Neurofilament depletion improves microtubule dynamics via modulation of Stat3/stathmin signaling

In neurons, microtubules form a dense array within axons, and the stability and function of this microtubule network is modulated by neurofilaments. Accumulation of neurofilaments has been observed in several forms of neurodegenerative diseases, but the mechanisms how elevated neurofilament levels destabilize axons are unknown so far. Here, we show that increased neurofilament expression in motor nerves of pmn mutant mice, a model of motoneuron disease, causes disturbed microtubule dynamics. The disease is caused by a point mutation in the tubulin-specific chaperone E (Tbce) gene, leading to an exchange of the most C-terminal amino acid tryptophan to glycine. As a consequence, the TBCE protein becomes instable which then results in destabilization of axonal microtubules and defects in axonal transport, in particular in motoneurons. Depletion of neurofilament increases the number and regrowth of microtubules in pmn mutant motoneurons and restores axon elongation. This effect is mediated by interaction of neurofilament with the stathmin complex. Accumulating neurofilaments associate with stathmin in axons of pmn mutant motoneurons. Depletion of neurofilament by Nefl knockout increases Stat3–stathmin interaction and stabilizes the microtubules in pmn mutant motoneurons. Consequently, counteracting enhanced neurofilament expression improves axonal maintenance and prolongs survival of pmn mutant mice. We propose that this mechanism could also be relevant for other neurodegenerative diseases in which neurofilament accumulation and loss of microtubules are prominent features. Electronic supplementary material The online version of this article (doi:10.1007/s00401-016-1564-y) contains supplementary material, which is available to authorized users.


Introduction
Destabilization of axon terminals and axon degeneration are key pathological features in amyotrophic lateral sclerosis (ALS) and spinal muscular atrophy (SMA), the most common forms of motoneuron disease [6,33,62,68]. Despite progress in the identification of underlying gene defects [41,58,60], the cellular mechanisms that are responsible for loss of motor function, degeneration of neuromuscular endplates, destabilization of axons and finally death of motoneuron cell bodies are not fully understood [20]. SMA and ALS are considered both clinically and genetically as distinct disorders. However, they share common features. On the clinical side, dysfunction and degeneration of neuromuscular endplates appears as an early symptom in both disorders, and alterations of the axonal cytoskeleton are characteristic at early stages of both diseases [16,29,35]. Accumulation of spheroids in proximal axons is commonly observed in ALS, and these spheroids primarily contain neurofilaments [13,23,52]. In a mouse model of SMA, the density of intermediate filaments in proximal and distal axons is highly increased [9], similar as in tg(SOD1*G93A) mutant mice [75], a model for a common form of familial ALS. This increased density of neurofilaments coincides with lack of axonal sprouting and increased dynamics and instability of axonal microtubules [19,22,37,38], indicating that the increased density of intermediate filaments and destabilization of microtubules are functionally connected.
Neurofilaments modulate microtubule stability and thus axon caliber and microtubule functions in axonal transport [49]. NFL is required for the assembly of neurofilaments, as NFM and NFH cannot form intermediate filaments in the absence of NFL [17,18]. Abnormal accumulation of neurofilaments causes neuronal degeneration by disrupting axonal transport of cargos required for the maintenance of axon terminals [10]. Depletion of axonal neurofilaments in Nefl−/− and tg(Prph);Nefl−/− neurons leads to improved axonal transport of mitochondria and lysosomes [56]. It also increases life span in tg(SOD1*G85R) mutant mice [79] and attenuates the neurodegenerative disease phenotype in tau T44 transgenic mice [28]. However, the precise molecular mechanisms how neurofilament depletion stabilizes microtubules, improves axonal transport and prevents axon degeneration have remained unclear.
Here we show that NFL depletion in pmn mutant mice [7,46,66] increases microtubule stability and delays axon degeneration. Disruption of the Nefl gene in pmn mutant mice increases microtubule numbers in motor axons, and ameliorates the disease phenotype. This effect appears to be caused by interaction of NFL with a protein complex including stathmin and signal transducer and activator of transcription-3 (Stat3). Interaction of Stat3 and stathmin is increased upon NFL depletion. Enhanced Stat3-stathmin interaction inhibits stathmin action in destabilizing microtubules and increases levels of soluble tubulin heterodimers making them available for the maintenance of axonal microtubules in motoneurons. This effect could explain why neurofilament depletion delays disease onset and prolongs survival in pmn mutant mice, and also how increased neurofilament levels lead to axonal destabilization in a wide variety of neurodegenerative disorders.

Description of mouse lines used in this study
Heterozygous pmn mutant mice originally maintained on Naval Medical Research Institute (NMRI) [64] genetic background were crossed with heterozygous Nefl knockout mice on a C57BL/6 genetic background [83] to produce double heterozygous Nefl+/−; pmn+/− mice. These mice were then backcrossed with NMRI genetic background for at least four generations to obtain a uniform NMRI genetic background. Subsequently, the heterozygous Nefl+/−; pmn+/− were intercrossed to obtain the mice investigated in this work. NMRI genetic background was preferred over C57BL/6 to obtain bigger litter sizes under a situation when two autosomal recessive traits (pmn and Nefl) had to be crossed to obtain the desired genotypes. Nefl knockout and corresponding control wild-type mice on a C57BL/6 genetic background were used for the immunoprecipitation experiments. All experimental procedures were approved by animal care and ethic committee of the University of Wuerzburg, the Veterinäramt of the City of Wuerzburg and the Regierung von Unterfranken, and were performed according to the guidelines of the European Union. All mice were maintained under a 12-h light/dark cycle with food and water ad libitum.

Survival of the mice
To compare survival of pmn mice lacking NFL, mice heterozygous for Nefl+/−; pmn+/− were intercrossed, and survival of mutant mice of either sex with different genotypes was compared. Nefl+/+;pmn mice were used as a control for comparison with Nefl+/−;pmn and Nefl−/−;pmn to analyze survival with Kaplan-Meier curves. The median survival for the mice was compared using log rank test. Mice analyzed for survival were not used for the behavior tests.

Functional analysis of mice with rotarod and grip strength tests
Motor coordination and motor performance of the mutant mice were tested using a rotarod (Accelerating Rotarod, Ugo Basile) by recording the latency (time from beginning of the trial until the mouse falls off) to fall off the rotating rod. Six mice from each genotype, Nefl+/+;pmn, Nefl+/−;pmn and Nefl−/−;pmn were tested at a presymptomatic stage of postnatal day 21 and on three consecutive days when the mice showed signs of disease at the age of 27, 28 and 29 days both on a constant speed (8 rpm, 10 min maximum per test) and an accelerating speed rotarod (linear acceleration from 4 to 40 rpm within 2 min). The latency to fall off the rotating rod (in seconds) was recorded for each mouse for three trials spaced by 10 min each. The average latency for all the 9 trials for each mouse for three consecutive days (27)(28)(29) was plotted and used for further analysis. Forelimb grip strength of mice was tested using an automatic grip strength meter (Chatillon, Columbus Instruments) [47]. Mice were allowed to grasp a horizontal metal grid and pulled by their tail until the grip was released. The peak pull-force (in Newton) was recorded on the digital display. This test was performed for 6 trials per day at day 21 when disease was not apparent and for three consecutive days (age of 27, 28 and 29 days) when the mice showed signs of disease. Average grip strength from day 21 and 3 days at disease stages is plotted. The behavior tests were performed during the light cycle. All tests were performed on each mouse with a time gap of at least 1 h.

Primary motoneuron culture
Lumbar spinal cord was dissected from embryonic day 13.5 mice as described [78]. Isolated spinal cord from each embryo was cleaned and trypsinized at 37 °C for 15 min in HBSS, containing trypsin at a final concentration of 0.1 %. Trypsinization reaction was stopped by adding trypsin inhibitor (Sigma-Aldrich; T9128) to a final concentration of 0.1 % and digested tissues were gently triturated to obtain single cells. Panning plates were prepared by coating 24-well Nunclon™ surface dishes with antibody against p75 NTR , clone MLR2 (a gift from R. Rush, Flinders University, Adelaide, Australia; [61] and also commercially available through Biosensis (M-009-100) and Abcam (ab61425)), diluted to a final concentration of 5 ng/ml. Cells were transferred to panning dishes for 45 min where motoneurons expressing p75 NTR receptor attach to the antibody coated surface of the dish. Enriched motoneurons were then plated on polyornithine and laminin-111 (catalog no. 23017-015, lot no. 1347084; Invitrogen) coated glass coverslips (Marienfeld) or dishes. These cells were cultured at 37 °C temperature, 5 % CO 2 in neurobasal medium (Invitrogen) containing 500 μM GlutaMAX (Invitrogen), 2 % horse serum (Linaris), 2 % B-27 supplement (Invitrogen), and the neurotrophic factor BDNF (5 ng/ml). Culture medium was changed on day 1, 3 and 5 after plating. 2000 cells were plated per 10 mm coverslip for measurements of axon length, 2000 cells per dish were plated for electron microscopic analyses and 5000 cells per 25 mm coverslip were plated for SIM.

Electron microscopy of primary motoneurons
Motoneurons were cultured on special Falcon dishes (BD Falcon™-Dish 35 × 10 mm non-TC Petri EZGrip 500cas-BD Biosciences) for 7 days following the motoneuron culture protocol as described above. These motoneurons were fixed with 2.5 % glutaraldehyde and 0.8 % tannic acid in 0.1 M cacodylate buffer pH 7.5 (CB), for 5 min at 37 °C followed by 90 min at room temperature. Neurons were then washed three times with CB and treated with 1 % OsO 4 in CB for 1 h. Cells were subjected to dehydration in 30, 50 % ethanol for 5 min each, followed by 30 min 0.5 % uranyl acetate in 70 % ethanol, followed by 90, 96, 100 % ethanol for 5 min each, and were finally embedded in a thin sheet of Epon (Serva, Heidelberg, Germany) resin. After polymerization, the resin sheets were stained with methylene blue for light microscopic identification of motoneurons. Resin pieces containing identified motoneurons were mounted on empty Epon blocks, ultrathin sections of ca. 80 nm were prepared, transferred to Formvar-coated nickel grids and contrasted with uranyl acetate and lead citrate [59]. Electron micrographs were obtained with a transmission electron microscope (LEO 912 AB; Carl Zeiss). Intermediate filaments have a diameter of 11.29 ± 0.46 nm. For measuring the IF density, 12 parallel lines at a distance of 173 nm, perpendicular to the long axis of the axon were drawn on the micrograph and the number of intermediate filaments intersecting each line was counted [12]; for proximal axons (within 50 μm distance to the cell body), distal axons (less than 100 μm distance to the axon tip), and intermediate parts (in-between these two regions). Width of the axon was measured for each line and IF density was calculated by dividing the mean number of IF intersection by mean axon width.

Electron microscopy of phrenic nerves
Mice were killed by excessive CO 2 exposure and transcardially perfused with a mixture of 4 % paraformaldehyde (PFA) and 2 % glutaraldehyde in CB. Distal phrenic nerves were collected and postfixed in the same fixative overnight at 4 °C. On the next day, nerves were washed with CB and treated with 2 % OsO4 in CB for 2 h, dehydrated in an ascending concentration of ethanol and embedded in Epon. Transverse ultrathin sections of the nerve were transferred to Formvar-coated nickel grids and contrasted with uranyl acetate and lead citrate. Electron micrographs were obtained with a transmission electron microscope (LEO 912 AB; Carl Zeiss). At least 3 mice from each genotype were used and microtubules were counted in transverse sections of axons from each mouse. Area of the axon was calculated using the ImageJ software (NIH).

Immunocytochemistry and light microscopic analysis
Motoneurons cultured for 3 or 7 days were fixed with 4 % PFA (freshly prepared) for 20 min at room temperature. PFA was washed out with PBS by 3 washes for 5 min each. Fixed neurons were treated with blocking solution (0.3 % Triton X-100, 0.1 % Tween-20, 10 % horse serum in PBS) for 30 min. Primary antibodies diluted in blocking solution were added onto the neurons and incubated overnight at 4 °C. On the following day, neurons were washed three times with washing solution (0.1 % Triton X-100, 0.1 % Tween-20 in PBS), 10 min each wash and incubated for 1 h with the corresponding fluorescently labeled secondary antibodies at room temperature. Cells were washed with washing solution three times for 10 min each wash and mounted on object glass slides using aqua-polymount (Polysciences) and imaged under a confocal microscope (SP2; Leica) using a HC PL-APO 20×/0.70 objective. Axon length was measured using LAS AF software (Leica) by observers who were blind with respect to the genotype.

Structured illumination microscopy
For structured illumination microscopy (SIM) analysis, motoneurons were cultured for 3 days and fixed with the above described protocols. Neurons were labeled by indirect immunofluorescence using secondary antibodies labeled with Alexa Fluor 488, Cy3 and Cy5. Specimens were imaged using a SIM Zeiss ELYRA S.1 microscope system with a 63×/1.40 oil immersion objective in x-y-z stacks. Raw images (16 bit) were processed to reconstruct high-resolution information using the provided commercial software package (Zeiss). Three-color images were aligned using a transformation matrix and were later processed with ImageJ. Shown are maximum-intensity projections of 5 z-stacks.

Western blot analysis
Sciatic nerves from 34-day-old mice were isolated and lysed with a glass-glass homogenizer in RIPA buffer (50 mM Tris pH 7.4, 150 mM NaCl, 1 % Triton X-100, 0.1 % SDS, 2 mM EDTA, protease inhibitor, 0.5 % sodium deoxycholate, 1 mM NaF, 10 mM sodium pyrophosphate, 1 µM okadaic acid and 2 mM sodium orthovanadate). Concentration of protein was calculated using bicinchoninic acid assay (BCA assay) and lysate was boiled in Laemmli buffer for 10 min at 99 °C. Equal amount of protein from different samples were subjected to SDS-PAGE and transferred to nitrocellulose membranes. After incubation with the desired antibodies, membranes were developed using ECL or ECL advance (GE Healthcare). Obtained blots were scanned and band intensities were quantified by densitometry analysis with ImageJ (NIH).

Immunoprecipitation
Sciatic nerves or NSC34 cells were lysed in IP buffer (50 mM Tris pH 7.4, 150 mM NaCl, 1 % Triton X-100, 2 mM EDTA, Protease inhibitor, 1 mM NaF, 10 mM sodium pyrophosphate, 1 µM okadaic acid, 2 mM sodium orthovanadate) using a glass-glass homogenizer (only for sciatic nerves lysis) and at least 200 µg of protein lysate was incubated with 5 µl anti-stathmin (rabbit) antibody, overnight at 4 °C on a rotating wheel. The protein-antibody mix was incubated with pre-equilibrated Protein A-agarose beads (Roche) for 1 h at 4 °C on a rotating wheel. Protein coupled beads were pelleted by centrifugation at 500 rpm for 3 min at 4 °C and supernatant was stored at −20 °C. Beads were washed three times with IP buffer and proteins were eluted by boiling the beads with 2X Laemmli buffer at 99 °C for 10 min. Eluted proteins were loaded on a SDS-PAGE and immunoblotting was done for stathmin, Stat3 and anti-NFL. The vector used for overexpression of NFL under a pRc/CMV promoter was supplied from Dr. JP Julien.

Microtubule-regrowth assay
An established MT regrowth assay [1] with following modifications was used to study the MT regrowth in cultured motoneurons. 6000 cells were plated on polyornithinelaminin coated 12 mm coverslips and incubated at 37 °C for 1 h. Cells were then provided with full medium (2 % horse serum, 1X B-27 and 5 ng BDNF) containing 10 µM nocodazole to depolymerize the microtubule network. After 6 h of depolymerization, cells were washed 5 times with warm neurobasal (NB) medium and incubated at 37 °C for 5 min with 500 µl of warm NB to investigate the regrowth of microtubules. These cells were then washed with MT stabilizing buffer PHEM (60 mM Pipes, 25 mM Hepes, 10 mM EGTA, and 2 mM MgCl 2 ) and MTs were extracted by adding PHEM with 0.5 % Triton X-100 and 10 µM paclitaxel for 3 min at 37 °C. Cells were rinsed with PHEM and fixed in 4 % PFA + PHEM (1:1). These motoneurons were then stained for α-tubulin to label polymerized MT and γ-tubulin to label the microtubule organizing center (MTOC). Images were taken at a SP5 confocal microscope (Leica, Bensheim, Germany) with a 63X oil objective, 4X magnification and 1.35 NA Image J was used to quantify the regrowth of microtubules. After background subtraction of radius 50, similar threshold was set for all the images and Sholl analysis was performed to quantify the number of microtubule intersections at a step size of 1 µm. Total length of MT for each cell was determined by multiplying the number of intersections at each step by its distance from the MTOC, starting from the periphery to the center. Mean length was calculated by summing up the length of MTs obtained and divided by the total number of MTs measured: Σ ((nx − (nx + 1)) × ∆MTOCx)/N, where n is the number of intersections, x is the circle of interest, x + 1 is the next outer circle, ∆MTOCx is the distance from MTOC to the circle of interest, and N is the total number of MTs.

Statistical analysis
Statistical analyses were performed using the GraphPad Prism 4.02 software (GraphPad, San Diego, CA, USA). A log rank test was used to test for the significance of differences in survival of mice as shown in the Kaplan-Meier curves in Fig. 2a. Parametric tests were used for normally distributed data and non-parametric tests for data which were not normally distributed. All tests were two-tailed unless otherwise mentioned. All data are expressed as mean ± SEM. Final processing of all images was performed with ImageJ (Rasband, WS, ImageJ, US National Institutes of Health, Bethesda, Maryland, USA, http://imagej.nih.gov/ij/, 1997-2015), [65] and Photoshop CS5 (Adobe). Brightness and contrast were enhanced for better visualization.

Increased intermediate filament levels in pmn mutant motoneurons
Axonal accumulation of neurofilaments is a pathological hallmark in many types of motoneuron disease. Pmn mutant mice suffer from a severe form of motoneuron disease which starts in third postnatal week and leads to death within 3-4 weeks [64]. In order to investigate the axonal pathology in pmn mice, we first analyzed the ultrastructure of phrenic nerves from 34-day-old pmn mice by electron microscopy. We observed a marked increase in number and density of intermediate filaments with typical diameter of approx. 10 nm in cross sections of phrenic nerve from pmn mutant mice (Fig. 1a). We then analyzed the ultrastructure of isolated embryonic motoneurons which were cultured for 7 days in vitro. Analysis of proximal, intermediate and distal compartments of pmn motor axons showed an increase in number of intermediate filaments in all the compartments as compared to the wild-type axons (Fig. 1b,  c). Next, we analyzed the levels of neurofilament proteins in peripheral nerves of pmn mutant mice. For this purpose, sciatic nerves were isolated from 34-day-old pmn mutant mice, representing an advanced stage of disease, and subjected to Western blot analysis for NFL and NFH protein.
Pmn mutant mice showed a significant increase in NFL and NFH protein levels in comparison to wild-type mice ( Fig. 1d-f). Equal loading was controlled by determining eIF2α levels. The increase of NFL expression did not occur on the transcriptional level (Fig. 1g), indicating that altered translation or altered stability of neurofilaments accounts for this observation.

NFL depletion prolongs life span and improves motor performance in pmn mutant mice
In order to study the role of increased NFL in pmn disease phenotype, mutant mice were crossbred with Nefl knockout mice. Deletion of one Nefl allele reduced the levels of NFL in Nefl+/−;pmn littermates to control levels, as observed in wild-type mice (Fig. 1d, e). Also, the increased levels of NFH protein in pmn mice were normalized to wild-type levels in Nefl+/−;pmn mice ( Fig. 1d, f).
To investigate the effect of increased neurofilament expression on the neurodegenerative process in pmn mutant mice, we generated pmn mice completely lacking neurofilament light chain. Western blot analysis of sciatic nerve protein extracts from 34-day-old mice confirmed the absence of NFL (Fig. 1d) in Nefl−/−;pmn mice. Nefl deletion per se does not reduce survival of the mice, and Nefl knockout mice do not show an overt disease phenotype [83]. We followed the life span of pmn mice that lack NFL. Nefl−/−;pmn mice showed a significant extension of life span with an average increase of ~21 % in median survival as compared to Nefl+/+;pmn mice (log  Figure 2b shows the hind limb of pmn mice from the same litter with (left) and without (right) NFL at postnatal day 29. Notably, the pmn mouse with NFL exhibited completely atrophied hind limbs, at the same age when Nefl−/−;pmn showed still preserved hind limb motor function. The average weight of pmn, Nefl+/−;pmn and Nefl−/−;pmn mice did not differ significantly at day 21 which represents a presymptomatic or early disease stage or at 27-29 days which marks a period of rapid disease progression (Fig. 2b, c).
In order to assess the motor performance and disease progression in Nefl−/−;pmn mice, we tested the grip strength of six mice per genotype and their performance on the rotarod on day 21 at disease onset and for three consecutive days (age of 27, 28 and 29 day) during a phase of fast disease progression (Fig. 2d, e). No significant difference was observed in the grip strength or rotarod performance between the three genotypes (Fig. 2d, e; supplementary Figure 3) at postnatal day 21. At day 27-29, we found that both Nefl−/−;pmn and Nefl+/−;pmn mice exhibited higher grip strength as compared to Nefl+/+;pmn mice (Fig. 2d), using a grip strength meter. There was a tendency towards higher latency to fall off the accelerated rotarod on day 21 in Nefl−/−;pmn mutant mice but this difference was not statistically significant (n = 6-7 for each group, P > 0.05). At day 27-29, Nefl−/−;pmn and Nefl+/−;pmn mice showed an increase in the latency to fall off the rotating rod, both on an accelerating (Fig. 2e) and a constant speed rotarod (Suppl. Figure 1) as compared to Nefl+/+;pmn mice. On the accelerating speed rotarod Nefl−/−;pmn could stay up to 50 s on the rod when the rotating speed was increased to 40 rpm, in contrast to Nefl+/+pmn mice which could not sustain the higher rotating speed at day 27-29. Interestingly, Nefl+/−;pmn mice in which the NFL levels are comparable to wild-type mice performed better in the rotarod test (Suppl. Figure 1) than Nefl−/−;pmn mice that completely lack NFL. Hence, deletion of Nefl from pmn mice does not influence disease onset, but prolongs survival of these mice, improves motor performance and partially ameliorates the disease phenotype.

NFL depletion increases microtubule number in motor nerves of pmn mice and rescues defective axon elongation in vitro
Pmn mutant mice exhibit reduced axonal microtubule density in the distal part of phrenic nerves [46,63]. Similarly, cultured motoneurons from pmn mutant mice show reduced axonal microtubule density in the proximal part of the axons [66]. On the other hand, Nefl−/− mice show increased microtubule numbers in axons within spinal cord and higher α-tubulin protein levels in the cerebral cortex [28]. In order to study the microtubule dynamics in Nefl−/−;pmn mice, we analyzed the ultrastructure of distal phrenic nerves in 34-day-old wild-type, pmn, Nefl−/− and Nefl−/−;pmn mice. Confirming previous analyses [46,62] we observed a reduction in the number of microtubules per axon in pmn mutant mice (Fig. 3a, b) as compared to wild-type mice (wild-type 86.91 ± 9.793 and pmn 41.52 ± 8.585) (P < 0.05; t = 3.640; one-way ANOVA). Deletion of Nefl prevented the loss of microtubules per axon in phrenic nerves of pmn mutant mice, and Nefl−/−;pmn (71.83 ± 4.858) axons showed microtubule numbers comparable to wild-type (Fig. 3a, b; P > 0.05; t = 1.306; oneway ANOVA). However, the total number of microtubules per axon was not significantly higher in Nefl−/− mice (120.6 ± 11.47) compared to axons from control wild-type mice (86.91 ± 9.793; P > 0.05; t = 2.701), indicating that depletion of NFL resulted in maintenance or stabilization of microtubules but not in the generation of additional microtubules. Pmn pathology did not alter axonal diameter and therefore the cross-sectional area of axons in phrenic nerve of pmn mice remained unaffected (Fig. 3d) at postnatal day 34. In contrast, NFL appeared as a major modulator of axon caliber [24]. Confirming previous studies, we observed that NFL depletion led to a drastic reduction in the cross-sectional area of the axons in Nefl−/− and Nefl−/−;pmn axons in phrenic nerves of 34-day-old postnatal mice as compared to wild-type controls (Fig. 3d). Consequently, the density of microtubules per axon increased drastically in Nefl−/− and Nefl−/−;pmn nerves as compared to wild-type and pmn controls (Fig. 3c) in postnatal nerves at stages when these mice show clinical symptoms. These data suggest that NFL depletion prevents destabilization of microtubules in pmn mice and leads to an increased density of microtubules in axons by reducing the axonal caliber in both wild-type and pmn mutant mice.
Arrangement, distribution and organization of microtubules are essential for axonal outgrowth and elongation [14,74]. Isolated motoneurons from pmn mutant mice show reduced axon outgrowth when cultured for 7 days in vitro [7], which could be rescued by CNTF application [66]. In order to study the effect of increased microtubule dynamics on axon elongation in pmn motoneurons lacking NFL, we measured and compared axon length of Nefl+/+;pmn and Nefl−/−;pmn mutant motoneurons cultured for 7 days in vitro in presence of BDNF. CNTF was not present in these cultures. Pmn motoneurons showed significantly shorter axons when cultured with BDNF ( Fig. 3e, f), whereas axons from Nefl−/− motoneurons grew comparable to those of wild-type motoneurons. Nefl deletion completely rescued the defect in axon elongation of pmn mutant motoneurons, and Nefl−/−;pmn motoneurons grew significantly longer axons as compared to axons of pmn motoneurons, reaching axon lengths comparable to those of wild-type. Survival of Nefl−/− motoneurons after 7 days in vitro did not differ from that of wild-type motoneurons (Fig. 3g).
Previous studies [66] showed that the axon diameter of cultured pmn motoneurons does not differ from wild-type controls, which is in line with the observations made in this study with axons from phrenic nerves of the same mutant mice at postnatal day 34 (Fig. 3d). In order to test how axons were altered in Nefl−/− motoneurons, longitudinally sectioned motor axons were analyzed under the electron microscope. Proximal, intermediate and distal compartments of wild-type and Nefl−/− motor axons were used to measure the axonal diameter and the average diameter of wild-type and Nefl−/− axons is plotted. Axons of Nefl−/− motoneurons cultured for 7 days in vitro did not show changes in diameter as compared to wild type (Fig. 3h, i), indicating that the rescue effect observed in Nefl−/−;pmn mutant motoneurons is not a consequence of altered axon diameter.

Nefl deletion rescues microtubule regrowth defect in cultured pmn motoneurons
In order to study microtubule dynamics in Nefl−/− motoneurons, we investigated microtubule regrowth and polymerization in cultured motoneurons after complete disintegration of microtubules. For this purpose, an established protocol for MT regrowth after nocodazole treatment was  [1,63,66]. Isolated E13.5 motoneurons were plated in culture dishes for 1 h and the established microtubule network was completely depolymerized by applying 10 µM nocodazole for 6 h. Nocodazole was then washed off and microtubules were allowed to regrow. Sholl analysis (Fig. 4a, b) was performed to quantify MT polymerization, average MT length and total length of MT after regrowth. Pmn mutant motoneurons showed reduced MT regrowth when compared to wild-type motoneurons, as reported previously [66] (Fig. 4a, c). Nefl−/− motoneurons showed no change in MT repolymerization and the MT intersections appeared comparable to wild-type neurons (Fig. 4a, d). On the other hand, Nefl deletion significantly increased MT regrowth and repolymerization in pmn motoneurons, and Nefl−/−;pmn motoneurons showed significantly higher microtubule regrowth than pmn motoneurons (Fig. 4a, f) when cultured with BDNF.
Next, we measured the average length of microtubules. For this purpose, length of all the microtubules growing from the microtubule organizing center (MTOC) was measured and average length was calculated, as reported previously [66]. Pmn mutant motoneurons exhibited shorter average microtubule length compared to wild-type neurons (Fig. 4g). Nefl−/− motoneurons showed microtubule length comparable to wild type. Average microtubule length within Nefl−/−;pmn motoneurons was longer than the length of microtubules in pmn motoneurons and reached levels comparable to wild type and Nefl−/− controls (Fig. 4g). We also measured the total length of microtubules and found that Nefl−/−;pmn motoneurons show an increase in total microtubule length originating from MTOC as compared pmn motoneurons, that showed reduced total microtubule growth (Fig. 4h).

NFL interacts with Stat3-stathmin complex and NFL depletion increases Stat3-stathmin interaction
CNTF-dependent restoration of axon length and maintenance of pmn motoneurons is mediated by local axonal activity of Stat3 [66]. Stat3 is activated upon CNTF application and the activated Stat3 interacts with stathmin thereby inhibiting its microtubule destabilizing activity and preventing the axonal degeneration in pmn motoneurons. Stat3-stathmin interaction has been shown to be increased in motoneurons upon CNTF application [66]. We therefore studied the interaction of Stat3-stathmin in Nefl−/− mice. We performed immunoprecipitation with protein lysates of sciatic nerves from 34-day-old wild-type and Nefl−/− mice via stathmin pulldown and investigated the interaction partners of stathmin. This pulldown experiment revealed that NFL also interacts with stathmin (Fig. 5a). Sciatic nerve lysates from Nefl−/− mice showed an increase in Fig. 5 Nefl deletion increases Stat3-stathmin interaction. a Stathmin was immunoprecipitated from sciatic nerve extracts of 34-dayold mice. Western blot shows from left to right, input (Ip) and eluate (E) from wild-type and Nefl−/− mice nerve extracts after anti-IgG control and anti-stathmin pulldown. b Quantification of band intensities in the eluate showed an increased interaction of Stat3 and stathmin in Nefl−/− mice (t = 17.08; P = 0.0034). Bars represent mean ± SEM (one sample t test, n = 3 independent experiments). c Nefl deletion increases Stat3-stathmin interaction in sciatic nerves from pmn mutant mice. d Western blot analyses show similar levels of stathmin in the sciatic nerve extracts of 34-day-old wild-type, pmn and Nefl−/− mice. Histone levels were determined to ensure equal loading of proteins. e Quantification of Stat3 per stathmin levels in the sciatic nerve extracts (input before immunoprecipitation) from 34-day-old wild-type and Nefl−/− mice (n = 3 independent experiments). f Stathmin was immunoprecipitated from NSC-34 cell extracts. Western blot shows from left to right input for IgG, mock, and NFL overexpressing NSC-34 cells. Right blot shows eluate after IgG control pulldown and stathmin pulldown from NSC-34 cells after transfection with mock (only lipofectamine) or, NFL overexpression vector, respectively. g Western blot analysis reveals increased levels of phosphorylated Stat3 (at Y705) in the sciatic nerve extracts of 34-day-old pmn and Nefl−/− in comparison to wild-type mice. Quantification of pStat3 normalized by total Stat3 levels in h pmn mouse compared to wild type (t = 4.928; P = 0.0044; n = 6 independent experiments) and i in Nefl−/− mouse compared to the wild-type mouse (t = 2.865; P = 0.0242; n = 8 independent experiments). Bars represent mean ± SEM Stat3-stathmin interaction as compared to wild-type mice (Fig. 5a, b). Also in pmn mutant mice, NFL depletion led to an increase in Stat3-stathmin interaction in sciatic nerve extracts (Fig. 5c). In Western blots of sciatic nerve extracts, the levels of Stat3 (Fig. 5a) and stathmin (Fig. 5d) and also the ratio of Stat3/stathmin (Fig. 5e) were not altered, excluding the possibility that the enhanced Stat3-stathmin interaction is due to altered expression of any of these two proteins. Similarly, Stathmin2 levels were unchanged in pmn mutant sciatic nerves (Suppl. Figure 2a) and the mRNA levels for Stmn1 were not changed in the adult pmn or Nefl−/− mouse sciatic nerves ( Supplementary  Figure 2b, c). Hence, depletion of NFL increases the interaction of Stat3-stathmin that antagonizes the MT-destabilizing activity of stathmin [53].
In order to confirm the interaction of NFL and stathmin, we performed these immunoprecipitation experiments with NSC34 cells. For this purpose, cell lysates were subjected to stathmin pulldown and NFL and Stat3 interaction were studied by co-immunoprecipitation. Our data confirmed the interaction of NFL with stathmin in this cell culture system (Fig. 5f). Overexpression of NFL led to an increased interaction of NFL with stathmin and in parallel a decreased interaction with Stat3 (Fig. 5f).
Since the phosphorylation of Stat3 at Y705 plays an important role in its interaction with stathmin upon CNTF treatment [66], we studied the levels of phosphorylation of Stat3 in sciatic nerve extracts of 34-day-old Nefl−/− and pmn mice. The levels of phosphorylation showed an about fivefold increase in pmn nerve extracts (Fig. 5g, h). Nefl−/− nerve extracts also showed an increase in the phosphorylation at Y705, indicating that the increase in the stathmin/Stat3 interaction is mediated via activation of Stat3 (Fig. 5g , i).

Nefl deletion affects the distribution and interaction of Stat3 and stathmin with tubulin in motoneurons
Dimers of αand β-tubulin are the basic components of microtubules. These heterodimers undergo a variety of post-translational modifications which assign specific functions to the tubulin subunits. Acetylation of α-tubulin occurs in polymerized microtubules, and the acetylated tubulin is enriched in stable microtubules with a low turnover. On the other side, tyrosinated α-tubulin marks recently assembled more dynamic microtubules with a high turnover [77]. In neurons, acetylated microtubules are found in more proximal parts of axons where microtubules are thought to be stable, and they appear to be excluded from axon terminals [66,80]. In the highly dynamic axonal growth cones and dendrites, tyrosinated tubulins are found [66]. Pmn motoneurons exhibit more highly dynamic MTs in axons and show higher levels of tyrosinated tubulins in axons (Suppl. Figure 3) as compared to wild-type motoneurons, whereas the levels of acetylated MTs are unchanged (Suppl. Figure 3 and [66]). We observed that NFL depletion increased the intensity of staining against acetylated tubulin in axons of motoneurons (Fig. 6a, b). In the same axons, levels of tyrosinated microtubules did not change. In summary, these data indicate that NFL depletion enhances the stability of microtubules, as could be seen by the higher levels of acetylation.
This effect could be explained by reduced activity of stathmin. Stathmin appears redistributed in axons of pmn mutant motoneurons, with higher levels of this protein in association with NFL (Suppl. Figure 4). Stathmin has two functions on microtubules dynamics. On the one side, it binds to heterodimers of αand β-tubulin and prevents the polymerization of new microtubules. On the other side, it interacts with polymerized microtubules and induces their disassembly in microtubule catastrophe reactions [8]. In order to study this possibility, we determined the subcellular distribution of Stat3 and stathmin. In wild-type axons, Stat3 was observed to be co-localized with the tyrosinated tubulin whereas in the Nefl−/− axons Stat3 co-localized with stathmin (Fig. 7). These findings emphasize that the association of axonal Stat3 with stathmin modulates microtubule dynamics in the neurons. In addition, they support the data obtained by stathmin immunoprecipitation and Western blot analysis showing that NFL modulates the interaction of Stat3 and stathmin. Thus, NFL depletion increases the interaction between Stat3 and stathmin and stabilizes microtubules by reducing the MT destabilizing activity of stathmin.

Discussion
In this study, we observed that NFL protein levels are at least twofold increased in pmn mutant motoneurons and that NFL depletion rescues defective axon growth in cultured motoneurons and prolongs survival of pmn mutant mice. This effect was found both in Nefl+/−;pmn motoneurons in which elevated NF expression was brought back to wildtype control levels and in Nefl−/−;pmn motoneurons in which axonal neurofilament was completely lost. In cultured pmn mutant motoneurons, NFL depletion also resulted in increased axon elongation and increased levels of acetylated tubulin, a marker for stable microtubules. Thus, NFL depletion modulates microtubule dynamics in a similar manner as observed after Stat3 activation. This effect was apparently due to increased phosphorylation of Stat3 and enhanced interaction of Stat3 with stathmin when NFL levels were reduced, resulting in enhanced stability of microtubules.
Accumulation of neurofilaments has been observed in a variety of neurodegenerative diseases and in corresponding mouse models [36,42,44]. Thus neurofilament pathology represents a point where pathomechanisms of distinct neurodegenerative disorders converge [40]. However, the reasons why NFL is upregulated in degenerating neurons and the precise mechanism how NFL accumulation participates in these neurodegenerative mechanisms are not fully understood so far. Several lines of evidence indicate that defective axonal transport leads to enhanced phosphorylation and accumulation of neurofilaments in axons [55]. The upregulation could also be a consequence of altered posttranscriptional control [73], or a consequence of defective intermediate filament degradation. The possibility of increased protein synthesis has been ruled out in degenerating neurons of ALS and Alzheimer disease patients in which NEFL mRNA levels are downregulated [3,48,71,72,81], whereas the transcript levels for NEFM or NEFH usually remain unchanged [81]. This indicates that both altered axonal transport, posttranscriptional mechanisms including deregulated translation or posttranslational mechanisms such as reduced degradation of NFL [60] could be responsible for the phenotype. Gigaxonin plays an essential role in the degradation of IFs [54] and the accumulated pool of IF in giant axonal neuropathy is lost upon Gigaxonin restoration in patient iPSC derived motoneurons [31], indicating that defective degradation could play a central role in this and other forms of neurodegenerative disorders.
Mouse models of motoneuron disease and other neurodegenerative disorders have provided support for the hypothesis that accumulation of NFL could be an early event in neurodegeneration. Overexpression of NFL, NFM or NFH transgenes causes NF aggregation and motoneuron dysfunction resembling motoneuron disease in mouse .011) as compared to wild-type motoneurons, whereas tyrosinated microtubules remained unchanged. Intensity of stathmin is used as control. Bars represent mean ± SEM (one-way ANOVA, n = 10 motoneurons from three independent experiments, *P < 0.05, ***P < 0.001) 1 3 models [11,21,34,82]. In mutant SOD1 mice, Nefl deletion delayed the onset of disease and slowed the disease progression [79], and in tau transgenic mouse it reduces the abnormal tau accumulation and motoneurons degeneration [28]. In pmn mutant mice, deletion of only one Nefl allele normalized NFL and NFH protein levels in sciatic nerves. Under these conditions, the motor function in pmn mutant mice was improved, thus providing evidence that the elevation of endogenous neurofilament levels contributes to axon destabilization and loss of motor function. Despite the drastic reduction of IFs and the resultant reduction in axon diameter in peripheral nerves, Nefl−/−;pmn mutant mice survived longer than Nefl+/−;pmn or pmn mutant mice. Thus, loss of IFs delays axon destabilization in pmn mutant mice.
Microtubules in pmn mutant mice are unstable, because the underlying gene defect leads to a massive reduction of αβ-tubulin heterodimers, the basic components of microtubules [46]. In cultured motoneurons, this leads to defective axon elongation [63,66], and levels of tyrosinated microtubules are increased in the axons of these motoneurons. Intermediate filaments are anatomically and functionally linked with microtubules. Neurofilament interacts with tubulin [50] and stimulates microtubule polymerization in mature neurons [4]. Any disturbance of NF protein levels does not only affect assembly of NF fibers, it also influences microtubules [30]. NFM and NFH make cross bridges between adjacent NFs and microtubules [51]. Thus, also removal of NFM and NFH sidearms delays the disease in SOD1 mutant mice [43] in a similar manner as deletion of the Nefl gene in this mouse model [79]. Disease onset is also delayed in the same mouse model after treatment with microtubule stabilizing agents [19]. However, it has not been shown so far whether stabilization of microtubules contributes to the beneficial effects of NFL depletion or removal of NFM and NFH sidearms in SOD1 mice. The observation made in our study that levels of acetylated tubulin increase in axons of NFL-deficient motoneurons points to this possibility, and indicates that the beneficial effects of normalizing IF levels or those of massive reduction of functional IFs could be due to the stabilization of microtubules.
The levels of tyrosinated tubulin are increased in all compartments of the pmn mutant axons (Suppl. Figure 3) as observed in cultured motoneurons using light microscopy. The ultrastructural analyses of these pmn mutant axons (Fig. 1c) indicate that the increase of IF also occurs in all compartments with a gradient from proximal to distal, which however is similar in motoneurons from wildtype and pmn mutant mice. Thus the structural alterations in NFL upregulation do not go in parallel with dying back mechanisms in the cultured motoneurons, but they are thought to be of relevance for axonal transport processes in vivo which could contribute the degeneration of distal parts of the axons.
Stathmin plays a central role in the regulation of microtubule stability [8]. It acts in two distinct ways on microtubule dynamics. First, it destabilizes existing microtubules by inducing microtubule catastrophes in a dose-dependent fashion in vitro [2,26,45]. Second, it binds αβ-tubulin heterodimers and sequesters them in a way that microtubule polymerization is inhibited [32]. In vitro, a change in the pH of the buffer can lead to a shift in the role of stathmin from sequestration of tubulin affecting microtubule elongation to increase in microtubule catastrophes [27]. The N-terminal of stathmin is required for the catastrophic role, whereas the C-terminal is essential for the inhibition of MT-polymerization rate in vitro [25,39]. These effects on regulating microtubule dynamics apparently are involved in plasticity processes when neurons change their shape and new neuronal connections are made, for example during learning and memory processes. Mice which lack stathmin-1 show severe defects in fear memory formation in the amygdala [69,76] and this process correlates with high levels of stahmin-1 expression found in this brain region. Not much is known about how stathmin function is regulated, but its spatial distribution within cells seems to play a role [67]. Several members of the stathmin family are associated with membranous structures by palmitoylation that orients these proteins to specific subcellular compartment and thus restricts their subcellular distribution [8]. On the other hand, stathmin-1 lacks a palmitoylation signal, but this protein is not evenly distributed in the cytoplasm. In our study, we find stathmin-1 within axons, mostly colocalized with microtubules. High-resolution light microscopy with structured illumination microscopy (SIM) allowing resolution of structures down to nearly 100 nm reveals close proximity of stathmin with tyrosinated microtubules in the axons of wild-type motoneurons. This proximity seems to be increased in Nefl−/− motoneurons, and stathmin colocalization with Stat3 and microtubules also increases, as shown in Fig. 7, right panel. This increased interaction of stathmin with Stat3 is confirmed by biochemical immunoprecipitation assays. Neurofilament appears as part of this complex in pulldown assays, indicating that IFs play a role in the formation of complexes between Stat3 and stathmin. This could explain why IF depletion modulates microtubule stability.
In wild-type motoneurons, NFL depletion had a much more pronounced effect on stability of existing microtubules that are acetylated at relatively high levels when compared to microtubule regrowth after nocodazole treatment. In pmn mutant motoneurons in which availability of αβ-tubulin heterodimers is reduced, NFL depletion also restored defective microtubule regrowth after nocodazole treatment. This differential effect of NFL depletion could be explained by the reduced availability of αβ-tubulin heterodimers in pmn mutant motoneurons, which increases upon release of αβ-tubulin heterodimers when stathmin is inactivated by enhanced interaction with Stat3 [66]. Thus, the increased interaction of Stat3 and stathmin in NFLdepleted motoneurons enhances the capacity for microtubule regrowth and microtubule plasticity in motoneurons from pmn mutant mice. This also indicates that enhanced NFL levels in neurodegenerative diseases reduce the capacity for microtubule regrowth and microtubule plasticity. Our data provide evidence that the destabilizing activity of stathmin is enhanced when NFs are increased in pmn pathology and probably in other neurodegenerative disorders, and this could make a major contribution to axonal degeneration.
When this idea is followed up towards therapeutic implications, this would mean that catastrophe-inducing endogenous MT deregulators such as stathmin proteins should be functionally blocked in order to stabilize microtubules and to enhance stability of axons in conditions involving IF accumulation or microtubule destabilization. In neurons, altered MT-based transport and aggregation of proteins is generally associated with neurodegenerative disorders [57,70]. Despite the fact that axons exhibit smaller diameter, the improvement in MT network in Nefl−/−;pmn motoneurons apparently stabilizes the axon and possibly also improves the transport of cargoes, leading to prolonged survival and delay in the decline of motor function.
In summary, our findings suggest that NF accumulation contributes to axonal destabilization in the pmn mouse model of motoneuron disease and possibly also other forms of neurodegenerative disorders. NFL depletion stabilizes the MT structure and leads to enhanced axon growth in pmn mutant motoneurons via increased activation of Stat3 by phosphorylation at Y705 and thereby increased Stat3stathmin interaction. Thus, targeting the NFL accumulation in neurodegenerative diseases could be a target for therapy in neurodegenerative disorders.