TorsinA is essential for neuronal nuclear pore complex localization and maturation

Abstract

As lifelong interphase cells, neurons face an array of unique challenges. A key challenge is regulating nuclear pore complex (NPC) biogenesis and localization, the mechanisms of which are largely unknown. Here we identify neuronal maturation as a period of strongly upregulated NPC biogenesis. We demonstrate that the AAA+ protein torsinA, whose dysfunction causes the neurodevelopmental movement disorder DYT-TOR1A dystonia and co-ordinates NPC spatial organization without impacting total NPC density. We generated an endogenous Nup107-HaloTag mouse line to directly visualize NPC organization in developing neurons and find that torsinA is essential for proper NPC localization. In the absence of torsinA, the inner nuclear membrane buds excessively at sites of mislocalized nascent NPCs, and the formation of complete NPCs is delayed. Our work demonstrates that NPC spatial organization and number are independently determined and identifies NPC biogenesis as a process vulnerable to neurodevelopmental disease insults.

Extended Data Figure 1:. Identification and analysis of nucleoporin puncta

a, SIM images of primary neurons aged DIV4, 6, 8, 10, 14, 18 and 24 labeled with anti-Nup153 and anti-Nup98 antibodies. Peaks column shows peaks identified from Nup153 (green) and Nup98 (magenta) puncta. Scale bar = 2μm. b, Schematic of image analysis pipeline. The flattest part of the nuclear envelope was imaged. Nuclear ROIs were determined by manually outlining each nucleus. ROIs were eroded to account for nuclear envelope curvature at the edge, then overlayed onto Otsu thresholded images. Local maxima were identified with extended-maxima transform and used to establish peaks corresponding to each Nup. c, Superplots of nuclear ROI area as determined by Nup153/Nup98 staining in maturing neurons aged DIV4, 6, 8, 10, 14, 18 and 24. Plots show mean ±SD, with color coding indicative of biological replicates. One-way ANOVA with Dunnett’s multiple comparisons test on 3 biological replicates was performed, with DIV4 as the reference condition. Only DIV14 showed statistical significance; *P=0.0331. d, e, Frequency distribution of nearest neighbor distances of Nup153 (d) and Nup98 (e) puncta. For each Nup153 and Nup98 peak, the distance to its nearest neighbor was calculated. The results from three biological replicates were combined for each timepoint.

Extended Data Figure 2:. Multiple NPC components exhibit density increase during neuronal maturation

a, Confocal images of DIV4 and DIV10 primary neurons labeled with anti-Nup153 and anti-Nup98 antibodies. Scale bar = 10μm. b, c, Normalized nuclear rim fluorescence intensity of Nup153 and Nup98 in DIV4 and DIV10 primary neurons. Plots show mean ± SD, with color coding indicative of biological replicates. *P=0.0196 (b), *P=0.0139 (c) using two-tailed paired t-test from 4 biological replicates. d, Confocal images of DIV4 and DIV10 primary neurons labeled with anti-FG nucleoporin (mab414) and anti-Nup210 antibodies. Scale bar = 10μm. e, f, Normalized nuclear rim fluorescence intensity of FG-Nups (mab414) and Nup210 in DIV4 and DIV10 primary neurons. Plots show mean ± SD, with color coding indicative of biological replicates. **P=0.0076 (e), ****P<0.0001(f), two-tailed paired t-test from 4 biological replicates.

Extended Data Figure 3:. Comparison of WT and torsinA-KO neuronal NPC spatial organization and nuclear morphology

a, Autocorrelation plot of WT and torsinA-KO neurons calculated from Nup98 WT and torsinA-KO SIM images. b, Normalized autocorrelation plot of WT and torsinA-KO neurons calculated from Nup98 WT and torsinA-KO SIM images. Autocorrelations were normalized to account for amplitude dependency on NPC density. c, Autocorrelation plot of DIV10 WT and torsinA-KO neurons calculated from Nup210 dSTORM images. d, Nearest neighbor distance of segmented DIV10 WT and torsinA-KO NPC centroids. e, f, WT and torsinA-KO Nup153 (e) and Nup98 (f) puncta density in maturing primary neurons aged DIV4, 6, 8, 10, 14. Plots show mean ± SD, with color coding indicative of biological replicates. Timepoints from the same biological replicate were matched. Repeated measures two-way ANOVA with Dunnett’s multiple comparisons test was performed on 3 biological replicates, with DIV4 as the reference condition. ns, non-significant; P values are reported on the plots. ****P<0.0001; repeated measures two-way ANOVA with Dunnett’s multiple comparisons test, using DIV4 as a reference, 3 biological replicates. g, Nuclear area of WT and torsinA-KO neurons calculated from manually drawn ROIs from SIM images. Plots show mean ± SD. Repeated measures two-way ANOVA with Sidak’s multiple comparisons test was used to compare genotypes at each timepoint, and no comparisons reached statistical significance. Repeated measures two-way ANOVA with Dunnett’s multiple comparisons test was used to compare all timepoints to DIV4 within each genotype, and only torsinA-KO DIV4 versus DIV14 reached significance (*P=0.0102). h, Nuclear ROI circularity of WT and torsinA-KO neurons calculated from manually drawn ROIs from SIM images. Plots show mean ± SD. Repeated measures two-way ANOVA with Sidak’s multiple comparisons test was used to compare genotypes at each timepoint. No statistical significance was detected. i, Expansion microscopy of DIV10 WT and torsinA-KO neurons labeled with Nup210 and Lamin A/C antibodies. Representative images from 3 biological replicates per genotype are shown. Scale bar = 20μm.

Extended Data Figure 4:. Validation of HaloTag-Nup107 mouse line

a, Schematic of HaloTag-Nup107 fusion, including 5’ untranslated region (UTR, light grey), HaloTag open reading frame (magenta), flexible linker (dark grey), and Nup107 coding sequence (cyan). b, Immunoblot of P0 cortical lysates from Nup107^+/+, Nup107^KI/+, and Nup107^KI/KI mice probed with anti-Nup107 and anti-HaloTag antibodies. GAPDH was blotted as a loading control. Each band represents an independent biological sample (3 animals each). c, Distribution of genotypes in litters derived from intercrossing Nup107^KI/+ mice. No deviation from the expected Mendelian ratio was observed (P=0.3729, Chi-squared test). 74 pups from 9 litters were analyzed. d, Superplots of nuclear ROI area from Nup153 and JFX554 SIM images. Plots show mean ± SD, with color coding indicative of biological replicates. ns, not significant; repeated measures one-way ANOVA with Dunnett’s multiple comparisons test on 3 biological replicates, with DIV4 as a reference condition. e, Representative image of Nup153 and HaloTag-Nup107 colocalization in DIV10 HaloTag-Nup107 neurons from 3 biological replicates. Scale bar = 2 μm. f, Violin plot showing the percent of Nup153 and JFX554 puncta that colocalize with JFX554 and Nup153, respectively, in DIV10 neurons. Nup153 colocalization with JFX554 was measured by calculating the % of Nup153 peaks that overlap with thresholded JFX554 puncta, and vice versa. g, Frequency distribution of nearest neighbor distances of JFX554 puncta. For each JFX554 peak, the distance to its nearest neighbor was calculated. Results from three biological replicates were combined. h, Frequency distribution of JFX554 puncta within two-pore diameter (240nm) distance. For each JFX554 peak, the number of neighboring puncta within a radius of 240nm was calculated. Results from three biological replicates were combined. i, Distribution of genotypes in litters derived from intercrossing Tor1a^+/−; Nup107^KI/KI mice. All genotypes were born at the expected Mendelian ratio (P-0.4880, Chi-squared test). 46 pups from 6 litters were analyzed.

Extended Data Figure 5:. Comparison of NPC density and distribution from HaloTag-Nup107 pulse-chase

a, b, c, Nup153 (a), JFX554 (b), and new JF646 (c) puncta density in maturing WT and torsinA-KO primary neurons aged DIV4, 6, 8, 10. Plots show mean ± SD from 3 biological replicates. Repeated measures two-way ANOVA with Sidak’s multiple comparisons test was used to compare genotypes at each timepoint. No comparisons reached statistical significance. Repeated measures two-way ANOVA with Dunnett’s multiple comparisons test was used to compare all timepoints to DIV4, with ****P<0.0001. d, Autocorrelation of JFX554 images over 0–500nm separation distance. Similar starting amplitudes reflect constant JFX554 density. e, Autocorrelation of JF646 images over 0–500nm separation distance. Decreasing starting amplitudes over neuronal maturation reflect increasing JF646 density. Broadening of the curve in torsinA-KO neurons indicates spatial correlation of NPCs over larger distances.

Extended Data Figure 6:. Bleb-associated pores are narrow and exhibit nucleoplasmic central plugs

a, Slices from a DIV10 torsinA-KO tomogram oriented as an en-face view of the neuronal nuclear membrane. Pores at clusters of blebs, isolated blebs, or regular NPCs without blebs are marked by color-coded spheres centered around the middle plane of the pore channel, denoted by larger circles. Tomogram shown here is representative of 8 tomograms from two biological replicates. b, Projection of all marked pores from the tomogram shown in (a). c, Superplot of WT, KO regular (non bleb-associated), and KO bleb-associated pore diameter (nm). Plots show mean ± SD, with color coding indicative of biological replicates. Swarmplot of individual pore diameters (small points) is overlayed with the mean of pore diameters for each biological replicate (larger points). Pores from two WT and two torsinA-KO tomograms were analyzed. KO regular and bleb-associated pores were identified from the same torsinA-KO tomograms. *P=0.0128, **P=0.0068; two-tailed unpaired t-test from two biological replicates. d, Blinded quantitation of the percentage of pores that have dense central plugs. Plots show mean ± SD with each point representing a cell. For KO regular and bleb-associated pores, points are color-coded to reflect quantitation from the same tomogram. Blinded quantitation was performed on 593 pores from 8 WT and 8 torsinA-KO tomograms obtained from two biological replicates each, with each tomogram from a different cell. *P=0.0302; two-tailed unpaired t-test with Welch’s correction, ****P<0.0001 two-tailed paired t-test, using values from 8 tomograms. e, Slices showing individual pores from DIV10 WT and torsinA-KO tomograms. Regular and bleb-associated torsinA-KO pores were sampled from the same tomogram. From top to bottom, the slices progress from the cytoplasmic side of the pore channel towards the nucleoplasm. Center plane of the pore channel is marked with a horizontal arrow. Scale bar = 100nm.

Extended Data Figure 7:. NE blebs spatially correlate with NPC clusters

Line scan analyses of 8 WT and 8 torsinA-KO DIV10 nuclei from two biological replicates labeled with anti-Nup153 and anti-Ubiquitin-K48 antibodies.

Extended Data Figure 8:. Nup358 is recruited to persisting NPC clusters in DIV18 torsinA-KO neurons

a, SIM images of DIV10 and DIV18 WT and torsinA-KO neurons labeled with Nup153 and Nup358 antibodies. Scale bar = 2μm. b, dSTORM images of Nup210 in DIV18 WT and torsinA-KO neurons. Scale bar = 2μm. Right panels show zoomed in view. Scale bar for right panels = 200nm. c, Autocorrelation plot of DIV18 WT and torsinA-KO neurons, determined from Nup210 dSTORM images. 18 WT and 19 torsinA-KO cells from 3 biological replicates each were analyzed. d, Normalized localization density along radial distance in averaged WT and torsinA-KO NPCs from DIV18 dSTORM images. Plots show mean ± SEM from eight bootstrapping rounds with 250 randomly selected pores each. *P=0.019. e, Confocal images from DIV10-18 pulse-chase of WT and torsinA-KO neurons expressing HaloTag-Nup107. Scale bar = 5μm. f, Plot of Nup358 nuclear rim fluorescence intensity from DIV10-18 pulse-chase. Plots show mean ± SD from two biological replicates with color coding to indicate replicates. All values were normalized to the mean DIV10 WT fluorescence intensity. Reported P values are from a two-way ANOVA test. “n” denotes the number of analyzed nuclei. g, Plot of JFX554 nuclear rim fluorescence intensity from DIV10-18 pulse-chase. Plots show mean ± SD from two biological replicates with color coding to indicate replicates. All values were normalized to the mean DIV10 WT fluorescence intensity. Reported P values are from a two-way ANOVA test. Same number of nuclei were analyzed as in (f).

Extended Data Figure 9:. Summary of the effects of torsinA loss on NPC spatial organization and dynamics

a, Model of interphase NPC assembly in WT and torsinA-KO neurons. Onset of NPC assembly is not affected by torsinA deletion. Nuclear basket, inner ring, and transmembrane nucleoporins are recruited to the nascent NPC as the INM starts to bud. Instead of the normal INM-ONM fusion found in WT neurons, excessive INM extrusion causes NE blebs to emerge and enlarge in torsinA-KO neurons. These blebs stall torsinA-KO NPCs at an intermediate stage while NPC assembly completes in WT neurons. As torsinA-KO neurons continue to mature, NE blebs resolve and INM-ONM fusion occurs. Completion of NPC biogenesis is delayed in torsinA-KO neurons. b, Model of NPC localization in maturing WT (top) and torsinA-KO (bottom) neurons. In WT neurons, newly forming NPCs (blue) localize to empty spaces between existing NPCs (red), thereby maintaining uniform spatial organization. In maturing torsinA-KO neurons, newly forming NPCs (blue) localize abnormally close to each other or to existing NPCs (red), causing aberrant clusters. NPC biogenesis is upregulated in both genotypes and total NPC number is not affected by the absence of torsinA.

Figure 1:. Nuclear pore complex biogenesis is upregulated during neuronal maturation

a, Structured illumination microscopy (SIM) images of primary neurons aged DIV4, 10, and 24 labeled with anti-Nup153 and anti-Nup98 antibodies. Bottom row shows identified Nup153 and 98 peaks. Representative images from 3 biological replicates. Scale bar = 2μm. b, Superplots of Nup153 density (puncta/μm²) over time in primary neurons. Plots show mean ± SD from 3 biological replicates. One-way ANOVA with Dunnett’s multiple comparisons test was performed, with DIV4 as the reference condition. ***P=0.0007, ****P<0.0001. Total number of analyzed nuclei are represented as “n”. c, Superplots of Nup98 density (puncta/μm²) over time in primary neurons. Plots show mean ± SD from 3 biological replicates. One-way ANOVA with Dunnett’s multiple comparisons test was performed, with DIV4 as the reference condition. ***P=0.0001 (DIV4 vs. DIV8), ***P=0.0003 (DIV4 vs. DIV10), ****P<0.0001. d, e, Frequency distribution of Nup153 (d) and Nup98 (e) puncta found within two-pore diameter (240nm) distance. For each identified Nup153 and Nup98 peak, the number of neighboring puncta within a radius of 240nm was calculated. Results from three biological replicates were combined at each timepoint.

Figure 2:. TorsinA is essential for uniform NPC distribution but not upregulation of NPC biogenesis

a, b, SIM images of WT and torsinA-KO primary neurons aged DIV4, 6, 8, 10, and 14 labeled with anti-Nup153 (a) and anti-Nup98 (b) antibodies. Representative images from 3 biological replicates. Scale bar = 2μm. c, dSTORM images of Nup210-labeled NPCs in DIV10 WT and torsinA-KO neurons. Scale bar = 2µm. Right panels show zoomed in view of boxed regions. Scale bar of right panels = 200nm. 16 WT cells from 4 biological replicates and 18 torsinA-KO cells from 5 biological replicates were imaged. d, Averaged aligned DIV10 WT and torsinA-KO pores. Scale bar = 100nm. e, Normalized localization density along radial distance in averaged WT and torsinA-KO pores. Plots show mean ± SEM from eight bootstrapping rounds with 250 randomly selected pores each. No adjustments were made. ***P=0.0005, two-tailed t-test. f, g, Nup153 (f) and Nup98 (g) density (puncta/μm²) in maturing primary neurons. Plots show mean ± SD from 3 biological replicates. Timepoints from each replicate were matched and repeated measures two-way ANOVA with Sidak’s multiple comparisons test was performed to compare means between genotypes. ns, not significant.

Figure 3:. Neurons bearing HaloTag-Nup107 knock-in alleles demonstrate endogenous NPC upregulation

a, Confocal images of DIV10 Nup107^+/+, Nup107^KI/+, and Nup107^KI/KI neurons labeled with JF646 HaloTag ligand. Scale bar = 20μm. Representative images from three biological replicates. b, SIM images of Nup107^KI/KI neurons at DIV4, 10, and 18 labeled with JFX554 HaloTag ligand and anti-Nup153 antibody. c, d, Superplots of Nup153 (c) and JFX554 (d) puncta density in primary neurons aged DIV4, 10, 18. Plots show mean ± SD from 3 biological replicates. ***P=0.0002, ****P<0.0001; One-way ANOVA with Dunnett’s multiple comparisons test, with DIV4 as the reference condition. “n”, number of analyzed nuclei. e, Confocal images of P0 and P14 cortical brain sections from Tor1a^+/+;Nup107^KI/KI mice labeled with JFX650 HaloTag ligand. Scale bar = 50μm. f, Normalized JFX554 HaloTag-Nup107 nuclear rim intensity from P0 and P14 cortical brain sections. All measurements were normalized to the mean intensity of P0 brain sections. Plots show mean ± SD; four biological replicates. *P=0.0379; two-tailed unpaired t-test with Welch’s correction. “n”, number of analyzed nuclei. g, Normalized nuclear area based on ROIs used in (f). All measurements were normalized to the mean nuclear area of P0 nuclei. Plots show mean ± SD; four biological replicates. ***P=0.0009; two-tailed unpaired t-test with Welch’s correction. “n”, number of analyzed nuclei. h, Confocal maximum intensity projection of DIV10 WT and torsinA-KO neurons expressing HaloTag-Nup107 labeled with JF646 HaloTag ligand and anti-Nup153 antibody. Merge shows Hoechst (blue), JF646 (magenta), and Nup153 (green). Scale bar = 10μm. h’, Zoomed-in view of cells in yellow boxes in (h). Scale bar = 10μm. i, Confocal maximum intensity projection of DIV10 WT and torsinA-ΔE/ΔE neurons expressing HaloTag-Nup107 labeled with JF646 HaloTag ligand and anti-Nup153 antibody. Merge shows Hoechst (blue), JFX554 (magenta), and Nup153 (green). Scale bar = 10μm. i’, Zoomed-in view of cells in yellow boxes in (i). Scale bar = 10μm. j, k, Normalized Nup153 nuclear rim intensity from DIV10 WT and torsinA-KO (j) and WT and torsinA-ΔE/ΔE (k) neurons. Intensity values were normalized to the mean of WT intensities. Plots show mean ± SD from two (j) or three (k) biological replicates. ns, not significant; two-tailed paired t-test. “n”, number of analyzed nuclei.

Figure 4:. Sites of NPC biogenesis are abnormal in torsinA-KO neurons

a, Diagram of potential mechanisms for NPC clustering in torsinA-KO neurons. In (1), NPCs redistribute after formation, leading to clusters of existing and newly formed NPCs. In (2), sites of NPC biogenesis are mislocalized and clusters exclusively contain new NPCs. Red circles represent existing NPCs (pulse; JFX554). Blue circles represent new NPCs (chase; JF646). b, Schematic of HaloTag pulse-chase experiment. c, HaloTag pulse-chase SIM images of WT and torsinA-KO neurons. Neurons were stained with anti-Nup153 antibody post-fixation to label the total NPC population. Scale bar = 2μm. d, Density of JFX554 puncta and new JF646 puncta in DIV 4, 6, 8, and 10 WT and torsinA-KO HaloTag-Nup107 neurons. Plots show mean ± SD from 3 biological replicates. Repeated measures two-way ANOVA with Dunnett’s test was performed with DIV4 values as a reference. **P=0.0024, ***P=0.003, ****P<0.0001 for WT; **P=0.0076, ****P<0.0001 for KO. e, Normalized autocorrelation of JFX554 SIM images over 0–500nm separation distance. Autocorrelation values <1 around 200nm in both WT and torsinA-KO neurons reflect nonrandom uniform distribution of JFX554-labeled NPCs. f, Normalized autocorrelation of JF646 SIM images over 0–500nm separation distance. Autocorrelation values >1 in torsinA-KO neurons reflect clustering of NPCs labeled with JF646.

Figure 5:. NE blebs spatially and temporally coincide with NPC biogenesis

a, Transmission electron microscopy (TEM) images of WT and torsinA-KO neurons at DIV4 and DIV10. N=nucleus. Fully-formed NPCs are marked with yellow arrowheads. NE blebs identified by blinded analysis are marked with red asterisks. Scale bar = 5μm for the top row, 2μm for inset. Images were acquired from at least three biological replicates for each genotype at each timepoint. b, Quantitation of TEM images. Plots show mean ± SD. Each point represents the % of cells with at least one NE bleb from each biological replicate. **P=0.0043, two-tailed paired t-test. Total number of analyzed nuclei are represented as “n”. c, Frequency of NE blebs per cell from all replicates. d, Slices from a DIV10 torsinA-KO tomogram overlayed with segmented contours of the outer nuclear membrane (cyan), inner nuclear membrane (magenta), membranes inside blebs (green), and central plug of pores (yellow). Tomogram shown here is representative of 8 tomograms from two biological replicates. e, Rotated view of tomogram in (d) to show an XY view intersecting the center of two blebs at the yellow dashed line. Red and white arrows label bleb-associated pores with central plugs. The rotated view (right) demonstrates that these pores form the base of each bleb. f, Segmented bleb clusters from a DIV10 torsinA-KO tomogram. Scale bar = 1μm. 3D view is available in Supplementary Video 2.

Figure 6:. NPC assembly completes following NE bleb resolution

a, TEM images of WT and torsinA-KO neurons at DIV10 and DIV18. N=nucleus. Fully formed NPCs are marked with yellow arrowheads. NE blebs identified in blinded analysis are marked with red asterisks. Scale bar= 5μm for the top row, 2μm for inset. Images were acquired from three biological replicates for each genotype at each timepoint. b, Quantitation of TEM images. Plots show mean ± SD from 3 biological replicates. Each point represents the % of cells with at least one NE bleb from each biological replicate. Total number of analyzed nuclei are represented as “n”. **P=0.006; two-tailed paired t-test. c, Frequency of NE blebs per cell from all replicates. d, Confocal images of DIV10 and DIV18 WT and torsinA-KO neurons labeled with Hoechst, Nup153, Nup358, and Map2 (shown in merge). Scale bar = 10μm. e, Zoomed in view of Nup153 and Nup358 channels of nuclei marked with yellow boxes in (d). Scale bar = 10μm. f, g, Normalized Nup153 (f) and Nup358 (g) nuclear rim intensity of DIV10 and DIV18 WT and torsinA-KO primary neurons. Plots show mean ± SD from 3 biological replicates. ns, not significant; **P <0.01, repeated measures two-way ANOVA with Sidak’s multiple comparisons test. For (f), **P=0.0019 for WT and **P=0.0012 for torsinA-KO DIV10 vs. DIV18; For (g), *P=0.0206 for DIV10 WT vs. torsinA-KO, **P=0.0087 for WT DIV10 vs. DIV18 and **P=0.0050 for torsinA-KO DIV10 vs. DIV18.