Spectrin condensates provide a nidus for assembling the axonal membrane-associated periodic skeleton

Abstract

In axons, α/β-spectrins, adducin, and actin filaments assemble into a lattice underneath the plasma membrane, but the mechanistic events leading to this membrane-associated periodic skeleton (MPS) are unclear. Visualizing MPS-components in developing axons, we found distal focal patches containing spectrins and adducin (but sparse actin filaments) with biophysical properties reminiscent of biomolecular condensation. Overexpressing spectrin repeats - constituents of α/β-spectrins - in heterologous cells triggered condensate formation, and preventing the association of βII-spectrin with actin filaments or membranes also facilitated condensation. Introducing a stretch of spectrin-repeats in neurons before MPS establishment triggered ectopic condensate-like structures in the soma and disrupted the axonal lattice, advocating a functional role for biomolecular condensation. We propose a condensation-assembly model where spectrin-repeats trigger focal phase separated condensates, providing a nidus for MPS assembly that recruits actin filaments to ultimately generate the stable lattice. Our overall model is supported by recent studies showing phase-separation via coiled-coil domains and recruitment/polymerization of actin by other condensate-forming proteins.

Keywords: Cell biology; Molecular neuroscience; Neuroscience.

Graphical abstract

Figure 1

Spectrin/adducin patches in distal axon during development (A) Schematic of the MPS and timing of experiments. Hippocampal neurons were obtained from P0 pups and cultured for 2, 4, 7, or 10 days in vitro (DIV) before fixation and immunostaining. (B) Schematic showing the development of the axon through neuronal stages 3–5. Note formation of the axon initial segment (AIS) during stage 4 and synaptic contacts with target neurons at stage 5. (C–E) Representative images of 2 DIV and 7 DIV neurons stained for βIII-tubulin (volume marker), αII-spectrin or βII-spectrin, and adducin. Note the accumulation of spectrin and adducin in the proximal axon. Soma/axon junction is marked by a dashed line, double arrowheads label the growth cone, and a bracket marks patches along the axon. Scale bars = 10 μm. (F) Size and density of spectrin patches were greater in axonal segments closer to the AIS/proximal axon (mean ± SEM, unpaired t test with Welch’s correction, data analyzed from a sample of 9 axons from 9 DIV neurons, ∗∗∗p < 0.0001, ∗∗∗∗p < 0.00001). (G) Quantification of axon patches. Fluorescence intensity was measured at every point along the axon. The maximum and minimum fluorescence within 5 μm of each point was calculated, and the difference was normalized to the average fluorescence within 20 μm, giving an estimate of the discontinuity of fluorescence. The logarithm of this normalized difference was the discontinuity factor δ (see STAR Methods for more details). Scale bars = 5 μm. (H) Discontinuity factors for αII/βII-spectrin, adducin, and βIII-tubulin in neurons fixed at 2, 4, 7, or 10 DIV. Note that a higher positive δ indicates a patchier distal axon, and in general, discontinuity in the distal axon increases as the axon develops (data from 9 to 46 cells for each time point, from 3 independent cultures).

Figure 2

Paucity of actin filaments in distal spectrin patches (A) Neurons at 7 DIV were fixed and stained for F-actin and αII-spectrin. In zoomed insets of the distal axon, note that many spectrin patches lack F-actin, best seen in the merged image on the right. Scale bars = 10 μm. (B) A maximum-intensity line-scan of the dashed yellow line from (A). Note that many patches of αII-spectrin do not coincide with peaks of F-actin (brackets), while some do (asterisks). (C) Quantification of αII-spectrin patch colocalization with F-actin (phalloidin) or G-actin (DNase). Note that more spectrin patches in the proximal and medial thirds of the axon contained F-actin, while patches in the distal third were more likely to colocalize with G-actin or contain no actin (mean ± SEM, two-way ANOVA, data from 31 cells from 3 independent cultures, ∗∗∗p < 0.001). (D and E) STORM of βII-spectrin (D) or αII/βII-spectrin (mixed antibodies) together with actin (E) in proximal (1), middle (2), and distal (3) regions of a rat hippocampal neuron fixed at 6 DIV. XY zooms and XZ axis projections of the regions delineated by yellow lines are shown later in discussion, together with a 3D rendering of the section. Note that while proximal axons have the expected periodic appearance of spectrin, middle and distal axonal regions have interrupted periodicity with incomplete annular structures, or patchy distribution with no periodicity. Though most of the actin in proximal axons was aligned with spectrin as expected, this was not the case in distal axons (E). Scale bars = 20 μm (top view), 5 μm (middle views), 500 nm (zooms and transverse sections).

Figure 3

Spectrin in axonal patches can dynamically exchange with the axoplasm (A) Schematic of live-cell imaging and FRAP experiments. 5 DIV neurons were co-transfected with αII-spectrin:mGL and mScarlet (volume filler). At 7 DIV, αII-spectrin:mGL was photobleached either in proximal axons (within 100 μm of soma), or in distal axonal patches. (B) Live imaging of αII-spectrin:mGL in proximal and distal axons. Note distinct patches of fluorescence in distal axon, reminiscent of the endogenous spectrin pattern. These patches were largely stationary over ∼15 min of imaging. Scale bars = 2 μm. (C) Timelapse images show the recovery of αII-spectrin:mGL fluorescence following photobleaching (yellow dashed region) in the proximal axon or distal patches. Note that while the proximal axon shows minimal recovery of fluorescence – consistent with a stable lattice – distal axonal patches partially recover within minutes. Scale bars = 2 μm. (D) Quantification of FRAP experiments shows significantly faster recovery in distal patches (n = 12/condition, in axons from 3 independent cultures, two-way ANOVA, interaction ∗∗∗p < 0.0001).

Figure 4

FL-αII-spectrin forms biomolecular condensates in heterologous cells (A) Overall schematic of experiments expressing mGL-tagged FL or truncated αII/βII-spectrins/adducin in HEK293T cells. (B) Expression of mGL-tagged FL αII-spectrin, βII-spectrin, and α-adducin in HEK293T cells. Note that FL-αII-spectrin formed distinct inclusions (zoomed in dashed inset), βII-spectrin showed a membranous localization (also see Video S2), and α-adducin was diffuse. Scale bars = 10 μm. (C) Expression of a construct known to induce phase separation (engineered dimer of the RGG domain of yeast LAF1 protein fused to GFP), or soluble (untagged) mGL in HEK293T cells. Scale bars = 10 μm. (D) Frames from timelapse imaging of αII-spectrin:mGL inclusions in HEK293T cells show that they are largely stationary and do not undergo merging or splitting. Scale bars = 1 μm. (E) Top panel: Schematic showing kinetics of FRAP when a single inclusion is completely photobleached. Note that the recovery is due to the diffusion of fluorescent molecules from unbleached regions into the bleached region, across the surface of liquid droplets (small yellow arrows). Middle/bottom panels: FRAP of αII-spectrin droplets shows recovery of 40–50% of fluorescence within 45 s (n = 8, cells from 3 independent cultures, scale bars = 1 μm). (F) Top: Schematic showing the principle of MOCHA-FRAP experiments, in which one-half of a fluorescent inclusion is photobleached. Note that FRAP in this scenario is due to the diffusion of fluorescent molecules from the unbleached region (long yellow arrow), as well as from across the surface of putative droplets (small yellow arrows). Later in discussion: Assuming that the fluorescence redistributes within the inclusion at diffusion rate D and fluorescence in the droplet as a whole recover at interface flux-rate F, theoretical graphs show fluorescence of both unbleached and bleached droplet halves in a scenario in which D ≫ F (left), or when D ≪ F (right). The depth of the fluorescence dip in the unbleached half correlates to the balance between these two rates and is an indicator of liquid-like behavior. (G) MOCHA-FRAP analyses of αII-spectrin:mGL inclusions. Top panel: Representative FRAP images of a single αII-spectrin:mGL inclusion (top). The bleached region is marked with a dashed red line; note recovery of fluorescence in the bleached region. Bottom panel: Quantitative graphs of unbleached and bleached regions, obtained using the MOCHA-FRAP workflow. Black vertical bar indicates a dip depth of 19.1% at the location of minimum fluorescence of the unbleached half, which falls within the expected range for phase separation in the MOCHA-FRAP model (n = 8, cells from 3 independent cultures). Scale bars = 1 μm. (H) MOCHA-FRAP experiments on RGG-GFP-RGG droplets display fast internal redistribution and fluorescence recovery, with a dip depth of 14.2%, which is consistent with phase separation (n = 9, cells from 3 independent cultures). Scale bars = 1 μm.

Figure 5

Spectrin-repeat domains within FL-αII-spectrin show droplet-like behavior (A) Schematic showing FL and domain-deletion constructs of αII-spectrin that were tested for droplet formation in HEK293T cells. Constructs lacking the SH3 domain (ΔSH3), the N- and C-termini (SR5-20), or spectrin repeats 10–20 (SR1-9) were tagged with mGL. (B) Discrete, rounded inclusions were seen with ΔSH3 and SR5-20, but not with SR1-9 (highlighted in zoomed inset). Large swaths of fluorescence are also seen (black arrowheads), which may be related to the extent of over-expression and biophysical properties (also see Figure S3 and results, scale bars = 10 μm). (C) Full-inclusion FRAP analyses of mGL-tagged ΔSH3 and SR5-20 (top: images –bleached region outlined, bottom: FRAP-curves). Note that both αII-spectrin fragments show ∼40–60% fluorescence recovery (data obtained in cells from three independent cultures, scale bars = 1 μm). (D) MOCHA-FRAP analyses of mGL-tagged ΔSH3 and SR5-20 inclusions (top: images – dashed red line marks bleached region, bottom: fluorescence-curves from unbleached and bleached halves). Black bars indicate the dip depths at the locations of minimum fluorescence of the unbleached half. Note that in both cases, the dip-depth falls within the predicted range for phase separation, highlighted in gray (n = 7–10/condition, cells from 3 independent cultures, scale bars = 1 μm). (E) Schematic depicting steps for determining internal diffusion coefficient D from MOCHA-FRAP experiments. D is calculated from the Brownian diffusion curves (red) fit to the internal diffusion (black), which is equal to the fluorescence recovery of the bleached half (orange) with the total droplet recovery (pink) subtracted. (F) Graphs of internal diffusion from MOCHA-FRAP experiments with FL-αII-spectrin as well as SR5-20 and ΔSH3 constructs, with diffusion equation fits in red (n = 7–10, cells from 3 independent cultures). (G) Quantification of D from individual diffusion curves from populations of droplets, summarized in (F). Deletion of the N/C-termini or SH3-domain results in faster internal diffusion than within full-length protein droplets (mean ± SEM, one-way ANOVA with Dunn post hoc ∗, p < 0.05).

Figure 6

Isolated spectrin-repeats in βII-spectrin can form condensates (A) Schematic showing full-length (FL) and domain-deletion constructs of βII-spectrin that were tested for droplet formation in HEK293T cells. Constructs lacking the actin-binding calponin homology domains (ΔCH-CH), containing only the 17 spectrin repeats (SR1-17), or containing only the C-terminal membrane-binding pleckstrin homology domain (PH) were tagged with mGL. (B) Discrete inclusions were seen with ΔCH-CH and SR1-17, but not with PH (highlighted in zoomed inset), suggesting that eliminating actin- and membrane-binding promoted inclusion-formation (scale bars = 10 μm). (C) Full-inclusion FRAP analyses of mGL-tagged ΔCH-CH and SR1-17 (top: images – dashed red line marks bleached region, bottom: FRAP-curves). Note that recovery was faster when both actin- and membrane-binding domains were deleted (SR1-17, compared to DCH-CH; n = 7–9/condition, in cells from 3 independent cultures; scale bars = 1 μm). (D) MOCHA-FRAP analyses of mGL-tagged ΔCH-CH and SR1-17 inclusions (top: images – dashed red line marks bleached region, bottom: fluorescence-curves from unbleached and bleached halves). Black bars indicate the dip depths at the locations of minimum fluorescence of the unbleached half. Note that only the dip-depth of SR1-17 falls within the predicted range for phase separation, highlighted in gray (n = 7–9/condition, cells from 3 independent cultures, scale bars = 1 μm). (E) Internal diffusion graphs with fitted curves in red, and graphs of D calculated from population analyses (see method in Figures 5E–5G). Internal diffusion rates for ΔCH-CH and SR1-17 are similar, suggesting that MOCHA-FRAP differences may be driven by interface flux. Mean ± SEM, data from 7 to 9 droplets/cells from 3 independent cultures. Scale bars = 1 μm. (F) Testing the role of actin filaments in βII-spectrin droplet-formation. HEK293T cells transfected with FL-βII-spectrin or SR1-17-βII-spectrin were treated with 20 μM latrunculin A for 30 min to disrupt F-actin (or 4% DMSO as controls). (G) Top panels show that upon latrunculin treatment, there is a redistribution of FL-βII-spectrin from its normal membranous distribution to intracellular inclusions (zoomed insets on the right). Bottom panels show the effects of latrunculin on SR1-17 inclusions. Representative images shown, scale bars = 10 μm. (H) MOCHA-FRAP analyses of FL-βII-spectrin inclusions formed after latrunculin-treatment. Top panel shows time-lapse images of an inclusion that was partially photobleached (bleached region highlighted by a red dashed line), bottom panel shows curves from MOCHA-FRAP analyses. Note that the dip-depth (22.3%, black vertical bar) is within the expected range for phase separation (gray box; n = 6 from 3 independent cultures; scale bars = 1 μm).

Figure 7

Actin overexpression attenuates spectrin discontinuity in distal axons (A) Schematic showing the design of experiments to overexpress actin and evaluate endogenous spectrin. (B and C) Examples of αII-spectrin staining in axons transfected with GFP:actin (or soluble GFP as controls). Note actin overexpression led to decreased spectrin patchiness (increased continuity) in distal axons, quantified in (C) using the “discontinuity analyses” algorithm (at least 5–6 neurons/condition, three separate cultures, ∗∗∗p < 0.0001). (D–F) Neurons were either transfected with actin-Chromobody (to label endogenous actin), or with GFP:actin (to overexpress actin), and incorporation of endogenous or overexpressed actin into spectrin-patches was evaluated. Note that while endogenously labeled actin showed scant colocalization with spectrin patches as expected (E), actin-overexpression led to more actin/spectrin colocalization (F), suggesting greater incorporation of actin into spectrin patches in the latter scenario (mean ± SEM, unpaired t test with Welch’s correction, ∗∗∗∗p < 0.0001). Scale bars in (E) = 20 μm.

Figure 8

Early overexpression of minimal spectrin repeats disrupts endogenous MPS assembly in neurons (A) Overall experimental design: Spectrin repeats of βII-spectrin (SR1-17:mScarlet) or soluble mScarlet were transfected in cultured hippocampal neurons either during (3–6 DIV), or after (7–10 DIV) the establishment of the endogenous MPS in the proximal axon (region within 100 μm from soma). Neurons were subsequently fixed, stained with an antibody against the C-terminus of βII-spectrin (that would label endogenous βII-spectrin but not the transfected SR1-17), and spectrin periodicity in axons was evaluated by SIM. (B) SR1-17:mScarlet formed large inclusions in the neuronal soma with scant fluorescence in axons (marked by line/arrowheads, transfected DIV 7, analyzed DIV 10). Relative intensities of FL-βII-spectrin:mScarlet and SR1-17:mScarlet overexpression is quantified on the right (6–7 transfected neurons were analyzed for each condition from 2 separate coverslips. ∗, p < 0.05, also see Figure S10A). (C and D) Neuronal SR1-17 inclusions also contained βII-spectrin (endogenous, see results for explanation) and actin, but no adducin (also see Figure S10B). Neurons were transfected at 3 DIV with SR 1–17:mScarlet and Actin-Chromobody, and fixed at 6 DIV (representative data from ∼15 neurons for each condition, 3 coverslips). FRAP recovery of neuronal SR1-17:mGL inclusions shown in (D). (E) Periodicity in proximal axons of neurons transfected with mScarlet-tagged SR1-17 (or soluble mScarlet) at 3 DIV. Note that the overexpression of SR1-17 impeded the assembly of the MPS; zoomed insets later in discussion. (E) Fluorescence intensity profiles and autocorrelation analysis of axonal segments from (E) using SIM – zoomed insets, analyzed region marked by white brackets. Note disrupted periodicity upon transfection with SR1-17. (F) Quantification of the βII-spectrin periodicity in proximal axons. Periodicity in neurons expressing SR1-17:mSc was significantly lower than in neurons expressing soluble mScarlet (mean ± SEM, unpaired t test, nonparametric, Kolmogorov-Smirnoff, data from 15 to 25 neurons from 3 independent cultures. ∗∗∗, p < 0.0001). Scale bars: (B) = 100 μm; (C) = 10 μm; (D) 2 μm; (E) Top panels = 2 μm, bottom panels = 1 μm.

Figure 9

Model for MPS assembly during development (A) The MPS is first established in the most proximal part of the axon near cell bodies, followed by patches of spectrin/adducin (condensates) in more distal regions. Over time, the incorporation of actin into the patches generates a periodic lattice that extends laterally from the patches. A greater size/density of patches in axonal segments closer to the soma – presumably due to biased deposition of components via axonal transport – leads to a proximal to the distal extension of the MPS as the axon extends. (B) A zoomed-in view of proposed mechanistic events.