A dynamic formin-dependent deep F-actin network in axons

Abstract

Although actin at neuronal growth cones is well-studied, much less is known about actin organization and dynamics along axon shafts and presynaptic boutons. Using probes that selectively label filamentous-actin (F-actin), we found focal "actin hotspots" along axons-spaced ∼3-4 µm apart-where actin undergoes continuous assembly/disassembly. These foci are a nidus for vigorous actin polymerization, generating long filaments spurting bidirectionally along axons-a phenomenon we call "actin trails." Super-resolution microscopy reveals intra-axonal deep actin filaments in addition to the subplasmalemmal "actin rings" described recently. F-actin hotspots colocalize with stationary axonal endosomes, and blocking vesicle transport diminishes the actin trails, suggesting mechanistic links between vesicles and F-actin kinetics. Actin trails are formin-but not Arp2/3-dependent and help enrich actin at presynaptic boutons. Finally, formin inhibition dramatically disrupts synaptic recycling. Collectively, available data suggest a two-tier F-actin organization in axons, with stable "actin rings" providing mechanical support to the plasma membrane and dynamic "actin trails" generating a flexible cytoskeletal network with putative physiological roles.

Figure 1.

F-actin dynamics in axons. Neurons were transfected with GFP:Utr-CH to label F-actin (and soluble mRFP to visualize morphology), and F-actin dynamics in axons were imaged as described in methods. (A) Selected frames from a representative video showing anterograde (left) and retrograde (right) dynamics of actin polymers (dynamic tips marked by single/double arrowheads). Image far below shows soluble mRFP in the same axon. Elapsed time in seconds on left/right (see Video 1). (B and C) Kymograph of the video in A with first frame on top shown in B. Boxed ROIs are zoomed in C. Note unusual F-actin dynamics including diagonal, vectorial plumes of fluorescence (actin trails) and hotspots (vertical interrupted lines). Yellow and red arrowheads represent the same dynamic structures in A and B. Elapsed time is given in seconds on the left. (D) Cropped kymographs from other videos show focal F-actin hotspots (some marked by asterisks). Instances of actin trails emerging from—and collapsing into—hotspots can also be seen. Note trails marked by arrows 1, 2, 6, and 7 are emerging from hotspots. Antero, anterograde; Retro, retrograde.

Figure 2.

Quantitative analyses of F-actin hotspots and trails. (A) Mean lifetime of F-actin hotspots was 80.40 ± 3.95 s (n = 575 events from 28 axons). (B) Mean number of F-actin trails (every 10 min of imaging) was 20.35 ± 1.98 for anterograde events and 14.92 ± 1.35 for retrograde events (992 events from 28 axons). (C) Mean total lengths of F-actin polymers were 8.85 ± 0.18 µm (n = 573 events) and 8.87 ± 0.22 µm (n = 419 events) for anterograde and retrograde trails, respectively. (D) Mean elongation rates of F-actin trails were 0.99 ± 0.01 µm/s in both directions. (E) On average, 80.59 ± 2.23% F-actin trails emerged from hotspots (n = 36 axons). (F) On average, 63.00 ± 2.80% F-actin trails collapsed into hotspots (n = 36 axons). Arrowheads and asterisks in E and F represent initiation/termination of actin trails and position of hotspots, respectively. All values represent means ± SEM. Antero, anterograde; Retro, retrograde.

Figure 3.

3D STORM imaging of actin reveals intra-axonal actin bundles. (A and A′) Widefield image of a hippocampal neuron labeled with phalloidin (F-actin) and Neurofascin-186 (NF-186; to label the axon initiation segment [AIS], brackets). Proximal axon marked with arrowheads. (A′) 3D-STORM image of the actin labeling corresponding to the boxed area in A. Color codes for Z depth from −300 nm (blue) to 300 nm (red). Note Y-shaped proximal axon, as well as distal axons from other neurons. (B and C) Zoomed XY projections (300–600 nm thick) of the boxed regions (B and C) in image A′. Note that several longitudinal actin filaments can be seen within the axon shaft (arrowheads), in addition to periodic actin rings (comb). (B′ and C′) XZ slices (800 nm thick) from axons shown in B and C (area within brackets) highlighting intra-axonal actin filaments (arrowheads, dashed ellipse localizes the axon boundary). (D) 3D STORM image of distal axons with a dendritic shaft on right. Color codes for Z depth from −300 nm (blue) to 300 nm (red). (E and E′) XY/XZ projections (300–400 nm thick) of the boxed regions E in image D. Note longitudinal actin filaments inside the distal axon (arrowheads), together with the periodic actin rings (comb). (F and F′) Same imaging parameters and scales as previous, showing an axon segment with actin rings. Bars: (A) 20 µm; (A′ and D) 5 µm; (B, C, and E–F′) 2 µm; (B′ and C′) 500 nm.

Figure 4.

Axonal F-actin dynamics are dependent on actin turnover. (A and B) Kymographs from an axon transfected with GFP:Utr-CH (to label F-actin) and imaged before/after treatment with 100 nM latrunculin or 100 nM jasplakinolide. (A) Note gradual change in the slopes of the F-actin trails over 15 min of incubation, indicating attenuated polymerization. Drug washout (W/O) restores F-actin dynamics (right). (B) Treatment with jasplakinolide for 10 min essentially eliminates the on/off kinetics, leading to stationary F-actin accumulations along axons. (C) Kymographs of F-actin dynamics before and after treatment with 10 µg/ml nocodazole (NOC; note minimal change; Fig. S3 A). (D–G) Quantification of all F-actin dynamics after pharmacological treatments. Note that in general, actin-modulating drugs (latrunculin [LAT], cytochalasin-D, and jasplakinolide [JASP]) attenuate actin dynamics, whereas the microtubule-disrupting agent nocodazole has no effect. Also note increase in hotspot duration upon cytochalasin-D treatment is likely caused by the actin-capping effect of this agent (Cooper, 1987). Increase in hotspot duration upon jasplakinolide treatment may reflect hyperstabilization of F-actin dynamics. All experiments were performed before and after treating the same axon with the stated drug. For latrunculin treatment, n = 8 axons; cytochalasin-D treatment, n = 8 axons; jasplakinolide treatment, n = 6 axons; and nocodazole treatment, n = 8 axons were imaged. At least three independent repeats were performed for each condition. All values represent means ± SEM; (***, P < 0.001; **, P < 0.01, paired t test). For detailed statistics, see Table S1. Arrows between images represent passage of time in before/after experiments.

Figure 5.

Correlation of axonal F-actin dynamics and stationary endosomes. (A) Kymographs from neurons transfected with GFP:Utr-CH (to label F-actin) and pHRodo (a pH-sensitive endosomal marker that largely labeled stationary endosomes in axons, see Materials and methods), simultaneously visualized by live imaging. Note colocalization of F-actin hotspots (green) with pHrodo (red, overlay on right). A′ shows a zoomed ROI from overlay highlighting two actin trails originating precisely from where two stationary endosomes are situated (marked by asterisks). (B) Kymographs from neurons transfected with GFP:Utr-CH (to label F-actin) and Rab5:mCherry (to label early endosomes). Note colocalization of F-actin hotspots with early endosomes. (C, left) Quantification of colocalization data. The mean frequency of F-actin hotspots that overlapped with stationary endosomes labeled with pHRodo, Rab5-mRFP, and Lamp1-mCherry was 46.81 ± 2.76% (n = 13 axons), 29.04 ± 3.75% (n = 10 axons), and 9.58 ± 3.61% (n = 6 axons), respectively. (right) F-actin dynamics also correlated with number of pHRodo-positive endosomes in axons (n = 16 axons). (D and E) Neurons weretransfected with GFP:Utr-CH (to label F-actin) and treated with Brefeldin-A (BFA) to deplete vesicles in axons (Fig. S4 B). Kymographs show that F-actin dynamics in axons were greatly attenuated upon BFA treatment and restored upon washout of the drug; quantified in E. All values represent means ± SEM. ***, P < 0.001, one way analysis of variance followed by Dunnett’s post hoc test. For detailed statistics, see Table S1.

Figure 6.

Axonal F-actin dynamics are Formin dependent. (A) Kymographs from an axon transfected with GFP:Utr-CH (to label F-actin) and imaged before (left) and after (right) treatment with the formin-inhibitor SMIFH2. Note dramatic attenuation of the F-actin trails with little effect on F-actin hotspots, suggesting that the nucleation of F-actin at hotspots is spontaneous, and not formin dependent. (B) Similar experiments as in A, except that neurons were treated with the Arp2/3 inhibitor CK-666. Note that addition of CK-666 essentially has no effects on axonal F-actin dynamics. (C–F) Quantification of F-actin dynamics from all experiments. Note that treatment with SMIFH2 attenuates various kinetic behaviors of the F-actin trails, whereas CK-666 has no effects. Also note that SMIFH2 treatment leads to an increase in the hotspot duration, perhaps a compensatory response to attenuated actin trails (also see Results). For SMIFH2 treatment, n = 9 axons and for CK-666 treatment, n = 6 axons were imaged. At least three independent repeats were performed for each condition. All values represent means ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001; paired t test. For detailed statistics, see Table S1. Arrows between images represent passage of time in before/after experiments.

Figure 7.

Presynaptic F-actin dynamics are Formin dependent. (A) Images (left) and kymographs (right) of boutons from neurons transfected with GFP:Utr-CH (to label F-actin) and synaptophysin:mRFP (SyPhy:mRFP to label synaptic vesicle clusters). Note that F-actin appears as dynamic patches, circumferentially organized around the synaptic vesicle cluster. (B) Same experiment as above, except GFP:Utr-CH kymographs are scaled to reveal actin trails terminating into boutons (small arrowheads). (C and D) Representative images from a FRAP assay to detect F-actin entry into boutons, quantified in D. Neurons were transfected with GFP:Utr-CH and F-actin–enriched boutons were identified; a single bouton (dashed circles) within a string of synapses was photobleached, and recovery of fluorescence was visualized over time (see Materials and methods for details). Note rapid recovery of F-actin that is attenuated upon formin inhibition by SMIFH2, indicating diminished entry of F-actin into boutons after formin inhibition. (E and F) Maximum intensity projection images from a representative synaptic bouton demonstrate reduction in total F-actin fluorescence upon SMIFH2. Graph in F shows that fluorescence intensities decrease from 1,859 ± 351.7 to 999.1 ± 146.7 arbitrary fluorescence units (n = 9 boutons paired) upon SMIFH2 treatment for 30 min. All values represent means ± SEM. Paired t test **, P < 0.01.

Figure 8.

The Formin inhibitor SMIFH2 suppresses synaptic recycling. (A) Schematic of vGlut1-pHluorin experiments. SV refers to synaptic vesicle. (B) Representative panels show the fluorescence intensity change of vGlut1-pHluorin upon 600 action potential (AP) stimulation and NH₄Cl perfusion. Note that NH₄Cl alkalinizes all vesicles, revealing the total (recycling + resting) pool in these neurons. (C) Ensemble mean of vGlut1-pHluorin traces from control, 50 µM CK-666, or 30 µM SMIFH2-treated neurons (n = number of boutons). Note that although SMIFH2 attenuates neurotransmitter release and decreases synaptic vesicle endocytosis compared with control, CK-666 has no effect, quantified in C (all data normalized to total pools). (D, left) Recycling: Total pool ratio for control = 49.55 ± 3.81%; CK-666 = 45.65 ± 3.72%; and SMIFH2 = 23.84 ± 2.82%. (middle) Exocytosis rate for control = 0.017 ± 0.002; CK-666 = 0.014 ± 0.002; and SMIFH2 = 0.007 ± 0.001. (right) Endocytosis rate for control = 0.007 ± 0.002; CK-666 = 0.007 ± 0.002; and SMIFH2 = 0.001 ± 0.001. All values represent means ± SEM. (∼100–200 boutons on 7–11 coverslips were analyzed for each group from three separate batches of cultures; ***, P < 0.001; **, P < 0.01 compared with control by one-way analysis of variance followed by Dunnett’s post hoc test.)

Figure 9.

Model of axonal F-actin. (A) Interpretative illustration derived from data. Actin monomers nucleate and polymerize at discrete microscopic zones in axons (light gray vertical bar). The hotspots are likely sites where actin nucleates on stationary endosomes. (B) Working model showing both subplasmalemmal actin rings and intra-axonal actin filaments. Note that actin monomers nucleate and extend polymers along the long axis of the axon, generating a steady-state scenario where actin is equilibrated along the axon, allowing spatiotemporal control throughout the long processes, and facilitating delivery into presynaptic boutons.