Ong and J.F.L. Chin from MBI Microscopy core, Singapore Microscopy and Bioimage Analysis (SIMBA), for help in correction of chromatic aberration in confocal imaging, and MBI Wet Lab Core for providing essential lab equipment. The authors are grateful to G. Thibault (NTU), B. Burke (A*STAR), T. Hiraiwa, G. Jedd (TLL), P.T. Matsudaira, J. Hu, M. Yao, W. Huang, L. Ruan (JHU), H. Wang, R. Das, T. Chew, Y. Tee, M. Pan, S. Ramachandran (JHU), B. S. Wong, X. Lan, K. Seah and Y. Lee for constructive discussions. This work was supported by R. Li’s startup grant (A-0007081-00-00), a Tier 3 grant (A-0006324-03-00) received from Singapore Ministry of Education, a Mid-sized Grant (NRF-MSG-2023-0001) from Singapore National Research Foundation, and by the A*STAR core funding and Singapore National Research Foundation (under its NRF-SIS “SingMass” scheme) endowed to R.M. Sobota and L.C. Wang. H. W. would like to specially thank those who have maintained faith in him during difficult times.