Golgi fragmentation is an early hallmark of many neurodegenerative diseases but its pathophysiological relevance and molecular mechanisms are unclear. We here demonstrate severe and progressive Golgi fragmentation in motor neurons of progressive motor neuronopathy (pmn) mice due to loss of the Golgi-localized tubulin-binding cofactor E (TBCE). Loss of TBCE in mutant pmn and TBCE-depleted motor neuron cultures causes defects in Golgi-derived microtubules, as expected, but surprisingly also reduced levels of COPI subunits, decreased recruitment of tethering factors p115/GM130 and impaired Golgi SNARE-mediated vesicle fusion. Conversely, ARF1, which stimulates COPI vesicle formation, enhances the recruitment of TBCE to the Golgi, increases polymerization of Golgi-derived microtubules and rescues TBCE-linked Golgi fragmentation. These data indicate an ARF1/TBCE-mediated cross-talk that coordinates COPI formation and tubulin polymerization at the Golgi. We conclude that interruption of this cross-talk causes Golgi fragmentation in pmn mice and hypothesize that similar mechanisms operate in human amyotrophic lateral sclerosis and spinal muscular atrophy.
The mammalian Golgi apparatus is a single copy membrane-bound organelle that comprises stacked flattened cisternae. It lays at the heart of the secretory pathway where it controls the processing and dispatching of proteins en route from their site of synthesis in the endoplasmic reticulum to their final subcellular destination (1). Remarkably, structural and functional alterations of the Golgi apparatus are a common hallmark of many neurodegenerative diseases including Parkinson, Alzheimer and prion disease (2). They are best characterized in amyotrophic lateral sclerosis (ALS) where Golgi fragmentation and atrophy are among the earliest pathological features of degenerating motor neurons (3,4).
Two main hypotheses have been proposed to explain pathological Golgi fragmentation in ALS. The first hypothesis focuses on microtubule defects. Indeed, Golgi fragmentation observed in degenerating motor neurons of transgenic mutant SOD1 mice (4), the most widely used familial ALS model, bears similarities with the change in the Golgi organization induced by microtubule-depolymerizing agents in cell culture (5–7). In addition, the microtubule-destabilizing protein Stathmin is strongly up-regulated in the spinal cord of these mice (8). The second hypothesis incriminates protein misfolding or mislocalization. Aggregation-prone mutant SOD1 proteins have been shown to mislocalize in the lumen of the Golgi apparatus (9) but see (10), to inhibit anterograde protein transport and to cause Golgi fragmentation in vitro (11,12).
To investigate whether microtubule defects or protein misfolding/mislocalization are the primary cause of Golgi fragmentation in degenerating motor neurons in vivo, we took advantage of the unique features of progressive motor neuronopathy (pmn) mice (13). These mice bear a mutation in the tubulin-binding cofactor E (TBCE) gene (14,15) which encodes one of at least five tubulin-specific chaperones (TBCA–TBCE) known to promote tubulin folding and microtubule polymerization (16,17). TBCE is prominently expressed in motor neurons (14) where it localizes to the Golgi and shuttles as a peripheral membrane-associated protein between the Golgi membrane and the cytosol (18). The importance of TBCE function in human disease is underscored by autosomal recessive TBCE mutations in patients with Sanjad–Sakati/Kenny–Caffey syndrome suffering from severe growth and mental retardation, facial dysmorphism and congenital hypoparathyroidism (19).
The pmn mutation, a homozygous Trp–Gly exchange at the C-terminus of TBCE, destabilizes the protein (14) which decreases the level of polymerized microtubules but increases the level of soluble tubulin (18), allowing to test the two above mentioned hypotheses on the molecular mechanism of Golgi fragmentation. Furthermore, microtubule loss in pmn mice starts in distal axons leading to axonal dying back (18) but leaves motor neuron cell bodies in spinal cord intact (20) enabling to investigate Golgi structure during the entire disease course.
Here, we show that in pmn mice as in human ALS, Golgi fragmentation is one of the earliest features of degenerating motor neurons. Loss of TBCE in pmn spinal motor neurons and in TBCE-depleted NSC34 motor neuron cultures impairs the polymerization of Golgi-derived microtubules, as expected, but surprisingly also results in the loss of COPI vesicle subunits, defective localization of Golgi tethering factors and increased levels of ER-Golgi v-SNAREs, yet impaired SNARE pair formation, all features sustaining the observed changes in Golgi architecture and leading to extensive Golgi vesiculation. Conversely, ARF1 that drives COPI vesicle formation recruits TBCE to the Golgi membrane, enhances polymerization of Golgi-derived microtubules and rescues TBCE-linked Golgi fragmentation. These data unravel an unprecedented ARF1/TBCE-mediated cross-talk that coordinates COPI formation and tubulin polymerization at the Golgi. We conclude that defective TBCE function in the cross-talk causes Golgi fragmentation in pmn motor neurons and hypothesize that similar mechanisms operate in other motor neuron diseases such as SOD1-linked ALS and SMA.
Golgi fragmentation and atrophy in pmn motor neurons involves severe progressive Golgi vesiculation
To investigate whether Golgi structure is altered in motor neurons of pmn mice, we first used a microscopy approach. When examined in lumbar spinal cord cryosections of 25-day-old mice by immunofluorescence for the Golgi transmembrane protein MG160, the Golgi of pmn motor neurons is fragmented and atrophied when compared with wild-type littermates (Fig. 1A–E and Supplementary Material, Fig. S1A–B), whereas ER (KDEL) and early endosomes (EEA1) appear normal (Supplementary Material, Fig. S1C–D).
When examined by electron microscopy, the Golgi in the cell body of wild-type motor neurons exhibits the typical morphology of stacked cisternae (Fig. 1F). In pmn motor neurons, however, three prominent types of Golgi alterations are observed (Fig. 1F). In the first, some Golgi cisternae are disrupted and replaced by tubules and vesicles whereas others remain intact (‘partially vesicular’). In the second, Golgi cisternae are completely transformed into dense clusters of numerous tubules and small vesicles of ∼40–50 nm in diameter (‘completely vesicular’). In the third, only small remnants of fragmented Golgi are observed (‘remnant’). Between Day 15 when the first symptoms appear and Day 35 just before the animal's death, the frequency of completely vesiculated Golgi increases fourfold (Fig. 1G), suggesting that Golgi fragmentation and atrophy in pmn motor neurons is due to progressive Golgi vesiculation.
Loss of TBCE results in COPI degradation and dispersion of the tethers p115/GM130
To start addressing the mechanisms of Golgi vesiculation in pmn motor neurons, we asked whether the formation and trafficking of COPI, COPII and clathrin-coated vesicles that mediate transport to, through and from the Golgi (1) is affected. To do so, we assessed the expression of specific coat proteins. The immunoreactivity of the COPI coat subunit β-COP is almost undetectable in pmn motor neurons (Fig. 2A) and the protein levels of β-COP and ε-COP in pmn spinal cords are strongly reduced when assessed by western blot (Fig. 2B), whereas the levels of clathrin, the COPII coat subunit Sec23 and also the intraluminal Golgi-resident enzymes Mannosidase II (MannII) and Galactosyltransferase (GalT) are similar in pmn and wild-type tissues (Fig. 2B). To investigate this further, we used NSC34 motor neuron cultures where endogenous TBCE was completely depleted (Fig. 2C–E). Similar to pmn motor neurons in vivo, TBCE-depleted NSC34 cells exhibit extensive Golgi fragmentation observed by the pattern of MannII-green fluorescent protein (GFP) (Fig. 2C), as well as strongly decreased β-COP levels observed by immunofluorescence (Fig. 2C), western blot (Fig. 2D) and flow cytometry (Fig. 2E).
Since β-COP directly interacts with a number of vesicle tethering proteins including p115 (22), we then tested whether the reduction of β-COP impacts on p115 and indirectly on GM130, a known p115 interactor (23). p115 and GM130 show reduced immunofluorescence signals at the Golgi of pmn motor neurons (Fig. 2F) but unaltered protein levels in pmn spinal cords (Fig. 2G), suggesting that the loss of the COPI coat affects the membrane recruitment of these tethering proteins but not their stability. Cell fractionation (Fig. 2H) confirmed that a significant pool of p115 and GM130 is cytosolic in pmn spinal cord contrasting with their exclusive association with membranes in wild-type. Similarly, TBCE-depleted motor neurons in culture display more hazy GM130 immunofluorescence (Fig. 2I) and decreased membrane-bound p115 (Fig. 2J).
To characterize these changes quantitatively, we fractionated cell compartments on sucrose gradients (Fig. 2K). In control cells, TBCE, MannII and GM130 are confined to light fractions containing Golgi membranes whereas β-COP is expressed in membranes of increasing densities, including fractions of 1.14–1.36 m sucrose known to contain COPI vesicles (21). Conversely, in TBCE-depleted cells, β-COP expression is specifically lost in the vesicle fractions and GM130 is no longer associated to Golgi membrane (Fig. 2K), confirming that loss of TBCE function leads to β-COP redistribution and degradation as well as reduced membrane recruitment of tethers. Since β-COP degradation is sufficient to cause vesicle accumulation and GM130 redistribution in HeLa cells (24), it most likely mediates Golgi vesiculation in pmn and TBCE-depleted motor neurons.
In the early secretory pathway, COPII vesicles normally fuse with themselves (25) and the ER/Golgi intermediate compartment (ERGIC) (26). β-COP-depletion is known to cause fragmentation of the ERGIC into vesicular structures containing ERGIC-53 (27) and Golgi-resident enzymes, such as GalT (22). In line with this, we found that ERGIC-53 and GalT are completely shifted to vesicle fractions in TBCE-depleted cells (Fig. 2K), suggesting that the accumulating vesicles represent ERGIC fragments and COPII vesicles unable to fuse.
Loss of TBCE leads to increased levels of Golgi v-SNAREs but impaired vesicle fusion
The fusion of COPI and COPII vesicles with their target membranes is mediated by the pairing of ER/Golgi v-SNAREs with cognate t-SNAREs (28). The vesicle accumulation in pmn and TBCE-depleted motor neurons might therefore suggest an impaired SNARE pairing. To test this, we assessed the formation of the two known ER/Golgi SNARE pairs GS28/Syntaxin 5a (29) and GS15/Ykt6 (30). By western blot, the total levels of the t-SNAREs Syntaxin 5a and Ykt6 (Fig. 3A) remain unchanged in pmn spinal cords whereas the levels of the v-SNAREs GS15 and GS28 are increased by >10-fold in pmn when compared with wild-type spinal cords (Fig. 3A). The level of the endosomal v-/t-SNAREs Vti1a/Syntaxin 6 remains unaffected demonstrating specificity (Fig. 3A). The increase in GS28 and GS15 levels in pmn versus wild-type motor neurons is also documented by immunofluorescence (Fig. 3B, upper rows). Quantitative polymerase chain reaction analyses showed no increase in GS15 and GS28 mRNA levels in pmn spinal cords at various time points (Supplementary Material, Fig. S2), suggesting that the increase in v-SNARE protein levels is due to enhanced protein stability or reduced protein degradation rather than to transcriptional up-regulation.
ER/Golgi v-SNAREs are recycled back to the ER by COPI vesicles (31,32) and loss of β-COP in pmn and TBCE-depleted cells might impede this recycling. To assess v-SNARE recycling, we examined the distribution of GS15 and GS 28 that accumulate in small dispersed puncta, as shown by 3D modeling (Fig. 3B, lower rows). These puncta likely correspond to Golgi vesicles or groups of Golgi vesicles as both v-SNAREs are shifted to vesicle fractions in TBCE-depleted fractionated cells (Fig. 3C), suggesting that loss of TBCE function impairs v-SNARE recycling to the ER.
To investigate why the v-SNARE-loaded Golgi vesicles accumulate instead of fusing, we assessed v-/t-SNARE pairing by immunoprecipitation and found that in TBCE-depleted cells GS15/Syntaxin 5a and GS28/Syntaxin 5a complexes are less abundant than in control cells (Fig. 3D upper rows), whereas unpaired GS15, GS28 and Syntaxin 5a are more abundant (Fig. 3D, lower rows). High resolution confocal imaging and pixel per pixel analysis supported that GS28/Syntaxin 5a signals are dissociated in TBCE-depleted cells but overlapping in control cells (Fig. 3E). Taken together, these data suggest that ER/Golgi SNARE pairing and recycling is strongly impaired in the absence of TBCE despite the increased protein levels of GS15 and GS28.
Golgi vesiculation in pmn motor neurons is caused by loss of TBCE and not a consequence of neuronal damage
Since Golgi fragmentation is one of the earliest pathological features of degenerating motor neurons in ALS (4), we thought to determine the onset and kinetics of Golgi vesiculation in pmn mice. Motor neuron disease in pmn mice first manifests at Day 15 with skeletal muscle paresis and atrophy and leads to the animal's death 3 weeks later (13). We here demonstrate that the loss of myelinated fibers in the phrenic motor nerve starts at Day 15 and then continues progressively (Supplementary Material, Fig. S3), in line with our earlier studies (20,33). We therefore monitored Golgi vesiculation in pmn and control litter mice at two presymptomatic stages (Days 5 and 10), disease onset (Day 15), midstage (Day 25) and endstage (Day 35). Altered GM130 and GS28 immunofluorescence in motor neurons is first detected at Day 10 (Fig. 4A–B), 5 days before first clinical and histopathological alterations. GM130 and GS28-labeled Golgi vesiculation then progressively affects the majority of motor neurons (Fig. 4A–B), matching a continuous rise of the v-SNARE levels in pmn spinal cords (Fig. 4C).
The presymptomatic onset of Golgi vesiculation in pmn mice suggests that this phenomenon is not a simple consequence of motor neuron damage. We tested this by unilateral transection of the sciatic nerve in wild-type mice and analysis of the retrogradely labeled lesioned motor neurons (Fig. 4D–F). Both p115 labeling (Fig. 4D–E) as well as β-COP, GS15 and GS28 protein levels (Fig. 4F) are unaffected in the lesioned motor neurons when compared with unlesioned ones, indicating that Golgi vesiculation in pmn motor neurons is not a consequence of motor neuron damage but may be a contributor to the disease.
To demonstrate that Golgi vesiculation in pmn motor neurons is due to loss of TBCE function, we complemented pmn mice with wild-type TBCE transgenes (18), which partially or completely restore TBCE expression at the Golgi apparatus (Fig. 4G–I). Transgenic TBCE complementation restores normal p115 labeling in individual motor neurons (Fig. 4G–H) and rescues the pathological spinal cord levels of β-COP, GS15 and GS28 (Fig. 4I). Transgenic TBCE complementation also normalizes the pathologically increased number of GS28-labeled Golgi elements in motor neurons of pmn mice, as shown by quantitation of 3D Golgi modelings (Fig. 4J–K). Whereas the number of small GS28-labeled Golgi elements increases in pmn motor neurons, it goes down to wild-type levels upon TBCE transgene expression in both lines of transgenic TBCE pmn mice, indicating full rescue of Golgi vesiculation. Taken together, these data indicate that loss of TBCE function in pmn motor neurons is responsible for Golgi vesiculation.
TBCE is instrumental for the nucleation and growth of Golgi-derived microtubules
How may loss of TBCE function cause Golgi vesiculation? According to previous studies (16), TBCE is essential for the folding of α-tubulin and its polymerization into microtubules. Tubulin polymerization in differentiated neurons mainly takes place at non-centrosomal sites (34) such as the Golgi (35,36) where TBCE is localized (18). We therefore searched for defects in Golgi-associated microtubules containing detyrosinated (Δ-Tyr) tubulin (37) and found that these are greatly diminished in mutant pmn (Fig. 5A) and TBCE-depleted motor neurons (Fig. 5B–C). To confirm that TBCE is required for the nucleation and growth of Golgi-derived microtubules, we performed washout experiments after treatment with the microtubule-disrupting drug nocodazole (10 µm, 5 h). Nocodazole causes Golgi fragmentation into ministacks (5,6) but does not displace TBCE from the Golgi membrane (Fig. 5D). In control cells, small microtubule asters begin to form immediately after nocodazole washout at dispersed Golgi profiles (Fig. 5E). These microtubules then grow rapidly (Fig. 5F) and reach neighboring Golgi profiles (Fig. 5E, g), leading to re-formation of a compact Golgi ribbon within 200 min (Supplementary Material, Fig. S4). In TBCE-depleted cells however, Golgi-derived microtubules are rare (Fig. 5E), grow much slower than in control cells (Fig. 5F) and only rarely reach neighboring Golgi profiles (Fig. 5E, G) leaving the Golgi disrupted at all time points (Supplementary Material, Fig. S4). These data indicate that TBCE is required for the nucleation and growth of Golgi-derived microtubules.
Accumulation of soluble α-tubulin is not sufficient to trigger Golgi vesiculation
During tubulin folding, α- and β-tubulins first interact with the chaperone proteins prefoldin (38) and cytosolic chaperonin (39), thus generating tubulin folding intermediates (16). α-tubulin folding intermediates interact with tubulin chaperones TBCB and TBCE and β-tubulin folding intermediates with TBCA and TBCD (16,17). Finally, TBCE/α-tubulin, TBCD/β-tubulin and TBCC form a supercomplex that releases native α/β-tubulin heterodimers competent for polymerization into microtubules (40). In line with the function of TBCE in this pathway, we found that TBCE-depleted NSC34 cells contain more soluble α-tubulin (Fig. 6A), less polymerized α-tubulin (Fig. 6A), and alterations of cellular microtubules (Fig. 6C–D). These changes are closely associated with the molecular and structural stigmata of Golgi vesiculation (Fig. 6B–D).
To determine whether an increase in soluble α-tubulin can trigger Golgi vesiculation, we expressed the folding-defective α-tubulin mutant TUBA1 R264C which is unable to stably interact with TBCB (41). In contrast to its wild-type counterpart, flag-tagged mutant TUBA1 is not efficiently polymerized into cellular microtubules (Fig. 6E, H) and accumulates in a soluble form (Fig. 6E). Mutant TUBA1 significantly increases the cellular levels of soluble α-tubulin (Fig. 6F), similarly to TBCE depletion (Fig. 6A), but without affecting the polymerization of endogenous α-tubulin (Fig. 6F) or the structure of the residual microtubule network (Fig. 6H–I). Strikingly, expression of mutant TUBA1 does not alter β-COP, GS15 and GS28 levels (Fig. 6G) or Golgi structure (Fig. 6H–I), indicating that mutant TUBA1 expression is not sufficient to drive Golgi vesiculation. The differential effects of mutant TUBA1 expression and TBCE depletion strongly suggest that Golgi vesiculation is caused by loss of microtubules rather than by accumulation of soluble α-tubulin.
Golgi maintenance depends on an ARF1/TBCE-mediated cross-talk that coordinates COPI vesicle formation and tubulin polymerization at the Golgi
Our data raise the hypothesis that TBCE maintains Golgi structure by adjusting the polymerization of Golgi-derived microtubules to the formation of COPI vesicles. To begin to test this, we modulated the activity of ARF1, the small GTPase that controls COPI vesicle formation (1). Inhibiting ARF1 activity with brefeldin A not only reduces the levels of β-COP at the Golgi, but also shifts TBCE into the cytosol and decreases the number of Golgi-derived microtubules (Supplementary Material, Fig. S5A–B), suggesting that ARF1 regulates TBCE activity.
We next enhanced ARF1 function by (over)expressing wild-type ARF1 or the constitutively active ARF1 mutant Q71L. Both forms of ARF1 accumulate at the Golgi and recruit β-COP (Fig. 7A), as expected. Remarkably, they also lead to an increased recruitment of TBCE to the Golgi membrane, as observed by immunolabeling (Fig. 7A, arrows on right panels), cell fractionation (Fig. 7B) and flow cytometry (Fig. 7C) and to increased polymerization of Golgi-derived microtubules (Fig. 7A and C). Overall, ARF1 Q71L is significantly more effective than ARF1 wt (Fig. 7C). Taken together, these data suggest a cross-talk between ARF1 and TBCE that coordinates tubulin polymerization and COPI formation at the Golgi.
To test whether this cross-talk is involved in Golgi maintenance, we attempted to rescue TBCE-linked Golgi alterations with ARF1. Remarkably, overexpression of ARF1 Q71L after TBCE knockdown strongly prevents alterations of Golgi-derived microtubules (Fig. 7D–E) and rescues defective Golgi architecture labeled by β-COP and GM130 (Fig. 7D–E and data not shown). Flow cytometry suggests that this occurs through ARF1-mediated stabilization of membrane-bound TBCE (Fig. 7F). These data show that defective ARF1/TBCE cross-talk causes Golgi fragmentation in motor neurons.
We here report a novel ARF1/TBCE-mediated cross-talk (Fig. 7G) in which membrane-bound active ARF1 catalyzes COPI coat assembly and recruits TBCE to the Golgi membrane where TBCE drives the nucleation and polymerization of Golgi-derived microtubules. We hypothesize that these microtubules then stabilize the COPI coat through their reported interaction with the β-COP subunit (7). The ARF1/TBCE cross-talk is reminiscent of the regulation of the tubulin-binding co-factor TBCD by the ARF1-related GTPase ARL2 (42). It also mirrors interactions between COPII and the microtubule-dependent motor protein dynein at ER exit sites (43), and between clathrin and actin at the trans Golgi network (44,45), illustrating emerging signaling pathways between vesicular coats, cytoskeleton and motor proteins that ensure efficient coordination of carrier formation and transport in the secretory pathway.
How does an interrupted cross-talk between ARF1 and TBCE cause Golgi fragmentation? We show that absence of TBCE causes loss of Golgi-derived microtubules, β-COP degradation and diminished membrane-recruitment of tethering factors p115/GM130. The degradation of β-COP, ε-COP and possibly other COPI subunits impedes the function of the ER/Golgi intermediate compartment as the target for COPII-derived vesicles (25,26), resulting in accumulation of the latter. This scenario of Golgi fragmentation (Fig. 7G) is recapitulated by β-COP depletion, causing accumulation of small vesicles containing Golgi enzymes at the cell center (22) and redistribution of GM130 (24).
How does loss of TBCE function cause the increased level of Golgi v-SNAREs GS15 and GS28? There are two possibilities. First, GS28 is known to be recycled by COPI vesicles (46) and to be degraded by the proteasome after interaction with the ubiquitin-like (UBL) protein GATE-16 and its regulator ORP7 (47,48). The pmn mutation results in the incorporation of the v-SNAREs in COPII vesicles, possibly away from GATE-16 and its function in proteasomal degradation, hence leading to their accumulation. Second, TBCE itself contains a C-terminal UBL domain which is structurally similar to GATE-16 (T. Levine, personal communication) and might thus directly participate in the v-SNARE degradation. The Trp524>Gly mutation in pmn would destabilize the UBL domain inhibiting its interaction with the proteasome (49) leading to the v-SNARE increase.
Does altered COPI vesicle formation or trafficking cause Golgi fragmentation in other neurodegenerative diseases than pmn? It may be the case in mice mutated in the COPI subunit δ-COP/Archain with Purkinje cell degeneration (50) and in mice mutated in the β-COP-interacting protein Scyl1 such as muscle deficient (51,52) or Scyl1 KO (53) with motor neuron degeneration which, however, need to be investigated for potential Golgi alterations. It is likely the case in motor neurons of mutant SOD1 mice since these display Golgi fragmentation and up-regulation of the microtubule-depolymerizing protein stathmin (8) together with similar dysregulation of β-COP, GS15 and GS28 (Supplementary Material, Fig. S6A-D) and similar GS15- and GS28-labeled Golgi vesiculation (Supplementary Material, Fig. S6E–F) as pmn mice. Defective coordination of vesicle formation and trafficking may thus cause Golgi fragmentation in human SOD-linked ALS and other neurodegenerative disorders.
What may be the contribution of altered COPI vesicle trafficking to motor neuron degeneration and dysfunction? Blocking COPI formation in neuronal cultures by inhibiting ARF1 causes collapse of growing axons (54), suggesting a critical role of COPI vesicles in axon growth or maintenance. The COPI subunit α-COP has been shown to interact with survival motor neuron (SMN), the protein deficient in spinal muscular atrophy, and to co-traffic with SMN in motor axons (55). Interrupting α-COP–SMN interactions leads to SMN accumulation at the Golgi (56), reduced neurite outgrowth (57) and growth cone defects similar to those observed after pharmacological Golgi fragmentation (56). Furthermore, the COPI subunits α-, β- and γ-COP bind numerous mRNAs (58,59) at least some of which require microtubules for their axonal transport (58). The Golgi apparatus is also responsible for packaging precursors of synaptic vesicles which are lost at an early disease stage in mutant SOD1 (60), SMN (61) and pmn mice (B.S., G.H., unpublished data). Defective coordination between the polymerization of microtubules and the formation of COPI vesicles may thus compromise the transport of critical axon and synapse constituents and thereby contribute to motor neuron degeneration and dysfunction.
MATERIALS AND METHODS
Antibodies and reagents
Primary antibodies were as follows [supplier, dilution in immunofluorescence (IF), western blot (WB) and cytometry (Cyt)]: TBCE [rabbit SA53 (18)], IF 1:300, WB 1:500, Cyt 1:50), TBCE (Abnova, IF 1:200), FLAG (Sigma, IF 1:1000, WB 1:2000), GFP (Roche, WB 1:2000), HA (Roche, WB 1:1000, cytometry 1:750), VAChT (Sigma, IF 1:2000), ChAT (Chemicon, IF 1:100), GM130 (Becton Dickinson, IF 1:300, WB 1:500), p115, GS15, GS28, Rab5, Syntaxin 6, Calnexin, Clathrin, Vti1a, (Becton Dickinson; IF 1:300, WB 1:1000), Syntaxin 5a (Abcam, WB1:1000), Sec23 (Abcam 50672, WB 1:1000), β-COP (Abcam, WB 1:2000, IF 1:500, Cyt 1:1000), β-COP (Dr R. Duden, University of Lübeck, WB 1:1000), MG160 (Abcam, IF: 1:500), ARF1 (Dr R.A. Kahn, Emory University, WB 1:1000), Ykt6 (Dr J.C. Hay, University of Michigan, WB 1:2000), MannII (Covance, WB 1:1000), GalT (PTG, WB 1:1000), KDEL (StressGen, IF 1:500), α-tubulin (Sigma, IF 1:2000, WB 1:5000), acetylated-tubulin (Sigma, WB 1:2000), β-actin (Sigma, WB 1:5000), α-tubulin-fluorescein isothiocyanate (FITC) (Sigma, Cyt 1:2000), βIII-tubulin (TuJ1, Babco, IF 1:2000, WB 1:10000), Δ-Tyr-tubulin (Dr A. Andrieux, IF 1:5000, WB 1:10000, Cyt 1:1000), Tyr-tubulin (Dr A. Andrieux, WB 1:2000). Fluorochrome-conjugated secondary antibodies were from Molecular Probes (Carlsbad, CA) and horse radish peroxidase-conjugated secondary antibodies from Jackson Immuno Research (West Grove, PA).
Small interfering (si) RNA duplexes against mouse TBCE (NM_178337, refs D-057655-00 and M-057655-01) or luciferase control (ref P-002099-01-20) were purchased from Dharmacon (Chicago, IL).
si TBCE 5′-GUAGAAGAGUUGAAGUUAA
si TBCE mix 5′-GUAGAAGAGUUGAAGUUAA,
si luciferase 5′-CAUUCUAUCCUCUAGAGGAUG
Plasmid expression vectors contained the following cDNAs: GFP-MannII (Dr M. Bornens, Institut Curie, Paris), TUBA1A wt or TUBA1A R264C (Dr J. Chelly, Institut Cochin, Paris), ARF1 wt, Q71L and T31N (Dr J. Donaldson, NIH, Bethesda).
Additional reagents were from the following suppliers: phosphate buffered saline (PBS), Hank's balanced salt solution, trypsin, culture media and supplements (Invitrogen, Carlsbad, CA), polyornithin, taxol, nocodazole (Sigma), Vectashield (Vector laboratories, Burlingame, CA), Complete protease inhibitors (Roche, Basel, Switzerland), Ketamine (Bayer, Leverkusen, Germany) and Xylazine (Mérial, Lyon, France), coverslips and Superfrost Plus glass slides (Menzel, Schwerte, Germany).
Mutant pmn mice, TBCEPA/+pmn and TBCEPC/+pmn mice were maintained on a mixed background (C57/BL6; 129/SvJ) using intercrosses (>F8) and genotyped by PCR (18).
Axotomy and retrograde labeling of motor neurons
Mice were anesthetized by intraperitoneal injection of Ketamine (50 mg/kg) and Xylazine (10 mg/kg). Under a stereomicroscope, a skin incision was made, the sciatic nerve identified and cut at the hip level. A piece of hemostatic sponge (Spongostan, Ferrosan, Denmark) 2 × 2 × 2 mm in size was soaked in tetramethylrhodamine–dextran (Molecular Probes, 10% w/v in PBS) and applied to the proximal stump. The wound was closed with metal clips and an antibiotic prophylaxis with ampicillin administered. All experiments with animals were performed in strict compliance with French and European legislation.
Confocal imaging, morphometry and 3D modeling
Images were obtained with an LSM 510 confocal microscope (Zeiss, Oberkochen, Germany). Motor neurons on spinal cord tissue cryosections were identified by staining for the motor neuron markers VAChT (vesicular acetylcholine transferase) or ChAT (choline acetyltransferase). Cells were imaged by confocal microscopy using a ×40 water immersion objective, an xy-resolution of 1024 × 1024 pixel, a z-interval of 0.3 µm and a mean scan depth of 30 µm.
Morphometric analyses were done with Metamorph software (Molecular Dynamics) using single confocal cross sections taken at the nuclear midplane. The boundaries of the ChAT-labeled cell soma and the DAPI-stained nucleus were manually delineated with the Metamorph drawing tool. Golgi surface area was automatically determined with the Metamorph morphometry tool by applying a fixed threshold to the MG160 signal.
Three-dimensional (3D) modelings of the Golgi apparatus were done with Imaris software (Bitplane, Zurich, Switzerland). Images were processed and Golgi membranes visualized using the IsoSurface mode of the Surpass module. The number of individual Golgi elements per motor neuron was determined in a blinded manner. Fluorescent line intensity profiles were determined with the linescan module of Zen software (Zeiss) on confocal z-stacks of NSC34 cells (z-interval 0.3 μm, scan depth 10 μm).
Deeply anesthetized mice were transcardially perfused with Sorensen's phosphate buffer (pH 7.4) followed by glutaraldehyde (2% v/v in cacodylate). Spinal cords were dissected out, postfixed for 24 h and cut into small segments comprising the ventral spinal cord before processing for resin embedding (epon) following standard protocols. Sixty nanometer ultrathin cross sections were contrasted with uranyl acetate and visualized under a JEOL electron microscope. At least two lumbar spinal cords from each genotype were analyzed and three sections (each containing up to 30 motor neurons) were analyzed per time point. Motor neurons were recognized on the basis of their frequency (1:20), large size and pale nucleus. Within individual motor neurons, all Golgi profiles were categorized.
NSC34 motor neurons were cultured in DMEM supplemented with 10% (v/v) fetal calf serum at 37°C and 7.5% CO2 and transfected with siRNAs and/or DNA plasmids using Lipofectamine 2000 (Invitrogen) as described (18).
Immunoblots and subcellular fractionation
Lumbar spinal cords from deeply anesthetized mice and NSC34 cells were homogenized in lysis buffer containing 50 mm Tris–HCl pH 7.5, 150 mm NaCl, 2 mm ethylenediaminetetraacetic acid (EDTA), 1% Triton X-100, Complete antiproteases and, for GM130, also 0.25% sodium dodecyl sulfate (SDS). 30 or 50 µg protein were subjected to SDS-polyacrylamide gel electrophoresis and blotted on Immobilon membranes (Millipore) which were processed by standard methods, reacted with ECL or ECL plus kits (Amersham) and revealed with XAR films (Kodak) or a Biorad imager. Band intensities were quantified by TotalLabQuant software.
Crude fractionation of membranes from lumbar spinal cord was done after tissue freezing (−80°C), thawing and homogenization in 50 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), pH 7.4, 250 mm sucrose, 1 mM Mg acetate and protease inhibitors (Complete, EDTA-free). Lysates were homogenized using a Dounce homogenizer (15 passes) and centrifuged at 1 000 g for 10 min. The postnuclear supernatant was centrifuged at 10 000 g for 30 min at 4°C yielding a P10 pellet and the supernatant was centrifuged at 100 000 g (Beckman TLA-110) for 1 h at 4°C yielding an S100 supernatant and a P100 pellet.
For membrane fractionation on sucrose density gradients, cells were homogenized in an isotonic sucrose solution by 10 passages through a Dounce homogenizer and centrifuged at 1 000 g for 10 min. The postnuclear supernatants were centrifuged at 190 000 g (43 000 rpm) for 1 h in a Beckman SW60 Ti rotor; 12 fractions of 330 µl each were collected from top to bottom and equal volumes analyzed by immunoblot. Sucrose concentrations in fractions and standards were determined by an enzymatic assay (Sigma, ref. SCA-20) using a Biochrom WPA spectrophotometer. Subcellular fractionation of NSC34 cells into membrane and cytosolic fractions was done as described (21).
Microtubule re-growth assay
Growth dynamics of Golgi-derived microtubule were studied by modifying an earlier reported procedure (18). NSC34 cells transfected with siRNAs and a GFP-MannII plasmid were treated at 5 days in vitro (DIV) with nocodazole (10 µm, 5 h, 37°C). Nocodazole was then washed out with warm culture medium, and neurons further incubated at 37°C. At time intervals ranging from 0 to 200 min, cultures were rinsed with PHEM 60 mm piperazine-N,N'-bis(2-ethanesulfonic acid), 25 mm HEPES, 10 mm ethylene glycol tetraacetic acid, 2 mm MgCl2, 1% formaldehyde (pH 6.9) and then treated for 3 min with PHEM containing 0.2% Triton X-100 and 20 µm taxol to extract soluble proteins and stabilize microtubules. Cells were fixed, blocked and counterstained for α-tubulin. After confocal imaging of entire cells, number and length of Golgi-derived microtubules were determined using Metamorph and ImageJ software respectively.
Transfected NSC34 cells from 6-well plates were harvested in PBS using a cell scraper and lysed on ice for 10 min in 50 mm Tris pH 7.5, 150 mm NaCl, 5 mm EDTA, 1% Triton X-100 and Complete protease inhibitors. After removal of cell debris, 300 µl of cell lysates were incubated with 3 µg antibodies and 30 µl protein G-magnetic beads (Millipore) on a tube rotator for 4 h. Supernatants and immune complexes were then collected using a magnetic stand holder.
Transfected cells were harvested at 3 DIV (ARF1) or 5 DIV (siTBCE), rinsed with PBS, centrifuged at 250g, washed with PHEM and treated for 3 min with PHEM containing 0.1% Triton X-100 and 20 µm taxol to extract soluble proteins and to stabilize microtubules. Cells were fixed for 15 min at RT by adding an equal volume of 8% (v/v) FA in PBS, rinsed in PBS, centrifuged and blocked for 30 min in 5% goat serum, 1% (w/v) BSA and 0.1% (v/v) Triton X-100. Cells were incubated for 1 h with primary antibodies and for 1 h with secondary antibodies or biotin/streptavidin reagents coupled to Alexa-594 (HA) or Alexa-488 (β-COP, TBCE and ΔTyr-tubulin). One thousand five hundred cells per duplicate sample and condition were analyzed with a FACS ARIA SORP (Becton Dickinson) and signals plotted with FlowJo software.
Each experiment was performed with several biological replicates, i.e. usually three pairs of wild-type and pmn mice and duplicate or triplicate cultures of control and siTBCE-transfected cells. All experiments were repeated once or twice. Data were analyzed with Microsoft Excel, SigmaStat 3.1 (Systat, Evanston, IL) or GraphPad Prism (GraphPad). Data from two groups showing Gaussian distribution were analyzed with Student's t-test; otherwise the Mann–Whitney U test was used. Data from more than two groups showing Gaussian distribution and equal variance were analyzed with one-way ANOVA test and Tukey's post hoc test; otherwise the Kruskal–Wallis test and Dunn post hoc test were used. Cytometry data were tested for significance with the χ2 test using FlowJo software.
Work in G. Haase's laboratory is supported by grants from Association Française contre les Myopathies (AFM), Agence Nationale pour la Recherche and ERANET Neuron. Work in C. Rabouille's laboratory is supported by NWO. S.B. was supported by student fellowships from Fondation pour la Recherche Médicale (FRM) and AFM. M.S. was supported by postdoctoral fellowships from Deutsche Forschungsgemeinschaft and INSERM.
We gratefully acknowledge the expert help of D. Xanthakis (Department of Cell Biology, UMC Utrecht, NL) in processing, sectioning and electron microscopic analysis of tissue samples, the expert help of P. Weber (IBDML, Marseille, France) and A. Bernadac (CNRS, Marseille, France) in confocal microscopy and imaging software and the contribution of Dr H. Schmalbruch (University of Copenhagen) to earlier stages of the work. We are grateful to Dr A. Spang for critical reading of the manuscript and to Drs A. Andrieux, M. Bornens, J. Chelly, J. Donaldson, R. Duden, J.C. Hay and R.A. Kahn for providing essential reagents.
Conflict of Interest Statement. None declared.