The membrane periodic skeleton is an actomyosin network that regulates axonal diameter and conduction

Neurons have a membrane periodic skeleton (MPS) composed of actin rings interconnected by spectrin. Here, combining chemical and genetic gain- and loss-of-function assays, we show that in rat hippocampal neurons the MPS is an actomyosin network that controls axonal expansion and contraction. Using super-resolution microscopy, we analyzed the localization of axonal non-muscle myosin II (NMII). We show that active NMII light chains are colocalized with actin rings and organized in a circular periodic manner throughout the axon shaft. In contrast, NMII heavy chains are mostly positioned along the longitudinal axonal axis, being able to crosslink adjacent rings. NMII filaments can play contractile or scaffolding roles determined by their position relative to actin rings and activation state. We also show that MPS destabilization through NMII inactivation affects axonal electrophysiology, increasing action potential conduction velocity. In summary, our findings open new perspectives on axon diameter regulation, with important implications in neuronal biology.


Introduction
When considering an adult axon, its diameter can oscillate depending on organelle transport (Greenberg et al., 1990), neuronal activity (Fields, 2011), deformations generated by movement or degeneration. The mechanisms controlling axonal diameter throughout the neuronal lifetime remain however unclear. The mature axon shaft is supported by a submembraneous actin-spectrin networkthe membrane periodic skeleton (MPS)-composed of actin rings regularly spaced by spectrin tetramers approximately every 190 nm (Xu et al., 2013). Although its assembly and function are largely unknown, the MPS may provide mechanical support for the long thin structure of axons (Hammarlund et al., 2007). In the initial MPS model, each ring was hypothesized to be composed of actin filaments capped by the actin-binding protein adducin (Xu et al., 2013). Recently, combining platinum-replica electron and optical super-resolution microscopy, the MPS actin rings were shown to be made of two long, intertwined actin filaments (Vassilopoulos et al., 2019). According to this novel view, adducin might be responsible to enhance the lateral binding of spectrin to actin. We have previously demonstrated that adducin is required to maintain axon caliber as its absence in vitro leads to actin rings of increased diameter, while in vivo it results in progressive axon enlargement and degeneration (Leite et al., 2016). We have additionally found that in vitro, the radius of axonal actin ring narrows over time (Leite et al., 2016), supporting that the MPS has dynamic properties. Since reduction in axon diameter with time occurs both in WT and a-adducin knock-out (KO) neurons, MPS dynamics is probably regulated by additional actin-binding proteins.
The role of actin in the control of axonal radial tension is emerging (Costa et al., 2018;Fan et al., 2017). NMII is a hexamer composed by two heavy chains, two regulatory light chains (RLC) and two essential light chains (ELC), being a conserved molecule for generating mechanical forces (Vicente-Manzanares et al., 2009). The NMII contractile ATPase activity and the assembly of myosin filaments that coordinate force generation is activated by phosphorylation of myosin light chain (MLC) (Vicente-Manzanares et al., 2009). Here, we provide evidence that the axonal MPS, similarly to actin rings present in other biological contexts, is an actomyosin-II network that regulates circumferential axonal contractility. Furthermore, we demonstrate that the MPS affects signal propagation velocity, a property with important functional implications.

Results and discussion
Modulation of NMII activity regulates the expansion and contraction of axonal diameter The MPS of both WT and a-adducin KO neurons contracts in vitro at a rate of 6-12 nm/day (Leite et al., 2016). Given the general role of NMII in promoting contractility, we tested whether axon thinning in vitro was dependent on NMII activity. For that, NMII-mediated ATP hydrolysis and thereby actomyosin-based motility, were inhibited by blebbistatin (Straight et al., 2003; Figure 1A). In the presence of the drug, axon thinning of hippocampal neurons from DIV8 to DIV22 was abolished as determined using Stimulated Emission Depletion (STED) microscopy ( Figure 1B,C). This supports that axon thinning in vitro occurs through a NMII-mediated mechanism. Additionally, DIV8 hippocampal neurons treated with blebbistatin had a 1.3-fold increase in axon diameter ( Figure 1D,E). Alternative modes of drug-mediated modulation of myosin activity were tested, including ML-7 (Saitoh et al., 1987), calyculin A (Ishihara et al., 1989), and myovin1 (Gramlich and Klyachko, 2017;Islam et al., 2010). The function of NMII is controlled by MLC kinase (MLCK) that phosphorylates the NMII RLCs leading to conformational changes and self-assembly in myosin filaments (Vicente-Manzanares et al., 2009; Figure 1A). ML-7, a selective MLCK inhibitor that decreases pMLC levels in hippocampal neurons (Figure 1-figure supplement 1A, B), led to an increase in axonal diameter similar to that produced by blebbistatin ( Figure 1D,E). As protein phosphatase 1 (PP1) is the major myosin phosphatase responsible for dephosphorylation of NMII (Matsumura and Hartshorne, 2008), calyculin A, a potent PP1 inhibitor that results in increased phosphorylation of NMII RLCs and increased myosin contractility, was used (Iizuka et al., 1999; Figure 1A). In the presence of calyculin A, sustained activation of NMII resulted in a decrease to 0.8fold in axonal diameter ( Figure 1D,E), supporting that NMII activity can positively and negatively regulate axonal caliber. Of note, the effect of calyculin A was reverted by blebbistatin ( Figure 1D,E). Myovin-1, a potent inhibitor of the ATPase activity of myosin V that inhibits myosin V-dependent intersynaptic vesicle exchange in hippocampal neurons (Gramlich and Klyachko, 2017), had no effect on the MPS diameter ( Figure 1D,E) suggesting a NMII-specific effect. In further support of the MPS dynamic nature, the effect of blebbistatin was reverted as drug treatment was released from DIV8 to DIV12 ( Figure 1F,G). In summary, we show that axon diameter can be controlled both at the level of NMII activation through phosphorylation, and at the level of its power-stroke activity.
We next asked whether reducing both adducin levels and NMII activity might have a cumulative effect. To test this hypothesis, the MPS was examined in blebbistatin-treated hippocampal neurons (1 hr before fixation); n = 6-8 axons/condition and 20-87 rings/ axon. (C) Representative STED images of bII-spectrin immunostaining using a STAR 635P secondary antibody related to (B). (D) Axon diameter of DIV8 Figure 1 continued on next page where adducin was knocked-down ( Figure 1H,I and Figure 1-figure supplement 1C, D). Whereas inhibiting NMII activity or knocking down adducin led to a 1.3-fold increase in axonal caliber, being the latter case in agreement with our previous studies in adducin KO neurons (Leite et al., 2016), the combination of both treatments did not result in a cumulative effect ( Figure 1H,I).
In mammalian cells, different genes encode three different NMII heavy chain isoforms, NMHCIIA, NMHCIIB and NMHCIIC that determine the NMII isoform (NMIIA, NMIIB and NMIIC, respectively) (Vicente-Manzanares et al., 2009). Different NMII isoforms have unique kinetic properties and specific cellular functions (Vicente-Manzanares et al., 2009). To understand the role of different NMII isoforms in the MPS, we induced their down-regulation using specific shRNAs ( Figure 1J,K and Figure 1-figure supplement 1C-L). Downregulation of either NMIIA or NMIIB induced a 1.3-fold increase in axon caliber, similar to that induced by either blebbistatin or ML-7 treatment, whereas downregulation of NMIIC was without effect ( Figure 1J,K). These results are in agreement with previous studies highlighting the distinctive mechanochemical profile of NMIIC and lower ability in generating contractile force when compared with NMIIA and NMIIB isoforms (Billington et al., 2013). As such, knockdown of either NMIIA or NMIIB was sufficient to induce a comparable outcome as drug-mediated inhibition of all NMII isoforms, whereas NMIIC probably does not participate in NMII-mediated regulation of axon diameter. shRNA-mediated knockdown of NMII RLC (Wang et al., 2008) had a similar effect to the knockdown of myosin heavy chains, leading to MPS expansion ( Figure 1J,K). Myosin phosphatase target subunit 1 (MYPT1) is the regulatory subunit of myosin light chain phosphatase and is responsible for the dephosphorylation of MLC. The inactivation of its enzymatic activity increases NMII phosphorylation and consequently its activation (Vicente-Manzanares et al., 2009; Figure 1A). Knocking down MYPT1 (Figure 1 and Figure 1-figure supplement 1E, F, K, L) caused a decrease to 0.8-fold in axon diameter ( Figure 1J,K). Of note, rescue experiments performed using co-transfection with shRNA-resistant constructs reverted the phenotype of the corresponding shRNA, confirming their specificity ( Figure 1J,K). Our data support that both NMIIA and NMIIB, contribute to the MPS actomyosin network.
The possible tilting of actin rings in relation to the axonal axis was measured in all the conditions where drug-or shRNA-mediated modulation of NMII activity was performed. No variations in the angle of actin rings in relation to the axonal axis were found ( Figure 1L,M) suggesting that NMII probably does not contract between adjacent rings. Of note, tilting may be hindered by the rat hippocampal neurons incubated with drug modulators of myosin activity: 3 mM blebbistatin (Bleb), 4 mM ML-7, 5 nM calyculin A (CalA), 5 nM calyculin A + 3 mM blebbistation (CalA+Bleb), 4 mM myovin1 (Myo1) and the respective vehicles (veh). Calyculin A was added at DIV8 (25 min before fixation), while the remaining drugs were added at DIV3 and DIV8 1 hr before fixation; n = 8-13 axons/condition and 10-77 rings/axon. (E) Representative STED images of bII-spectrin immunostaining using a STAR 635P secondary antibody related to (D). (F) Axon diameter of DIV8 rat hippocampal neurons incubated with either 3 mM blebbistatin (Bleb) or vehicle (veh) at DIV3 and DIV8 (1 hr before fixation). In cells treated with blebbistatin up to DIV8, drug was replaced by either vehicle (Bleb+veh) or drug treatment was continued (Bleb+Bleb) and axonal diameter was analysed at DIV12; n = 7-10 axons/condition and 12-79 rings/axon. (G) Representative STED images of bII-spectrin immunostaining using a STAR 635P secondary antibody related to (F). (H) Axon diameter of DIV8 rat hippocampal neurons upon shRNA-mediated knockdown of adducin (sh Adducin) or transfection using a control shRNA (CTR) and subsequent incubation with either Bleb or veh; T-transfected; NT-non-transfected; n = 4-13 axons/ condition and n = 12-88 rings/axon. (I) Representative STED images of bII-spectrin immunostaining using a STAR 635P secondary antibody related to (H). (J) Axon diameter of DIV8 rat hippocampal neurons upon shRNA-mediated knockdown of different NMII isoforms (NMIIA, NMIIB and NMIIC), RLC or MYPT1; scramble shRNA was used as control (CTR); shRNA resistant constructs (R) for each specific shRNA were used to rescue their effect; Ttransfected; NT-non-transfected; n = 5-15 axons/condition and 6-103 rings/axon. (K) Representative STED images of bII-spectrin immunostaining using a STAR 635P secondary antibody related to (J). (L) Tilting (effective angle a eff ) of the actin rings in relation to the axonal axis after incubation with NMII chemical modulators; similar concentrations and incubation times as defined in (D); n = 8-14 axons/condition and 14-72 rings/axon and (M) after shRNA-mediated downregulation; n = 5-15 axons/condition and 7-70 rings/axon. Tilting was quantified in STED images after bII-spectrin immunostaining. (N) MPS periodicity after incubation with myosin-targeting drugs; similar concentrations and incubation times as defined in (D). The average distance between peaks was quantified by STED after bII-spectrin immunostaining; n = 8-14 axons/condition and 14-72 rings/axon. (O) MPS periodicity after shRNA-mediated knockdown of different NMII isoforms (NMIIA, NMIIB and NMIIC), RLC or MYPT1; a scramble shRNA was used as control (CTR). The average distance between peaks was quantified by STED after bII-spectrin immunostaining; T-transfected; NT-non-transfected; n = 5-15 axons/condition and 7-70 rings/axon. In all panels: Scale bars, 200 nm. *p<0.05; **p<0.01; ***<0.001; ****p<0.0001; ns-non significant. Graphs show mean ± s.e.m. In every case displayed in this Figure  existence of the bII-spectrin scaffold. Also, having several active NMII filaments attached to adjacent axon rings, may generate a net force that would be averaged out and thus close to zero. Importantly, neither drug-nor shRNA-mediated modulation of NMII activity, resulted in alterations of MPS periodicity ( Figure 1N,O, respectively), thus arguing that a possible participation of NMII in axonal longitudinal tension does not interfere with the 190 nm length of extended spectrin tetramers (Brown et al., 2015).

Propagation velocity is altered by manipulation of NMII activity
Neuronal electrophysiology is intimately related to morphology. Besides the relation between conduction velocity and axonal caliber (Hodgkin, 1954;Waxman, 1980), sodium channels are distributed in axons in a periodic pattern coordinated with the MPS (Xu et al., 2013). To uncover possible electrophysiological implications of manipulating NMII activity, we used a combination of in vitro microelectrode arrays (MEA) and microfluidics (mEF platforms) to characterize axonal propagation velocity between DIV11 and DIV14. This mEF platform was previously shown by us to allow axon alignment on top of a sequence of microelectrodes, providing detailed electrophysiological information about signal conduction on isolated axons ( Figure 2A) while also recording global activity levels (Heiney et al., 2019;Lopes et al., 2018). Blebbistatin-treated and untreated neurons were electrophysiologically competent regarding signal conduction and no significant differences in global electrical activity were found, as determined by assessing their mean firing rate (MFR) ( Figure 2B,C). However, axonal conduction velocity between treated and untreated axons was different, with distinct velocity distributions ( Figure 2D,E). From DIV11 on, blebbistatin-treated neurons consistently showed higher propagation velocities when compared to vehicle-treated cells. In unmyelinated axons, velocity conduction is proportional to the square root of the axonal diameter (Hodgkin, 1954) and blebbistatin treatment produces a 1.3-fold increase in axon diameter that is not altered throughout time ( Figure 1F). Our data favors that NMII modulation affects axonal electrophysiology through changes in axonal diameter. This effect should be further explored by using additional modulators of NMII activity and shRNA-mediated downregulation approaches.
Phosphorylated NMII light chains are colocalized with actin within the MPS and organized as circular periodic structures persisting throughout the axon shaft We next analyzed the localization of NMII in axons using STED microscopy, and 3D Single Molecule Localization Microscopy (SMLM) with a dSTORM/GSDIM-protocol (Fö lling et al., 2008;Heilemann et al., 2008). pMLC has been described to be specifically localized in the axon initial segment (AIS) (Berger et al., 2018). pMLC regulates the head groups of the NMII heavy chains, mediating their binding to actin and the organization of NMII into bipolar filaments (Berger et al., 2018). Despite the enrichment of pMLC at the AIS, not only non-phosphorylated MLC, but also active pMLC could be observed throughout the entire axon shaft ( Figure 3A). By applying SMLM to the AIS of DIV8 hippocampal neurons and focusing on their lateral outermost points up to the axonal surface, pMLC displayed a bilateral~200 nm periodicity and a circular structure (Figure 3B,C; Video 1). Periodic (~190 nm) pMLC staining, coincident with phalloidin staining restricted to the AIS, has been previously shown using STORM (Berger et al., 2018), suggesting a direct binding of pMLC to actin. Here, we show that pMLC periodicity is not restricted to the AIS, extending to the axon shaft, being visible in the lateral outer part of the axon ( Figure 3D-highlighted by white ruler; and 3E). Interestingly, in some axonal regions, pMLC showed a periodic distribution, consistent with anchoring in different positions of adjacent actin rings ( Figure 3D-highlighted by red ruler; and Figure 3F). The absence of a generalized striped pattern for pMLC (in contrast to what is observed for axonal actin and bII-spectrin staining), supports that in each actin ring a limited number of pMLC molecules is anchored. Alternatively, technical limitations of the antibodies used may preclude the visualization of the entire pool of pMLC molecules present in the structure. However, in some regions of the axon shaft of hippocampal neurons, pMLC not only displays the expected striped pattern but also is colocalized with actin, supporting that throughout the axon, pMLC is bound to actin rings that compose the MPS ( Figure 3G). Simultaneous analysis of bII-spectrin and pMLC using dual color SMLM showed that in the axon shaft the two molecules intercalate ( Figure 3H, arrowheads highlight pMLC and asterisks bII-spectrin), whereas a periodic distribution of pMLC consistent with anchoring in different positions of adjacent actin rings also occurs, in both cases non-overlapping with bII spectrin ( Figure 3H).
NMII heavy chains are present along the longitudinal axonal axis, being able to crosslink adjacent rings To characterize the spatial location of NMII heavy chains, we used antibodies that recognize the C-terminal tail region of NMIIA and NMIIB heavy chains, which label the middle portion of NMII bipolar filaments ( Figure 4A). Using these antibodies, we started by performing two-color STED in combination with phalloidin staining. When focusing the axonal lateral outermost points, both NMII heavy chain A ( Figure 4B) and NMII heavy chain B ( Figure 4C) were mainly detected within the axon core. At discrete sites, NMIIA and NMIIB colocalized with phalloidin ( Figure 4B,C-highlighted with arrowheads), suggesting that NMII heavy chains might additionally be within individual actin rings. Although no simultaneous labeling of NMIIA and NMIIB was performed given host antibody constraints, their localization within axons did not differ significantly ( Figure 4B,C). When observing axons single-labeled for NMIIA using SMLM, a pattern emerged suggesting the existence of multiple NMIIA heavy chain filaments along the longitudinal axonal axis ( Figure 4D). Of note, this organization clearly contrasted with the circular periodic distribution of pMLC ( Figure 3B; Video 1).
To further assess the localization of NMII heavy chains in axons, the SH-SY5Y neuroblastoma cell line and primary hippocampal neurons were transfected with fluorescent NMII fusion constructs that allow visualizing simultaneously the N-terminal (eGFP tag) and C-terminal (mApple tag) domains of the NMIIA heavy chain (Beach et al., 2014; Figure 4E). Of note, in differentiated SH-SY5Y cells, the approximately 190 nm MPS periodic pattern of bII-spectrin was observed ( Figure 4F,G). We used super- hippocampal neuron immunostained against pMLC (red) and stained for actin (gray), using a STAR 580 secondary antibody and phalloidin 635, respectively. pMLC molecules, highlighted with arrowheads, co-localize with actin. Scale bar: 200 nm. The raw image was deconvolved using the CMLE algorithm (Huygens Professional, Scientific Volume Imaging). (H) Z projection of 3D two-colour SMLM of a DIV8 hippocampal neuron immunostained against bII-spectrin (red) and pMLC (green) using secondary antibodies labeled with Alexa Fluor 532 and 647, respectively. Molecules intercalating at outermost positions are highlighted using arrowheads (pMLC) and asterisks (bII-spectrin). Scale bar: 500 nm. resolution spinning disk microscopy to study differentiated SH-SY5Y ( Figure 4H) and primary hippocampal neurons ( Figure 4I). We show that in axons of both cell types, NMIIA can assemble into bipolar filaments of~300 nm in length, consisting of eGFP puncta at the ends of the filament (N-terminus of head domains) with a single mApple punctum (C-terminus of tail domains) in the middle ( Figure 4H,I). Similarly to the data obtained by STORM ( Figure 4D), fluorescent NMII fusion proteins revealed the existence of multiple consecutive myosin filaments positioned along the axonal axis ( Figure 4H,I).
Dual color SMLM for the simultaneous analysis of bII-spectrin and NMIIA, further suggested that NMII heavy chain A is mainly positioned within the axon core forming several filaments with different orientations in relation to the axon shaft ( Figure 4J; Video 2). The analysis of the NMIIA labeling envelope, that is, the smooth curve outlining the extremes of NMIIA labeling, showed variable orientations with respect to the axonal longitudinal axis ( Figure 4K). This supports that NMIIA filaments are not strictly aligned with the axonal axis. 3D reconstructions were performed to quantify spatial properties ( Figure 4L-N). SMLM showed sites of colocalization of NMIIA heavy chain and bII-spectrin ( Figure 4N), which suggests the existence of NMII filaments crosslinking adjacent rings ( Figure 4Q, model a). Recently, immunogold labeling of pMLC followed by platinum-replica electron microscopy showed gold beads along filaments perpendicular to actin rings (Vassilopoulos et al., 2019). This led to the hypothesis that these filaments might correspond to myosins associated with the MPS, cross-linking neighboring rings, which goes along our observations. One should however note that in our analysis, some of the NMIIA staining was non-coincident with bII-spectrin, thus pointing towards a possible NMIIA filament distribution within actin rings ( Figure 4Q, model b). As expected, bII-spectrin showed a prominent peak in the spatial frequency domain, corresponding to a 197 nm periodicity ( Figure 4O, highlighted with #). Interestingly, a second pronounced peak at 1.71 mm was also observed ( Figure 4O, highlighted with *), suggesting a possible secondary structure arrangement of bII-spectrin in axons, that should be the subject of further analysis.

Structural organization and dynamics of actomyosin axonal rings
Our findings highlight possible distinct spatial positions of NMII filaments with respect to the MPS, which correlate with different biomechanical roles. NMII filaments crosslinking adjacent actin rings ( Figure 4P, model a) are not expected to provide for radial contractility but provide for scaffolding. NMII filaments within individual actin rings ( Figure 4P, model b), represent the conformation capable of generating the highest contractile force leading to actin filament sliding along the ring. Assuming that the MPS actin rings are composed of two parallel, intertwined for NMII heavy chain A using an Alexa Fluor 647 labeled secondary antibody. NMIIA labeling following different orientations in relation to the axonal axis is highlighted with arrowheads. Scale bar: 500 nm. (E) Representation of a bipolar NMIIA filament, with the N-terminal eGFP tagged and the C-terminal mApple tagged. (F) SH-SY5Y was immunolabelled with bII-spectrin using a STAR 635P secondary and imaged using STED. Scale bar: 200 nm. The raw image was deconvolved using the CMLE algorithm (Huygens Professional, SVI). (G) Analysis of SH-SY5Y bII-spectrin periodicity related to (F). (H) Representative spinning disk image of SH-SY5Y after co-transfection with eGFP-NMIIA and NMIIA-mApple. Scale bar: 1 mm. Insets highlight bipolar NMIIA filaments of~300 nm. (I) Representative spinning disk image of a primary hippocampal neuron axon after co-transfection with eGFP-NMIIA and NMIIA-mApple. Scale bar: 1 mm. (J) Z projection of a 3D SMLM double stained for bII-spectrin (red) and NMII heavy chain A (green) using anti-mouse Alexa Fluor 532 and anti-rabbit Alexa Fluor 647, respectively. Scale bar: 500 nm. (K) Analysis of the envelope of NMIIA labeling (Z projection) relative to the region highlighted by the white box in (J); the axonal membrane (dashed line) and centerline (solid line) are depicted. Scale bar: 1 mm. (L) Computer 3D reconstruction of the SMLM double stained for bII-spectrin (red) and NMII heavy chain A (green) of the image shown in (J). Solid line follows the axon centerline, whereas the dashed line marks the cellular membrane. The red scale bar is 1.7 mm, which corresponds to the size of the bII-spectrin secondary structure (see panel O). A Z projection is shown. (M) Y projection related to (J). (N) Co-localization of bII-spectrin and NMII heavy chain A for the reconstruction shown in (L). (Q) Fluorescence intensity spatial frequency of bII-spectrin on the axonal axis, analyzed by Fourier transform. In addition to the 0.20 mm peak (marked with '#"), there is also a consistent peak at 1.7 mm (marked with '*"). (P) Models for the distribution of the NMII along axons. NMII filaments may crosslink adjacent actin rings (model a) or span individual rings (model b), in both cases with variable angles relative to the axonal axis.
https://elifesciences.org/articles/55471#video2 actin filaments (Vassilopoulos et al., 2019), NMII activity could provide for contraction by putting the two actin filaments in registry with each other. One should however note that from the images presented (Vassilopoulos et al., 2019), a non-uniform distribution of NMII along the actin structures is observed which, to our understanding, cannot generate significant contractile force but rather hold a crosslinking function.
It is possible that in a single ring, NMII filaments with scaffolding and contractile roles might coexist. Additionally, scaffolding and contractile roles may switch via change in activity of the NMII motors. Pharmacological NMII inhibition may lead to ring expansion given loss of both scaffolding and contractile roles; as a consequence, the inner axonal tension will lead to increased axon diameter. When axonal diameter is restored, contractile NMII filaments may perform ring contraction, favored by transient absence of scaffolding filaments. Eventually, scaffolding NMII may be restored and contraction may slow down to its original low rate. In physiological responses (such as axon diameter increase upon cargo trafficking) at an initial stage, scaffolding NMII may be reduced to a minimum, while contractile filaments cannot prevent ring expansion. At a second stage (when cargo moves), contractile NMII may lead to a fast return to the original diameter. Eventually, scaffolding filaments may be restored and the contraction rate will return to a very low value. One should however bear in mind that despite the fact that NMII activity modulates MPS diameter, and that pMLC colocalizes with MPS actin both in the AIS and in the axon shaft, the interplay between NMII and deep axonal actin filaments may also occur. In this scenario, the possibility that these alternative axonal actin arrangements may also serve as myosin anchors cannot be ruled out and should be certainly explored in the future. In summary, our data supports that NMII regulates circumferential axonal contraction and expansion. These findings have important implications on our understanding of neuronal biology, including fluctuations in axonal diameter observed during trafficking, action potential firing and axon degeneration. Continued on next page

Immunolabeling
Primary hippocampal neurons were fixed with 4% PFA, in PBS at pH 7.4 for 20 min at room temperature. Fixed cells were permeabilized with 0.1% (v/v) triton X-100 (in PBS) for 5 min and autofluorescence was quenched with 0.2M ammonium chloride (Merck, cat# 1.01145.0500). Non-specific labeling was blocked by incubation with blocking buffer (5% FBS in PBS) for 1 hr. Primary antibodies diluted in blocking buffer were incubated overnight at 4˚C: mouse anti-bII-spectrin ( ImmunoResearch,1:200). After three 5 min PBS washes, coverslips were mounted in 80% glycerol for STED microscopy; for SMLM imaging, coverslips were mounted in GLOX/MEA buffer (detailed below) in depression slides and sealed with twinsil (Picodent, cat# 13001000) and for immunofluorescence, the coverslips were mounted in fluoroshield with DAPI (Vector Laboratories, cat# H-1200) and sealed with nail varnish. In the case of shRNA validations, images were acquired using a Zeiss Axio ImagerZ1 widefield microscope (Carl Zeiss) equipped with oil-immersion 63x/1.4 (Plan-Apochromat) and with Differential Interference Contrast (DIC) and a TCS Leica SP8 confocal microscope (Leica Microsystems). To validate the effectiveness of shRNA-mediated downregulation of NMIIA, NMIIB, NMIIC, RLC and MYPT1, the cell body from transfected and non-transfected cells, was delineated with the segmented line tool from Fiji, and the mean fluorescence intensity was measured. To confirm that ML-7 inhibits the phosphorylation of NMII, the AIS was delineated with the segmented line from Fiji and the mean fluorescence intensity was compared in treated and nontreated cells.

STED imaging
STED imaging was performed on an inverted Leica TCS SP8 STED 3X (Leica Microsystems), using DIV8 hippocampal neurons, unless otherwise indicated. Hippocampal neurons were imaged, at a fixed distance of 80-100 mm from the cell body, with a HC PLAPO CS2 100x NA 1.4 STED WHITE oil immersion objective (Leica Microsystems) using confocal and STED modes. The 2D vortex STED images with lateral resolution enhancement were recorded with 20 nm pixel size in xy and dwell times of typically 600 ns. First, the STED far-red channel (Abberior STAR 635P) was recorded with 633 nm excitation using the pulsed white light laser with 80 MHz repetition rate and STED depletion was performed with a synchronized pulsed 775 nm depletion laser. The detection bandpass was set to 650 to 750 nm and the pinhole was set to 0.93AU. The following acquisition settings were applied: 16 x line averaging and detector gating on a Hybrid Detector (HyD, Leica Microsystems) of 0.3 ns to 6 ns. The second STED channel (Abberior STAR580) was recorded in line sequential mode with 561 nm excitation and 775 nm depletion using a detection window from 580 to 620 nm. All other settings remained constant. We alternatively used an Abberior Instruments 'Expert Line' gated-STED coupled to a Nikon Ti microscope with an oil-immersion 60x 1.4 NA Plan-Apo objective (Nikon, Lambda Series) and a pinhole size set at 0.8 Airy units. The system features 40 MHz modulated excitation (405, 488, 560 and 640 nm) and depletion (775 nm) lasers. The microscope's detectors are avalanche photodiode detectors (APDs) which were used to gate the detection between~700 ps and 8ns. To analyze ring periodicity, the maximum intensity of peaks was determined and the interpeak distance was measured. To determine axon diameter, the distance between the outer points (brighter, in the focus plane) that formed the MPS was determined. Only axons unequivocally focused in the maximum wide plan were considered. Under distinct control conditions axon diameter varied, which is probably inherent to the different cultures used throughout the study. Given the use of specific controls in each experimental setting, this variation did not interfere with the interpretation of results. The tilting of actin rings was determined by measuring the angle of each actin ring regarding the axonal axis using the angle tool from Fiji. The tilting angle a was measured relative to the longitudinal axon axis. Angles larger than 90˚were mirrored to the first quadrant to yield an effective angle a eff =90-|90-a|.

SMLM imaging with dSTORM/GSDIM
For super-resolution SMLM-imaging with the Leica SR GSD using the dSTORM/GSDIM protocol, 18 mm coverslips (50000 cells/slide) were stored in PBS after fixation and immunolabelling at 4˚C. The coverslips were mounted onto a single depression slide (76 mm Â26 mm) and the cavity filled with 90-100 ml GLOX-MEA buffer (0.5 mg/ml glucose oxidase (Sigma-Aldrich, cat# G7141, 40 mg/ml catalase (Sigma-Aldrich, cat# 02071) 10% w/v glucose (Sigma-Aldrich, cat# 49163), 50 mM Tris-HCl pH 8.0, 10 mM NaCl and 10 mM b-mercaptoethylamine (Sigma-Aldrich, cat# M9768-5G)). The buffer was freshly prepared before imaging. Imaging was performed with a Leica SR GSD system using a HC PL APO 160Â/NA 1.43 oil objective. The images were recorded with an Andor iXon 897 EMCCD camera at 40 Hz using a central 180 pixel x 180 pixel subregion. For excitation, a 532 nm laser (500 mW maximum power output) and a 642 nm laser (500 mW maximum power output) were used and attenuated using an AOTF when appropriate. The two fluorophores were recorded sequentially and image acquisition, single molecule analysis and image reconstruction was performed with Leica LAS X 1.9.0.13747.

Spinning disk imaging
Differentiated SH-SY5Y cells (6500 cells/well) and DIV6-hippocampal neurons (50000 cells/well) were co-transfected with CMV-eGFP-NMIIA (Addgene, cat# 11347) and NMIIA-mApple (a kind gift from Dr John Hammer) using 1 mg:1 mg of each construct/well and Lipofectamine 3000 following the manufacturer's instructions. Two days later, in the case of SH-SY5Y cells, and four days later (at DIV10) in the case of primary hippocampal neurons, the cells were fixed. Transfected cells were then imaged using an Olympus SpinSR10 spinning disk confocal super-resolution microscope (Olympus, Tokyo, Japan) equipped with an PlanAPON 60 Â/1.42 NA oil objective (Olympus), a CSU-W1 SoRa-Unit (Yokogawa, Tokyo, Japan) with 3.2x magnification and ORCA-Flash 4.0 V3 Digital CMOS Camera (Hamamatsu, Hamamatsu City, Japan).

Preparation of microelectrode-microfluidic devices and electrophysiology recordings
Custom designed mEF devices were prepared following Lopes et al. (2018). Briefly, coated MEA chips (MultiChannel Systems MCS GmbH, Germany), with 252 recording electrodes of 30 mm in diameter and a center-to-center inter-electrode spacing of 100 mm, were combined with polydimethylsiloxane (PDMS) microfluidic chambers with an appropriate microgroove spacing for compartmentalization and monitoring of axonal activity. MEAs were coated with 0.01 mg/ml of poly-D-lysine (PDL, Corning) overnight at 37˚C, and then washed with sterile water. Microfluidic devices were sterilized with 70% ethanol and were gently attached to PDL-coated MEAs, creating a mEF chamber composed of two separate compartments connected by 700 mm length Â9.6 mm height Â10 mm width microchannels. The medium reservoirs were loaded with 150 ml of 5 mg/ml laminin isolated from mouse Engelbreth-Holm-Swarm sarcoma (Sigma-Aldrich Co.) and incubated overnight at 37˚C. The unbound laminin-1 was removed, and chambers were refilled with Neurobasal medium and left to equilibrate for at least 2 hr at 37˚C before cell seeding. Hippocampal neurons at DIVs 11, 12 and 14 were used in the electrophysiology experiments, where either blebbistatin (3 mM), or vehicle were added. Recordings were performed using a MEA2100 recording system (MultiChannel Systems MCS GmbH, Germany). The mEF devices prepared with these MEAs had 16 microchannels with 7 electrodes positioned along each microchannel, as well as 126 electrodes dedicated to the somal compartment. For each time point, recordings were obtained at a sampling rate of 20 kHz for the characterization of the overall network activity. Then, high-temporal resolution recordings were obtained at a sampling rate of 50 kHz, for a duration of 60 or 120 s, for the calculation of the conduction velocity. Throughout the experiments, the temperature was maintained at 37˚C and all recorded activity was spontaneous activity. Data analysis was carried out in MATLAB R2018a (The MathWorks Inc) using custom scripts (available in GitHub at: https://github.com/paulodecastroaguiar/Calculate_APs_velocities_in_MEAs; copy archived at https://github.com/elifesciences-publications/Calculate_APs_velocities_in_MEAs; Aguiar, 2020) and the mSpikeHunter tool (Heiney et al., 2019). Raw signals were band-pass filtered (200-3000 Hz) and spikes were detected by a threshold set to 6 Â STD of the electrode noise. Electrodes with a mean firing rate (MFR) of at least 0.1 Hz were considered active. For the propagation velocity calculations, the extracted spike times were further corrected based on the voltage waveforms. To be considered a propagating event the following requirements had to be fulfilled: detection over the entire microchannel (7 electrodes); time delay between electrode pairs lower than or equal to 1 ms (minimum propagation velocity of 0.1 m/ s); isolated spike in a 3 ms time window (as to ensure spike identity in all electrodes). Propagation velocity was then calculated by dividing the first-to-last electrode distance (600 mm span) by the delay between spike times. This stringent detection method eliminated any ambiguity during bursts and excluded sequences with missing spike times on at least one electrode, which drastically reduced the size but strengthened the quality of the action potentials dataset.

Statistical analysis
All measurements were performed with the researcher blinded to the experimental condition. Data are shown as mean ± s.e.m, which the exception of propagation velocity values which are shown as median ± s.d. Statistical significance was determined by Student's t-test using Prism (GraphPad Software), with exception of actin ring measurements in hippocampal neuron cultures, where one-way ANOVA was used (GraphPad Software). Sample sizes are indicated in Figure legends and significance was defined as p*<0.05, p**<0.01, p***<0.001, p****<0.0001, ns -not significant. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.