Formin Activity and mDia1 Contribute to Maintain Axon Initial Segment Composition and Structure

The axon initial segment (AIS) is essential for maintaining neuronal polarity, modulating protein transport into the axon, and action potential generation. These functions are supported by a distinctive actin and microtubule cytoskeleton that controls axonal trafficking and maintains a high density of voltage-gated ion channels linked by scaffold proteins to the AIS cytoskeleton. However, our knowledge of the mechanisms and proteins involved in AIS cytoskeleton regulation to maintain or modulate AIS structure is limited. In this context, formins play a significant role in the modulation of actin and microtubules. We show that pharmacological inhibition of formins modifies AIS actin and microtubule characteristics in cultured hippocampal neurons, reducing F-actin density and decreasing microtubule acetylation. Moreover, formin inhibition diminishes sodium channels, ankyrinG and βIV-spectrin AIS density, and AIS length, in cultured neurons and brain slices, accompanied by decreased neuronal excitability. We show that genetic downregulation of the mDia1 formin by interference RNAs also decreases AIS protein density and shortens AIS length. The ankyrinG decrease and AIS shortening observed in pharmacologically inhibited neurons and neuron-expressing mDia1 shRNAs were impaired by HDAC6 downregulation or EB1-GFP expression, known to increase microtubule acetylation or stability. However, actin stabilization only partially prevented AIS shortening without affecting AIS protein density loss. These results suggest that mDia1 maintain AIS composition and length contributing to the stability of AIS microtubules. Supplementary Information The online version contains supplementary material available at 10.1007/s12035-021-02531-6.


Introduction
The axon initial segment (AIS) serves as a barrier that controls membrane protein diffusion and cytoplasmic transport towards the axon, maintaining neuronal polarity [1][2][3][4][5]. Simultaneously, the AIS is characterized by a high density of voltage-gated ion channels that allow action potential (AP) generation [6]. This high density of voltage-gated ion channels is achieved by a not completely understood protein complex formed by scaffold proteins, such as ankyrinG or βIV-spectrin, and a specialized actin and microtubule cytoskeleton [7]. The whole complex is characterized by a high stability and detergent extraction resistance [8], but has a surprising high degree of plasticity [9], which regulatory mechanisms at cytoskeleton level are relatively unknown. AnkyrinG is the main scaffold protein that anchors voltage-gated sodium channels [10,11] and is bound to actin cytoskeleton through βIV-spectrin [12]. Moreover, ankyrinG can also bind to microtubules through EB1/3 proteins [13].
AIS actin and microtubules contribute to control axonal traffic and AIS structure [3,5,[14][15][16][17][18]. However, our understanding of the mechanisms controlling AIS cytoskeleton dynamics is limited. AIS microtubules are highly fasciculated and contain post-translational modifications of tubulin, such as acetylation and detyrosination, which allow kinesin-1 to transport cargoes into the axon [19][20][21]. Changes in tubulin acetylation by the tubulin deacetylase HDAC6 contribute to modifications in AIS [21,22]. A higher phosphorylation of the microtubule-associated protein Tau, due to V337M mutation, contributes to AIS shortening [23]. Recently, TRIM46 (tripartite motif-containing protein 46) was identified as a microtubule fasciculation protein at the AIS [15], and the microtubule-associated protein MAP7D2 was also localized to the proximal axon promoting kinesin-1-mediated cargo transport [17]. Besides that, AIS actin cytoskeleton is necessary to maintain the cytoplasmic barrier that impairs somatodendritic cargoes to enter the axon [5,16,24]. Different actin structures have been described in the AIS, such as actin patches, actin rings, or actin filaments [25]. The AIS also contains an actin-rich structure, the cisternal organelle (CO), supposed to be a calcium regulatory compartment [26] and related to AIS plasticity [27]. Some AIS actin-related proteins have been identified, such as synaptopodin, α-actinin-2, or myosin II and V [14,24,28]. Despite our increasing knowledge of actin-and microtubule-related proteins in the AIS, the mechanisms that regulate the crosstalk between both cytoskeleton components at the AIS are not fully understood.
Some studies have shown a mutual regulation of actin and microtubule stability in axons [29,30]. The crosstalk between actin and microtubules has been well-studied during acquisition of neuronal cell shape or cell migration, modulated by shared regulators, such as MAPs, EBs, or formins [29]. Among them, EB1/EB3 and MAP7D2 are localized at the AIS [3,17], while only one previous study points to localization of formin 2 at the proximal axon [31]. Fifteen different formins have been described in mammals [32], defined by their most conserved formin homology 2 (FH2) domain [33], which is the core element for formin physiological functions [34]. The regulatory mechanisms of formins are not fully understood, but their activation and autoinhibition depend on different mechanisms [35,36]. Formins are a major class of effectors for cytoskeleton modulation, with the ability to modulate and coordinate actin and microtubule structures [34,37,38]. Among them, formins modulate actin filaments elongation and nucleation, and contribute to microtubule stabilization, promoting indirectly their acetylation [39]. Among their multiple physiological roles, formins are necessary to maintain apico-basal polarity in neuroepithelial and epithelial cells [40][41][42], and to regulate axonal actin structural dynamics [43]. Furthermore, formins are related to degenerative diseases and mental disorders [44].
Thus, we hypothesized that formins may play a role in the regulation of AIS structure. We found that formin inhibition modifies actin and microtubule AIS cytoskeleton and reduces scaffold and functional protein density at the AIS, both in cultured neurons and brain slices. This leads to a reduced intrinsic neuronal excitability. Moreover, formin activity has a dual role in microtubule or actin filament regulation leading to the maintenance of AIS length or AIS protein density. Finally, we have identified mDia1 formin as a main formin to maintain AIS protein density and AIS length.

Materials and Methods
Animals Animals, Swiss CD1 mice, were housed in the Experimental Animal facility at Cajal Institute and maintained according to national legislation (53/2013(53/ , BOE no. 1337) and guidelines of the Council of the European Communities (2010/63/UE). Protocols were previously approved by the CSIC bioethics committee.

Cell Cultures
Mouse hippocampal neurons were prepared as previously described [45,46]. Briefly, mice hippocampi were dissected, incubated in a 0.25% trypsin solution in Ca 2+ /Mg 2+ free Hank's buffered salt solution (HBSS) and dissociated using fire-polished Pasteur pipettes. Neurons were placed on poly-lysine-coated coverslips (1 mg/ml) at a density of 6000 cells/cm 2 for 2 h in a plating medium (MEM, 10% horse serum, 0.6% glucose and Glutamax-I). Then coverslips were inverted and transferred to culture dishes containing an astrocyte monolayer in a neuronal medium (Neurobasal, B27 supp l e m e n t a n d G l u t a m a x -I ) . A 5 μ M 1-β -Darabinofuranosylcytosine (AraC) was added 2 days after to curb glial cells proliferation. Neurons were maintained replacing one-third of neuronal medium every week. In the case of pharmacological treatments in the absence of glial cell layer, coverslips were transferred to plates containing glial cellconditioned medium.
Nucleofection was performed after neuronal dissociation using the P3 Primary Cell 4D-Nucleofector TM X Kit L for primary mammalian neural cells (Cat# V4XP-3012, Lonza), using the CL133 program, 2 × 10 6 cells and 3 μg of total DNA for each nucleofection. After nucleofection, neurons were plated at a density of 15.000 cells/cm 2 .
Neuronal transfection was performed after plating at the times described in the "Results" section. Hippocampal neurons (20,000 cells/cm 2 ) were transfected with Lipofectamine 2000 (Cat#11668-030, Invitrogen) and maintained for 24 h or 48 h.
Neuro-2a cells were cultured in DMEM containing 10% FBS and glutamine (2 mM). Cells were transfected using Lipofectamine 2000 (Invitrogen) following manufacturer instructions, and kept for 3 days before fixation and immunocytochemistry.
Whole-cell patch-clamp recordings were obtained from CA3 pyramidal neurons. The external solution contained (mM): 125 NaCl, 26 NaHCO3, 3 CaCl2, 2.5 KCl, 2 MgCl2, 0.8 NaH2PO4, and 10 D-glucose, and was equilibrated with 95% O2-5% CO 2 . Patch-pipettes (5-10 MΩ) were filled with a solution containing (mM) 120 potassium gluconate, 20 KCl, 0.5 EGTA, 10 HEPES, 2 Na2ATP, 0.3 NaGTP, and 2 MgCl2, pH 7.4. Recordings were made at 30°C in a temperature-controlled recording chamber (Luigs & Neumann). The voltage and current signals were acquired on a PC computer fed by a Digidata1440A at a sampling frequency of 10 kHz for input-output curves and 40 kHz for spike threshold measurements. CA3 neurons were identified by their firing behavior. Neurons were held at their resting membrane potential (~−77 mV). Input-output curves were established by 10 pA increments of current step, and spikes were counted. Rheobase and gain were measured for each neuron as previously reported [48]. Spike threshold was measured as the first value of voltage corresponding to a rate exceeding 30 mV/ms.

Immunohistochemistry
Brains from mice P26-P30 days old were quickly removed and acute coronal slices (300 μm) containing sensorimotor cortex were cut with a vibratome (4°C) in a solution containing: 234 mM sucrose, 11 mM glucose, 26 mM NaHCO 3 , 2.5 mM KCl, 1.25 mM NaH 2 PO 4 , 10 mM MgSO 4 , and 0.5 mM CaCl 2 (equilibrated with 95% O 2 -5% CO 2 ). SMIFH2 treatment was carried out at a concentration of 15 or 30 μM for 3 h, and the same volume of vehicle DMSO was used as control. Slices were fixed 30 min at RT with 4% PFA and washed in PBS. Slices were incubated for 2 h at RT with PBS containing 0.1% TritonX-100 and 10% goat serum, followed by 4°C overnight incubation with anti-AnkG antibody (IgG2a, 1:150) in PBS containing 0.1% TritonX-100 and 1% goat serum. After secondary antibody incubation (1:500) for 2 h at RT, sections were counterstained with bis-benzimide (5 μg/ml, Cat# B2261, Sigma-Aldrich) for nuclei staining and mounted in Fluoromount-G. Images were acquired on a Leica SP5 confocal microscope using Leica DM6000B 40× 1.25 N.A. oil objective. Sections were imaged in z stacks with a 0.5-μm step size. For qualitative analysis, a Z-projection was obtained in the Fiji-ImageJ software. We measured individual AISs in cortical layers II/III, IV, and V/VI. Cortical layers were distinguished according to nuclei patterns. In the case of CA1 staining quantification, total ankyrinG fluorescence was measured in the whole CA1 region as shown in figures.

Fluorescence Recovery After Photobleaching
For FRAP experiments, hippocampal neurons were plated directly on 35-mm culture dishes with glass bottom (Ibidi). Neurons were plated at a density of 20.000 cells/cm 2 and supporting glial cells were place on inverted 24-mm coverslips. Neurons were lipofected with AnkyrinG-GFP at 8 DIV and kept for 24 h. Prior to FRAP experiment, neurons were treated with 15 μM SMIFH2 or DMSO for 3 h. Live imaging was performed on a Leica SP5 inverted confocal microscope, using 63× 1.40 NA oil objective and maintaining 37°C and 5% CO 2 in the live cell chamber. Time-lapse images were acquired using bidirectional scan at 1400 Hz with the format of 256 × 256 pixels and the pinhole size was 2 airy units. AIS was distinguished by AnkyrinG-GFP concentration at the proximal axon. A ROI with an area of 5 μm × 5 μm on the proximal axon was photobleached at 100% 488-nm Laser Line power with 25% Argon Laser Power, reducing local fluorescence by 80-95%. FRAP was performed with 10 frames of 0.1-s intervals for pre-bleach and 15 s for bleach with 0.1-s intervals, followed by post-bleach with 240 frames of 0.5-s intervals. FRAP analysis was performed as previously described [13]. The recovery curve was generated by pooling together at least 15 neurons in each condition and averaging each frame the recovery rate. The maximum recovery percentage was obtained through averaging recovery data in 20 s at the plateau from every neuron in Sigmaplot v12.5 (Systat Software Inc., San Jose, CA, USA).

AIS Parameter Quantification
Confocal microscope settings were adjusted to prevent signal saturation and the images were taken in z stacks with a 0.5-μm step size. Images were obtained randomly. All images from every experiment were obtained using the same settings, standardized in the control samples of each experiment. For quantitative analysis, all coverslips in each experiment experienced the same procedures for labeling. AIS was identified by ankyrinG staining. A stack of 1.5-μm 3 confocal sections containing the AIS was used to quantify AIS parameters. Analysis was performed using the Fiji-ImageJ software (NIH) drawing a line along the axon, starting at the soma. Data were smoothed every 1 μm using the Sigma Plot 12.5 software. AIS start, maximum, and end positions were determined as described previously [50]. Total fluorescence intensity for each marker was obtained by adding fluorescence intensity values from start to end position after determining AIS length using ankyrinG staining. Data where normalized in each experiment to the value of the mean fluorescence in control neurons. For AIS profiles data were normalized to the maximum fluorescence value in control neurons.

Statistical Analysis
All statistical analyses were carried out in GraphPad Prism 5 and Sigmaplot v12.5. Sample Gaussian distributions were first assessed for normality using the D'Agostino and Pearson omnibus test or Shapiro-Wilk's test. Statistical analysis was performed using a two tailed t-test for two group comparisons: Unpaired t-test for parametric data and a post hoc Mann-Whitney test for non-parametric data. For multiple group comparisons, one-way analysis of variance with Tukey post hoc test was used for parametric data and a post hoc Kruskal-Wallis with Dunn's test for non-parametric data. Graphs were represented as the mean ± SEM. Differences were considered significant when p-value less than 0.05, and represented as *p < 0.05, **p < 0.01, *** < 0.001.

Formin Inhibition Modify AIS Protein Density
To determine the effects of formin activity on AIS structure and protein composition, we treated neurons with SMIFH2, an inhibitor of formin conserved FH2 domain [34,37,51]. Treatment with SMIFH2 (15 μM) in 14 DIV-cultured hippocampal neurons led to a significant ankyrinG decrease after 3 h of application (−25%, Fig. 1A, B) that was detected after 30 min (−15%, Fig. 1C). This decrease was more pronounced in young neurons (7 DIV, −32%), than in older neurons (21 DIV, −15%) after 3 h of application (Fig. 1B), and was due to a decrease all along the AIS (Fig. 1D). Moreover, a similar reduction was also detected in 14 DIV neurons for other AIS proteins (Fig. 1E), such as βIV-spectrin (−20%), voltage-gated sodium channels (PanNaCh, −20%), and phospho-myosin light chain (pMLC, −10%), suggesting an effect on the underlying AIS cytoskeleton (Fig. S1). To discard that SMIFH2 treatment was generating an indirect effect on neurons through formin inhibition in astrocytes, we applied SMIFH2 to neurons in the absence of astrocytes (Fig.  S2A) and detected again an ankyrinG reduction (−28.80 ± 1.89%) compared to DMSO-treated neurons. Moreover, after add itiona l i nc ubation i n new plates co ntaining astrocyte-conditioned neuronal medium for 6 h (Fig. S2B), a further decrease of 10% of ankyrinG intensity was observed. After 24 h, ankyrinG decrease showed an inverse correlation with increased expression of the somatodendritic marker MAP2 along the AIS, entering the axon (Fig. S2C-E). Finally, we analyzed whether ankyrinG decrease could be due to calpain activation as previously described [45,52]. Hippocampal neurons (14 DIV) were pre-treated with the calpain inhibitor MDL-28170 (50 nM) for 1 h, and then neurons were incubated with SMIFH2 (15 μM) or DMSO for 3 h (Fig. S3). Our results show that ankyrinG decrease is not impaired by calpain inhibition, and other mechanisms are responsible for AIS protein reduction.  In order to confirm these results, 10 DIV hippocampal neurons were transfected with a plasmid expressing ankyrinG-GFP for 48 h and treated with SMIFH2 for 3 h (Fig. 1F). Analysis of the GFP signal at the AIS confirmed the results obtained for endogenous ankyrinG. The ankyrinG-GFP signal was reduced by a 35% along the AIS (Fig. 1G, H). In addition, ankyrinG-GFP signal was analyzed "in vivo" in hippocampal neurons after SMIFH2 treatment (Fig. 1I, J). Fluorescence recovery after photobleaching (FRAP) experiments demonstrated that ankyrinG-GFP signal recovery after photobleaching was significantly lower after SMIFH2 treatment (36 ± 5% vs 55 ± 6% in control neurons, Fig. 1J).
Thus, we analyzed whether formin inhibition was modifying microtubule characteristics leading to AIS protein density reduction. Actually, formins also contribute to increase microtubule acetylation and stability [39,54]. Therefore, we quantified immunofluorescence intensity of acetylated-α-tubulin after detergent extraction in control and SMIFH2-treated neurons (Fig. 4A). SMIFH2-treated neurons showed a significant decrease in acetylated-α-tubulin staining (46 ± 8%) compared to 100 ± 4% in control neurons (Fig. 4B). AnkyrinG and βIV-spectrin signals were also significantly decreased (71 ± 4% and 56 ± 3% of the control level, respectively). Moreover, analysis of acetylated-α-tubulin vs total tubulin ratio without detergent extraction shows a significant reduction in the AIS (0.82 ± 0.02 vs 1.0 ± 0.02 in DMSO neurons) that was not found in proximal dendrites (Fig. 4C). Kinesin-1 selectively interacts with acetylated microtubules guiding this motor into the axon [19,55]. Thus, we analyzed KIF5C distribution after SMIFH2 treatment. KIF5C showed a significant increase in 5 DIV neurons soma (124 ± 4.62%) accompanied by a reduction in axons (89 ± 4%) compared to DMSO-treated neurons (Fig. 4D, E). We detected a 30% KIF5C reduction in the first 30 microns of the axon compared to that in control cells (Fig. 4F). The above results suggest a participation of formin activity on AIS protein maintenance through a mechanism that involves microtubules.

mDia1 Contributes to Maintain AIS Proteins
A previous study in neurons confirmed that mDia1 is the primary target of SMIFH2 activity on microtubules in hippocampal neurons [49]. Moreover, mDia1 is necessary to maintain neuroepithelial cell polarity [42] and modulates axon and dendrites structures [49,56], being related to microtubule stability and acetylation, and actin dynamics modulation [54]. Thus, we stained 14 DIV hippocampal neurons with 3 different mDia1 antibodies (see "Methods" section), one of them validated in mDia1 knockout cells (Abcam). The staining Fig. 2 Formin inhibition decreases ankyrinG intensity in brain slices and reduces neuronal intrinsic excitability. A Brain slices (300 μm) containing hippocampi from P26 mice were treated with DMSO or 30 μM SMIFH2 for 3 h. Brain slices were stained to visualize ankyrinG (green) and nuclei (magenta). Scale bar = 100 μm. B Quantification of corrected total fluorescence intensity (CTF) in CA1 regions. **p < 0.01, unpaired t test. Each dot represents a CA1 zone as shown in A. Analysis was done in 4 independent experiments using 8 animals, and a total of 26 (control) and 34 (SMIFH2) slices. C Normalized ankyrinG intensity in DMSO (black) or 15 μM (red) and 30 μM SMIFH2 (magenta) in AISs of layers II/III, layer IV, and layers V/VI of P26 mice coronal sensorimotor cortex slices (300 μm). ***p < 0.001, Kruskal-Wallis, Dunn's multiple comparison test were analyzed. Data are represented as the mean ± SEM and acquired in 150 AISs, for each condition and layer, from three independent experiments. D Representative images of ankyrinG-stained cortical sections treated with DMSO or SMIFH2. E CA3 neurons from rat hippocampal slice organotypic cultures treated for 3 h in the presence of 30 μM SMIFH2 (red) or DMSO alone (control, black) were recorded under current-clamp and the number of evoked action potentials plotted against injected current. Representative traces are shown in the bottom panel. F, G Spike threshold, dV/dt curves, and representative traces (inset = focus on spike threshold) in control and SMIFH2-treated slices. H, I Rheobase (H) and gain (I) in control (black) and SMIFH2 (red)-treated slices pattern of all three mDia1 antibodies was similar and not specifically enriched at the AIS, though mDia1 staining appeared at the AIS (Fig. S4A-C). All three antibodies stained mainly the soma with a Golgi and perinuclear staining, as well as dendrites and axons (Fig. S4A-C). We observed a similar pattern in hippocampal neurons expressing CAmDia1-GFP (Fig. S4D). To analyze the potential role of mDia1 on AIS modulation, we used two different mDia1 interference RNAs (shmDia1-1, shmDia1-2) that express RFP protein as reporter (Fig. S5). Interference RNAs were validated in Neuro2a cells by immunofluorescence using the mDia1 knockout-validated antibody (Abcam). Both interference RNAs decreased mDia1 signal by 70%, 3 days after transfection, compared to cells transfected with scrambled shRNA (Fig. S5A-D).
To analyze mDia1 shRNA effects on neurons, first we checked that mDia1 shRNA was not affecting axon and dendrite development and did not find differences between scrambled shRNA-and shmDia1-1-nucleofected neurons regarding  Fig. 4 Formin inhibition modifies microtubules in the AIS and axonal transport. A Images of acetylated-tubulin (green), ankyrinG (red) and βIV-spectrin (blue) in the AIS of 13 DIV control and SMIFH2-treated neurons after detergent extraction for 5 min before fixation. Scale bar = 10 μm. B Normalized fluorescence intensity of markers indicated in A. *p<0.05, ***p < 0.001, Mann-Whitney test. C Acetylated-α-tubulin/α-tubulin ratio in the AIS or proximal dendrites in control or 15 μM SMIFH2-treated neurons. D KIF5C distribution in control or 15 μM SMIFH2-treated 5 DIV neurons for 3 h. Insets show magnifications of KIF5C expression in soma. Scale bar = 100 μm. E KIF5C normalized fluorescence intensity in soma (circles) and axon (squares) in DMSO (black) or SMIFH2 (red) treated neurons. *p < 0.05, ***p < 0.001, Mann-Whitney test. F KIF5C normalized fluorescence intensity in proximal axon of 5 DIV neurons treated with 15 μM SMIFH2 or DMSO. *p < 0.05, Mann-Whitney test. Data were acquired from three independent experiments and represented as the mean ± SEM axon length at 3 DIV or dendrites length at 7DIV (Fig. S5G,  H). Next, we nucleofected hippocampal neurons with shmDia1-1, shmDia1-2, or srcambled shRNA plasmids for 10 DIV and analyzed mDia1 staining in the soma of hippocampal neurons, where it is more densely localized (Fig. S5E-F). We found that mDia1 signal was reduced around 45% (shmDia1-1) and 30% (shmDia1-2) compared to scrambled shRNA-nucleofected neurons (Fig. S5F). Therefore, we analyzed ankyrinG expression at the AIS of 10 DIV neurons nucleofected with scrambled shRNA or mDia1 shRNAs (Fig. 5A, B). Both shmDia1-1 and shmDia1-2 shRNAs reduced significantly ankyrinG expression (Fig. 5B) to 68 ± 2% or 78 ± 2%, respectively, compared to scrambled shRNA (100 ± 2%). The higher reduction produced by shmDia1-1 compared to that by shmDia1-2 was in agreement with the previously observed higher mDia1 signal decrease (Fig. S5F). The AnkyrinG intensity profile along the AIS showed that this decline was homogenous all along the AIS (Fig. 5C). Moreover, 10 DIV shmDia1-1-nucleofected neurons showed a significant fluorescence intensity decrease of voltage-gated sodium channels and βIV-spectrin signals (47 ± 6% and 33 ± 4%, respectively, Fig. 5D). Besides that, ankyrinG intensity in shmDia1-1-nucleofected neurons was not further reduced after SMIFH2 (15 μM) treatment for 3 h (Fig. 5E, F) compared to that after DMSO treatment (67 ± 3% vs 73 ± 3%, respectively). However, SMIFH2 produced a s i g n i f i c a n t a n k y r i n G r e d u c t i o n i n s c r a m b l e d shRNA-nucleofected neurons (68 ± 2% vs 100 ± 3% in DMSO-treated scrambled shRNA neurons), suggesting a SMIFH2 action through mDia1 inhibition. Next, we investigated whether mDia1 suppression in later developmental stages did also reproduce the ankyrinG reduction observed after SMIFH2 treatment. Neurons were transfected at 10 DIV with shmDia1-1 or scrambled shRNAs (Fig. 5G) and analyzed 2 days after. We found a significant ankyrinG reduction (72 ± 4%) in neurons transfected with shmDia1-1 plasmid compared to scrambled shRNA-transfected neurons (Fig. 5H). These data suggest that neuronal expression of mDia1 contributes to the maintenance of AIS proteins from early developmental to mature stages of AIS.

Formin Inhibition or mDia1 Suppression Shortens the AIS
Next, we analyzed how formin inhibition or mDia1 downregulation may produce changes on AIS length or position. We first analyzed AIS length in brain slices treated with SMIFH2 (30 μM) for 3 h (Fig. 7A, B). We found that AIS length was significantly different between cortical layers in P30 female mice (19.39 ± 0.41 μm in layers II/III, 21.49 ± 0.47 μm in layer IV and 23.04 ± 0.28 μm in layers V/VI) and analyzed this parameter by layer groups after SMIFH2 treatment (Fig.  7B). Formin inhibition produced a similar and significant AIS length reduction, around 12%, in each layer group (Fig. 7B). SMIFH2 (15 μM) treatment in 10 DIV hippocampal neurons for 3 h reduced 20% of the AIS length (21.15 ± 0.72 μm compared to 27.86 ± 0.79 μm in DMSO-treated neurons, Fig. 7C, D, E). Despite that actin stabilization did not recover ankyrinG levels, pre-incubation with jasplakinolide 1 h before SMIFH2 treatment partially prevented AIS shortening (Fig.  7F). We observed the same partial AIS shortening prevention in neurons expressing EB1-GFP after SMIFH2 treatment (Fig.  7G). Transfection of shmDia1-1 plasmid (Fig. 7H, I, J) in 10 DIV hippocampal neurons did also reduce a 20% the AIS  . Data in graphs were acquired from three independent experiments and represented as the mean ± SEM SMIFH2-treated neurons or neuron-transfected or nucleofected with shmDia1-1 plasmid (Fig. 7E, J, L) proving a distal shortening of the AIS. Finally, we analyzed whether shHDAC6 expression, shown to recover ankyrinG density (Fig. 6F), could prevent mDia1 shRNA-mediated AIS shorte n i n g ( F i g . 7 M ) . A I S s w e r e a g a i n s h o r t e r i n shmDia1-1-transfected neurons (28.25 ± 1.71 μm vs 34.41 ± 1.40 μm in scrambled shRNA neurons), while those expressing shHDAC6 and shmDia1-1 interference RNAs had AISs not significantly shorter (32.69 ± 1.53 μm). Neurons only expressing shHDAC6 had slightly but significantly shorter AISs (31.19 ± 1.05 μm). From our results, we propose a dual role of formins in AIS actin and microtubules, where microtubule stability/acetylation contributes to maintain mainly ankyrinG density and AIS length, while actin modulation is only involved in AIS length regulation (Fig. 7N).

Discussion
Axon initial segment plays an important role in the control of intrinsic neuronal excitability, maintains neuronal polarity, and is capable of a high and dynamic structural plasticity [9,59]. These important roles require not fully understood mechanisms that modulate AIS cytoskeleton. In our study, we hypothesized that formins, as important cytoskeleton regulators [34,39] that participate in different cellular mechanisms and cell polarity [60], may regulate AIS cytoskeleton contributing to its maintenance and plasticity.
We report that formin activity is necessary to maintain structural ankyrinG and βIV-spectrin protein or sodium channel density at the AIS. A short time inhibition of formins, using 15 μM SMIFH2 concentration lower than the IC 50 (28 μM) previously described in 3T3 fibroblasts [51], is enough to decrease ankyrinG density, both in vitro and ex vivo. We also detected a myosin II light chain phosphorylation (pMLC) decrease, previously identified in relation to AIS plasticity [14], and a recent study proposed a SMIFH2 effect on myosins [61] in filopodia adhesion. However, even high SMIFH2 concentrations up to 100 μM do not affect non-muscle myosin II phosphorylation [62], suggesting a decrease due to AIS structural protein modifications. In fact, although SMIFH2 (30 μM) effects are reversible upon washout on fibroblasts in a short time [51], we found that SMIFH2-induced modifications Scrsh-RFP nucleofected neurons were used as control to normalize data. n.s., not significant, *p < 0.05, ***p < 0.001, Kruskal-Wallis, Dunn's multiple comparison test. G AnkyrinG profile along the proximal axon of neurons shown in F. Data were acquired from three independent experiments and represented as the mean ± SEM. H Schematic representation summarizing the role of formins in the regulation of ankyrinG levels. Decreased formin activity leads to lower acetylation and stability that is counteracted by increasing microtubule acetylation through shHDAC6 or stability by EB1. High tubulin acetylation by shHDAC6 alone also decreases ankyrinG levels after 3 h are not reversible in neurons and led to a later loss of axonal polarity, as previously described after ankyrinG knockdown [1]. Polarity loss happens also in mDia-deficient mice that show a disrupted integrity of their neuroepithelium [42], while in epithelial cells, mDia1 knockdown do not affect the formation of tight junctions, but their protein content decreases and epithelial barrier is compromised [40] as observed in our study in neurons. The effects produced by SMIFH2 could be associated to different formins expressed in neurons, as it is a general formin inhibitor [51]. However, we show that two different mDia1 shRNAs also mimic SMIFH2-mediated ankyrinG, βIV-spectrin, and sodium s., not significant, *p < 0.05, **p < 0.01, ***p < 0.0001, Mann-Whitney test. N Schematic representation that summarizes the role of formins in the regulation of ankyrinG levels and AIS length. Reduced formin activity or mDia1 actin modifications only affect AIS length, while microtubule modifications reduce AIS length and ankyrinG levels channel density decrease, maintaining a correlation between mDia1 staining reduction in soma and ankyrinG decrease. These results are in line with previous results in neurons demonstrating that mDia1 is the primary target of SMIFH2 [49] and our results showing any further change in ankyrinG density after SMIFH2 treatment in neurons expressing mDia1 interference RNA.
Previous studies identified mDia1 in dendrites [49], Golgi, or axonal growth cones [44]. mDia1 antibodies used in this study, including one validated in knockout cells, detected mDia1 staining at the AIS. However, most mDia1 antibodies signal and CAmDia1-GFP signal were detected in soma, dendrites, and axons opening the possibility that SMIFH2 effects on AIS composition and structure may come from changes in these structures. Some studies suggest that higher AIS protein density correlates with a more complex dendritic arbor [63,64]. Previous studies have shown that mDia1 interference RNAs reduce dendritic ramification in 18 DIV hippocampal neurons [65], while a higher ramification index is observed in 21 DIV cortical neurons after 7 days without affecting spine density [49]. We did not find any significant change in dendrite length after 7 DIV in neurons nucleofected with mDia1 shRNAs. Thus, ankyrinG decrease found in our study does not seem to be related to changes in dendrites, although we cannot completely discard an effect due to presynaptic or postsynaptic activity changes that may affect AIS structure and composition. Indeed, mDia1 suppression do not alter spine density or modify mEPSC frequency or amplitude [49] in physiological conditions, but SMIFH2 (30 μM)-mediated formin inhibition dramatically disrupts synaptic recycling in hippocampal neurons [43], what can modify neuronal inputs and generate changes on AIS structure.
Our data show changes in AIS actin and microtubule cytoskeleton after formin inhibition. Indeed, formins are important regulators of actin nucleation and elongation [66], and regulate microtubule dynamics and stability [57,67]. We found that formin inhibition was able to disrupt actin filaments and cisternal organelle. Actin cytoskeleton stabilization by jasplakinolide did partially impaired cisternal organelle loss and AIS shortening, but did not impair ankyrinG density decrease. Thus, formin inhibition-mediated actin modifications seem to play only a minor role on AIS length, due to cisternal organelle alterations, previously shown to play a role in AIS length plasticity [27]. Moreover, actin rings do not change during AIS developmental reorganization and shortening [68]. In this sense, mDia1 modulates actin-independent mechanisms, such as the spine density recovery, after Aβ peptide treatment, by an actin-binding-deficient K994A mDia1 mutant [67]. We found that formin inhibition decreased tubulin acetylation in AIS microtubules, and not in proximal dendrites, which correlated with ankyrinG decrease. The AIS contains stable microtubules characterized by high detyrosinated and acetylated-tubulin levels compared to dendrites [19,21], which influences the binding and motility of kinesin-1. Indeed, mDia1 contributes to the control of microtubule dynamics and increases tubulin acetylation [54], while decreased mDia1 expression leads to lower detyrosinated tubulin levels [49]. This explains the reduced kinesin-1 localization to the axon and ankyrinG density after formin inhibition. However, ankyrinG density and AIS length remain to control levels when the tubulin deacetylase HDAC6 is inhibited or knockdown in SMIFH2-treated neurons or neurons expressing mDia1 shRNA. HDAC6 knockdown also decreases AIS length and ankyrinG density, as previously studied [21], to a lesser extent than mDia1 knockdown. A suitable explanation is that both decreased or increased tubulin acetylation modify AIS integrity and lower ankyrinG density. Formin inhibition reduces tubulin acetylation, but HDAC6 suppression increases tubulin acetylation and compensate ankyrinG loss (Fig. 6H). Actually, HDAC6 inhibition compensates transport deficits in Huntington's disease promoting kinesin-1 recruitment by increasing microtubule acetylation [69]. Previous studies showed that HDAC6 and mDia formins collaborate to modulate cytoskeleton [70]. In fact, HDAC6 cooperates with Dia1 in the formation of links between actin filaments and microtubules, affecting microtubule dynamics [71]. Besides, HDAC6 can physically associate to EB1 [72] that has been proposed to function downstream of mDia1 in microtubule stabilization [58] and can explain ankyrinG density and AIS length maintenance in neurons over-expressing EB1 when formins are inhibited. Moreover, EB1 not only participates in microtubule dynamics regulation [58,73], but also is a component of the AIS [3], binding ankyrinG to microtubules [74].
AnkyrinG serves as the anchoring scaffold for voltage-gated sodium channels which density also decreased. In effect, lack of formin activity led to a reduced intrinsic neuronal excitability compatible with a sodium channel density decrease. Despite that the spike number is decreased, spike amplitude was not modified, which can be explained by the higher sensitivity of spike frequency compared to the mild reduction in sodium channel density (~20%, Fig. 1E) [75] observed after formin inhibition. In addition, formin inhibition depolarized the spike threshold, both in hippocampal and in cortical neurons. Such spike threshold increase may result from the reduction in sodium channel density or AIS length decrease due to the disorganization of the AIS by a formin inhibitor. In fact, a lower spike threshold has been associated to sodium channel density increase during AIS development in nucleus magnocellularis (NM) neurons [68]. Otherwise, spike threshold depolarization happens after AIS shortening due to blast wave exposure without any change in protein density [76]. A recent work suggests a differential regulation of sodium channel accumulation and AIS cytoskeleton reorganization, which work synergistically to optimize neuronal output [68].
In conclusion, our study describes a role of formins, and specifically mDia1, in the maintenance of AIS protein density and AIS length because of its direct or indirect action on AIS microtubules, but to a lesser extent on AIS actin-related structures (Fig. 7N). Formin loss of function or mutation is a characteristic in several mental disorders, as well as in neurodegenerative diseases [44]. Further studies are necessary to understand which physiological or pathological stimulus can modulate neuronal cytoskeleton through formins or mDia1 at the AIS or other neuronal compartments in order to regulate AIS structural plasticity.
Acknowledgements We thank the people of the confocal microscopy facility at Cajal Institute for their help on image acquisition. We thank Dr Wandosell for critical reading of the manuscript.

Code Availability Not applicable
Author Contribution WZ, JJG, PM, MR, DD, and MC performed data analyses; WZ, PM, MC, DR, BA, and NB-G performed experiments. JJG., WZ, and DD contributed to experimental design and data analysis supervision. JJG wrote the paper. WZ, DD, and PM revised the manuscript.
Funding Open Access funding provided thanks to the CRUE-CSIC agreement with Springer Nature. The work was supported by a grant from Ministerio de Ciencia y Universidades (RTI2018-095156-B-100) to JJG and INSERM funding to DD. Wei Zhang was supported by a fellowship from China Scholarship Council (No.201506300085) and Beatriz Achon by a Master fellowship from Universidad Autónoma de Madrid.
Data Availability Data generated and analyzed during this study are included in the published article and its supplementary information files.

Declarations
Ethics Approval and Consent to Participate Animals, Swiss CD1 mice, were housed in the Experimental Animal facility at Cajal Institute and maintained according to national legislation (53/2013, BOE no. 1337) and guidelines of the Council of the European Communities (2010/63/ UE). All protocols and animal experiments were previously approved by the CSIC bioethics committee and Comunidad de Madrid (PROEX-358/ 15).

Consent for Publication Not applicable
Competing Interests The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.