Dynamin (Dyn) is a multidomain and multifunctional GTPase best known for its essential role in clathrin-mediated endocytosis (CME). Dyn2 mutations have been linked to two human diseases, centronuclear myopathy (CNM) and Charcot-Marie-Tooth (CMT) disease. Paradoxically, although Dyn2 is ubiquitously expressed and essential for embryonic development, the disease-associated Dyn2 mutants are autosomal dominant, but result in slowly progressing and tissue-specific diseases. Thus, although the cellular defects that cause disease remain unclear, they are expected to be mild. To gain new insight into potential pathogenic mechanisms, we utilized mouse Dyn2 conditional knockout cells combined with retroviral-mediated reconstitution to mimic both heterozygous and homozygous states and characterized cellular phenotypes using quantitative assays for several membrane trafficking events. Surprisingly, none of the four mutants studied exhibited a defect in CME, but all were impaired in their ability to support p75/neurotrophin receptor export from the Golgi, the raft-dependent endocytosis of cholera toxin and the clathrin-independent endocytosis of epidermal growth factor receptor (EGFR). While it will be important to study these mutants in disease-relevant muscle and neuronal cells, given the importance of neurotrophic factors and lipid rafts in muscle physiology, we speculate that these common cellular defects might contribute to the tissue-specific diseases caused by a ubiquitously expressed protein.
Eukaryotic cells are characterized by their highly specialized organelles with unique protein and lipid compositions. Accurate and efficient membrane trafficking among these organelles is critical for their functions and important for human health (1). GTPases are important regulators in multiple membrane trafficking events (2–4). The large GTPase dynamin (Dyn) is best known for its role in clathrin-mediated endocytosis (CME) (5–8), but it has also been implicated in other cellular processes, including cytokinesis, export from the Golgi, caveolae-dependent and/or lipid raft-dependent endocytosis, macropinocytosis and microtubule stability (9–15). Mammals express three Dyn isoforms in a tissue-specific manner, of these dynamin 2 (Dyn2) is ubiquitously expressed (16).
Mutations in the gene encoding human Dyn2 (DNM2) have been linked to two neuromuscular diseases, the intermediate and axonal forms of Charcot-Marie-Tooth (CMT) neuropathies and centronuclear myopathy (CNM) (17–20). Thirty-one genes have been linked to CMT (21,22), which encode for proteins that are important for many different aspects of cellular function, including membrane structure, transcription, translation, signal transduction and membrane trafficking. CNM is a slowly progressive congenital muscular disorder characterized by a high rate of centrally localized nuclei in muscle fibers. Mutations in several proteins are linked to CNM, including myotubularin, amphiphysin 2 and sporadic mutations in ryanodine receptor 1 and myotubularin-like protein (19,23). For both diseases, the age of onset and syndrome severity vary between patients, even for those carrying the same mutant alleles. To date, for both CMT and CNM, the disease-related Dyn2 mutations are all autosomal dominant.
Dyn is composed of five functional domains: an N-terminal GTPase domain, followed by the middle and pleckstrin homology (PH) domains, the GTPase effector domain (GED) and the C-terminal proline/arginine domain (PRD) (Figure 1A). The GTPase domain binds and hydrolyzes GTP (24). The middle domain and GED interact with each other and are involved in higher order self-assembly (25). The PH domain binds phosphatidylinositol (PI) lipids through three variable loops (VL1–3) that form a Phosphatidylinositol-(4,5)-bisphosphate binding pocket (26). The PRD binds to SH3 domain-containing partners and is involved in targeting Dyn to clathrin-coated pits (CCPs) (5,27). While the structure of full-length Dyn is not yet available, high-resolution structures for a minimal Dyn1 GTPase–GED fusion (24), the PH domain (28,29), and for the middle domain/GED of the Dyn-related MxA protein (30) have been solved.
The first Dyn2 mutations linked to CMT and CNM were found to cluster in the PH and middle domains, respectively (17,18). However, several other mutations, scattered throughout the protein, have also been identified (20). To characterize the cellular defects associated with CMT- and CNM-linked Dyn2 mutations, several groups have transiently overexpressed the mutant proteins and studied the resulting phenotypes in cultured cells (15,17,18,31). These studies have shown dominant-negative effects on CME (15,31), microtubule dynamics (15), epidermal growth factor (EGF) signaling (31) and centrosome cohesion (18). However, a problem with these studies is that the phenotypes observed at the high levels of protein obtained by transient overexpression might not reflect the physiologically relevant effects of these autosomal dominant alleles. In addition, the strong cytotoxicity of Dyn2 (32) may cause non-specific effects when overexpressed.
To overcome these problems, we utilized retroviral vectors to express disease-related Dyn2 mutants at near-endogenous levels in conditional null Dyn2 mouse fibroblasts. This system has allowed us to study Dyn2 mutant effects under both disease-relevant heterozygous conditions and after Cre recombinase-mediated knockout (KO) of endogenous Dyn2. In addition, to reveal subtle defects, we have employed sensitive and quantitative assays to analyze several membrane trafficking events in vivo.
Generation of disease-related Dyn2 mutant cell lines
We previously generated Dyn2 conditional null (Dyn2 flox/null) mouse fibroblasts (herein referred to as Dyn2+/− cells) and showed, after excision of the remaining Dyn2 allele with Cre recombinase, that Dyn2 null cells had defects in CME, cytokinesis, export from the Golgi and macropinocytosis, despite the fact that these cells also expressed Dyn1 (14). All these defects were rescued by reconstitution with Dyn2 or Dyn2-GFP at physiological levels (14). Thus, we used this system to characterize the cellular phenotype(s) associated with CNM- and CMT-linked Dyn2 mutations.
Two CNM mutations and two CMT mutations were chosen for study based on residue locations and/or their frequency found in patients (Figure 1A). We chose the Dyn2 CNM mutant p.Arg465Trp (R465W) because it is the most common allele, and p.Glu368Lys (E368K) because this charge conversion is expected to have greater effects. By homology to MxA (30), these mutations map on the outward facing sides of the middle/GED dimer stalks (Figure 1B). The CMT-linked triple amino acid deletion, p.Asp555_Glu557del (Δ3), in the strand between VL1 and VL2 is one of the most common alleles, and the p.Leu570His (L570H) mutant maps to the VL2 (Figure 1C). Previous studies have shown that both the VL1 and VL3 are important for Dyn function in mediating membrane fission (33,34). To generate stable cell lines expressing the disease-related Dyn2 mutants, Dyn2+/− cells were infected with retrovirus encoding C-terminally green fluorescent protein (GFP)-tagged wild-type (WT) or mutant Dyn2 proteins for chromosomal integration of these constructs. Cells expressing comparable levels of WT and mutant Dyn2-GFP were selected via fluorescence-activated cell sorting (FACS) as previously described (14).
We initially selected cell lines based on similar GFP intensity; however, after culturing for 1 week to expand the sorted cells, the expression levels of all Dyn2 mutants were significantly lower (about 60–70%) than those expressing WT Dyn2-GFP (data not shown). This observation likely reflects the impairment of some function(s) in cells expressing higher levels of these dominant alleles and the resulting outgrowth of a healthier subpopulation of cells expressing lower levels of these Dyn2 mutants. Therefore, to directly compare the activity of WT Dyn2 to these mutant alleles, we selected cells expressing similar levels of Dyn2-GFP WT (Figure 1D). These reduced levels of expression, corresponding to 23–44% relative endogenous Dyn2 (Figure 1D), subsequently remained stable in culture for multiple passages. However, because we were unable to fully reproduce the 1:1 heterozygous conditions in the disease state, we also examined the ability of these hypomorphic alleles to support Dyn2 cellular functions after deletion of endogenous Dyn2 (Figure 1E). These experimental conditions should reveal cellular defects that may be too subtle to detect, given the ratio of mutant:endogenous Dyn in the heterozygous cell lines. Importantly, even at these low levels of expression, all the Dyn2 mutants were able to rescue the previously reported growth defect (14) of Dyn2 KO cells (data not shown), and the levels of Dyn2 mutant expression remained unchanged after Cre-mediated excision of endogenous Dyn2 (Figure 1E).
Subcellular localization of CMT- and CNM-related Dyn2 mutants
Endogenous Dyn2 is targeted to both CCPs at the plasma membrane (PM) and to the trans Golgi network (TGN) (Figure S1). Dyn2 exists as four splice variants: aa, ab, ba and bb, and targeting to the TGN is dependent on the splice variant of Dyn2 (14). Therefore, for these studies we have used the ba splice variant, which is efficiently targeted to both PM CCPs and the TGN (14), and expressed equally with other splice variants in both human skeletal muscle and peripheral nerve (31).
We first compared the subcellular localization of WT and the mutant Dyn2-GFP using fluorescence microscopy in fixed cells without (Figure 2A) or with (Figures 2B and 3) prior permeabilization to reduce cytosolic pools. The micrographs in Figure 2 show Dyn2-GFP localization in Dyn2 KO cells to avoid potential effects of hetero-oligomerization with endogenous Dyn2. Those in Figure 3 show the distribution of Dyn2-GFP relative to total Dyn2 in Dyn2+/− cells (Figure 3). Consistent with previous reports (14,35), WT Dyn2-GFP expressed in our stable cell lines could be detected diffusely in the cytosol, as well as enriched at PM CCPs [colocalizing the adaptor protein 2 (AP2)] and at the TGN [partially colocalizing with adaptor protein 1 (AP1)]. A nearly identical distribution was observed for total Dyn2 (Figure 3), consistent with the ability of Dyn2-GFP to fully rescue the Dyn2 null phenotype. Although all the Dyn2 mutants retained their ability to be localized to PM CCPs (Figure 2A), they differed in their extent of localization to the TGN (Figure 2B). The PH domain mutants retained their TGN localization, consistent with previous studies showing that the lipid-binding properties of the PH domain are not required for Dyn targeting to CCPs in vivo(34,36). In contrast, the middle domain mutants showed little overlap with AP1. As expected, these dominant mutants also significantly reduced the intensity of total Dyn2 staining at the TGN (Figure 3). Given that the localization of Dyn can be affected by both splice variations (14) and point mutations (Figure 2B) within the middle domain, we conclude that, in addition to self-assembly (25), the middle domain plays a role in targeting Dyn to the Golgi.
Contrary to previous reports (15), the structure of the Golgi apparatus, as detected by staining either the TGN-associated AP1 γ-adaptin (Figure 2B) or the cis Golgi protein, GM130 (Figure S2), was not affected in either the Dyn2 KO cells (14) or in cells expressing any of the Dyn2 mutants under these conditions (i.e. in the absence of endogenous Dyn).
Disease-related Dyn2 mutants impair p75/neurotrophin receptor export from the Golgi
Given the observed alterations in TGN targeting, we next examined the ability of Dyn2 mutants to mediate transport of the neurotrophin receptor, p75, from the Golgi to the PM, a Dyn2-dependent trafficking process (10,14,37). On the basis of the tissue specificity and slow progressive pathology of many patients with these diseases, we expected that these disease-related mutant Dyn2-associated phenotypes might be milder than those observed upon overexpression of strong dominant-negative Dyn2 mutants. Therefore, we developed a new and more quantitative strategy to measure trafficking of p75 from the TGN to the PM. For this we utilized a system for site-specific biotinylation via peptide-based tags (38). A 15-amino acid acceptor peptide (AP) sequence was introduced into p75 just after its signal sequence (Figure 4A), and the protein was tagged at its C-terminus with GFP. A tet-regulatable adenoviral expression vector encoding AP-p75-GFP and an endoplasmic reticulum (ER)-localized Escherichia coli enzyme biotin ligase (BirA) driven by an internal ribosome entry site (IRES) was used to express monobiotinylated AP-p75-GFP in the presence of biotin. Trafficking of this double-tagged transmembrane protein could be monitored using fluorescence microscopy (Figure 4B). Cells were coinfected with adenovirus encoding AP-p75-GFP and the tetracycline-regulatable transactivator (tTA) overnight in the presence of tetracycline to suppress expression. Tetracycline was removed and after 5-h induction of AP-p75-GFP and BirA expression, biotinylated p75-GFP localized mostly in the ER and cis Golgi (Figure 4B, panel a). After an additional 3-h incubation at 20°C, the protein accumulated at the TGN (Figure 4B, panel b). Subsequent shift to 32°C allowed for the time-dependent delivery of p75 from the TGN to the PM, detected as diffuse surface labeling (Figure 4B, panels c–e). The kinetics of delivery of biotinylated p75-GFP from the TGN to the PM were indistinguishable for those previously reported for p75-GFP (14,37), indicating that the biotinylation did not interfere with the proper trafficking of p75.
To estimate the relative kinetics of delivery to the PM, cells were transferred to 4°C after incubation at 32°C for 40 min, a relatively early time-point in surface appearance (Figure S3). Surface-accessible biotinylated p75-GFP was detected with Alexa-568-labeled streptavidin. After extensive washing, the cells were lysed in detergent. The ratio of Alexa-568 (surface)/GFP intensity (total) in the different cell lysates provides a quantitative measurement for the efficiency of delivery to the PM. The export of biotinylated p75-GFP from the TGN and delivery to the PM was indistinguishable in Dyn2+/− cells expressing either WT or the mutant Dyns (data not shown), indicating that under these conditions (∼1:3 ratio mutant:WT protein) the mutants did not exhibit a dominant-negative effect. Therefore, we tested their ability to support p75 transport after excision of endogenous DNM2 with Cre recombinase. None of the Dyn2 mutants were as effective as WT in rescuing p75-GFP export from the Golgi, including the PH domain mutants that were still targeted to the TGN (Figure 4C).
To test whether these mutants might have a dominant-negative effect on p75 trafficking if expressed at higher levels, we used the tet-regulatable adenoviral expression system to coexpress a subset of these Dyn2 mutants, as hemagglutinin (HA)-tagged proteins, together with AP-p75-GFP in tTA-expressing HeLa cells following with the same procedure as described earlier to quantify the export efficiency of biotinylated p75-GFP. Under these conditions, the HA-tagged Dyn2 is expressed at approximately equal levels to endogenous Dyn2 (Figure S3A). As can be seen, each of these Dyn2 mutants significantly inhibited p75-GFP export in HeLa cells (Figure 4D). Closer examination of the effects of one of these mutants, Dyn2-R455W, on the kinetics of p75 transport from the Golgi showed a clear delay of p75 export compared with Dyn2WT (Figure S3B). These results establish that the disease-related Dyn2 mutants share a defect in their ability to support trafficking of the p75/neurotrophin receptor from the TGN to the PM. Importantly, at equal levels of expression the mutants have a dominant-negative effect on endogenous Dyn2 activity.
The disease-related Dyn2 mutants do not affect microtubule acetylation or macropinocytosis
Dyn1 was originally identified based on its ability to bind microtubules (MTs) in vitro(39) and Dyn2 shares this ability (40). While there is little evidence for the in vivo association of WT Dyn with MTs, overexpressed mutants of Dyn2, including Dyn2ΔPRD (41), and interestingly, the CMT-linked mutant, Dyn2Δ3, have been shown to colocalize with MTs (15). It has also been suggested that Dyn2 plays a role in regulating the dynamic instability of MTs (15). Indeed, this study showed that transient overexpression of Dyn2Δ3 led to an accumulation of acetylated tubulin, an indicator of stabilized MTs. Increased acetylation of tubulin was also observed upon siRNA knockdown of endogenous Dyn2 (15).
The defect of these disease-associated Dyn2 mutants in p75 transport from the TGN to the PM did not correlate with their targeting efficiency to the TGN. Thus, we wondered whether it might reflect an indirect consequence of a role for Dyn2 in microtubule dynamics rather than in vesicle formation at the TGN. We first confirmed previous findings regarding the accumulation of acetylated tubulin in our Dyn2 null cells by indirect immunofluorescence staining of fixed cells and western blotting of cell lysates (Figure S4A,B). However, as for WT Dyn2, all the disease-related Dyn2 mutants, including Dyn2Δ3, reduced the amount of acetylated tubulin back to control levels. Thus, the defect in p75 trafficking is unlikely to be a result of the acetylated tubulin accumulation and altered microtubule dynamics.
Dyn has also been reported to directly or indirectly regulate actin reorganization (42,43). Consistent with this, we previously showed that Dyn2 null cells were defective in platelet-derived growth factor (PDGF)-induced, actin-based membrane ruffle formation and macropinocytosis (14). In the absence of PDGF, the mutant Dyn2-GFP-expressing cells displayed normal actin structures, mostly actin stress fibers, even in the Dyn2 null background (data not shown). Upon 5-min stimulation with PDGF, all disease-related Dyn2 mutant cell lines showed dorsal ruffle formation (Figure S4C) and rates of macropinocytosis of horse radish peroxidase (data not shown) that were indistinguishable from Dyn2 WT-expressing cells. Dyn2-GFP was localized to the membrane ruffle in the presence or absence of endogenous Dyn2, although the two middle domain mutants (E368K and R465W) showed significantly decreased enrichment to the dorsal ruffle compared with WT and the two PH domain mutants. Thus, again the middle domain mutants appear to alter Dyn2 membrane targeting. These results confirm that Dyn2 is important for both actin and microtubule dynamics; however, the disease-related Dyn2 mutations studied here retain these activities.
The disease-related Dyn2 mutants can fully support CME
The best-known functions of Dyn are in regulating CME and catalyzing membrane fission leading to clathrin-coated vesicle formation (1–4). Overexpression of some, although not all, disease-related Dyn2 mutants has been shown to inhibit CME (15,17,31). Therefore, we next tested the ability of the disease-related Dyn2 mutants either to act as dominant-negative inhibitors of CME in the presence of endogenous WT Dyn2 or to support CME in its absence. CME was quantitatively assessed using an enzyme-linked immunosorbent-based assay (ELISA) that measures the internalization of biotinylated transferrin (bTfn) into avidin-inaccessible endocytic compartments (44). The results show that CME was unaffected in cells expressing either of the autosomal dominant CMT- and CNM-associated Dyn2 mutations in the presence of endogenous Dyn2 (Figure 5A).
To further confirm that the lack of a dominant-negative effect on CME was not due to the low expression of these Dyn2 mutants, CMEs were analyzed in tTA-HeLa cells infected with tet-regulatable adenoviruses and expressing WT and Dyn2 mutant proteins at fivefold over endogenous levels (Figure S3C). Once again, under these conditions, neither of the CMT- or CNM-causing mutants are dominant-negative inhibitors of CME (Figure 5B).
Finally, just as for p75 export, we tested whether these Dyn2 mutants could support CME in Dyn2 null cells, as this could reveal potential subtle defects not detected when assayed in the presence of endogenous Dyn2. Surprisingly, all the Dyn2 mutants tested could fully rescue the bTfn uptake defect of Dyn2 null cells (Figure 5C). Together, these results indicate that constitutive CME is not a common defect associated with disease-related Dyn2 mutations, and thus is unlikely to contribute to the pathology of these diseases.
Overexpression of a Dyn1 PH domain mutant (Y597F) was shown to selectively impair internalization of the epidermal growth factor receptor (EGFR) without affecting TfnR uptake (45). Endocytosis of receptor tyrosine kinases, such as the EGFR, through clathrin-dependent or independent mechanisms is known to regulate their downstream signaling (46–48). Slight modifications in membrane trafficking might alter specific signaling pathways and thus could account for the disease phenotype. Under the conditions used in these experiments [i.e. in the presence of 10 ng/mL EGF; (45)], EGFR can be internalized through both clathrin-dependent and independent pathways (49–51). Thus, we next tested whether these Dyn2 mutants might also specifically inhibit EGFR versus TfnR uptake.
We first confirmed, using an ELISA, that at low concentrations (2 ng/mL biotinylated EGF, bEGF) the internalization of EGF into these mouse fibroblasts was unaffected by filipin, an inhibitor of raft-dependent endocytosis, whereas at 10 ng/mL the EGF uptake was partially inhibited by filipin (Figure S5A). Then, we tested the effects of the CNM and CMT mutants on bEGF internalization at 10 ng/mL and found that, in contrast to bTfn endocytosis, all exhibited a significant dominant-negative effect on EGF uptake (Figure 6A), even when expressed at these low levels relative to endogenous Dyn2. Given that the delivery of at least one receptor, p75, to the PM is impaired in these mutant cells, we confirmed that there was no significant difference in total surface binding of bEGF under these conditions. Moreover, the data in Figure 6A are normalized to total EGF surface binding and thus measure endocytosis independently of changes in surface EGFR expression. The extent of inhibition (∼20%) was similar to that observed with filipin under these conditions (Figure S5A). We also confirmed that none of the disease-related mutants were able to fully rescue the bEGF uptake defect in Dyn2 null cells (Figure S5B), as compared with WT Dyn2. Together, these data suggest a specific defect in the clathrin-independent pathway for EGFR uptake.
Fluorescently conjugated cholera toxin B (Alexa-594-CTB) binds to the sphingolipid GM1, a well-established marker for raft-dependent endocytic pathways (52,53). We had previously shown that this pathway was not impaired in Dyn2 null cells, but was inhibited by overexpression of dominant-negative Dyn mutants (14). From this we concluded that endogenous Dyn1 was sufficient to support CTB uptake in the absence of Dyn2. The data in Figure 6B show, using fluorescence microscopy, that the CNM- and CMT-Dyn2 mutants also inhibit the raft-dependent endocytosis of Alexa-594-CTB. To quantify the CTB internalization efficiency, the cells were analyzed for total surface-bound and internalized Alexa-594-CTB, as described in Materials and Methods. All the cells expressing the disease-related Dyn2 mutants exhibited an ∼30% reduced ability to internalize CTB (Figure 6C). Thus, another common property of all the disease-related Dyn2 mutations we have studied is their ability to inhibit raft-dependent endocytosis in the presence of endogenous Dyn2.
Disease-related Dyn2 mutants alter EGFR downregulation
A recent study has shown that EGFRs internalized via a clathrin-independent, filipin-sensitive pathway are more effectively targeted for degradation (50), whereas EGFRs internalized by CME are more efficiently recycled. As these disease-related Dyn2 mutants appear to selectively inhibit the former pathway, we examined their effects on EGF-dependent EGFR downregulation. As predicted from previous results (50), we also found that the extent of EGF-induced EGFR degradation at 90 min was significantly reduced in cells expressing the Dyn2 mutants compared with WT cells (Figure 7A,B).
The altered trafficking of EGFR in these mutant cells could affect downstream signaling (50), which might contribute to disease. While this hypothesis merits further exploration, especially in relevant muscle and neuronal cells, we did not detect a dominant-negative effect on extracellular signal-regulated kinase (ERK) activation in Dyn2+/− cells expressing the Dyn2 mutants (Figure S5C,D). However, ERK activation was decreased in Dyn2 KO cells, and the expression of disease-related Dyn2 mutants in the absence of endogenous Dyn2 resulted in slightly, but nonetheless significantly higher levels of ERK activation than in either control (Dyn2+/−) cells or KO cells reconstituted with WT Dyn2 (Figure S5E). While further analysis is needed, these data are nonetheless consistent with a selective, but mild impairment of a clathrin-independent, filipin-sensitive pathway for EGFR endocytosis that can modulate EGFR signaling. That these effects are seen with all the disease-related Dyn2 mutants studied suggests that this cellular phenotype might also be linked to the disease pathogenesis.
Although best known for its role in CME, our understanding of the versatility of Dyn's functions in vivo is still expanding. The discovery of Dyn2 mutations in neuronal and muscular diseases raises even more questions: (i) Is there a common cellular defect associated with different Dyn2 mutations? (ii) What is the pathological mechanism of Dyn2 mutations that cause disease? (iii) Dyn2-related CMT and CNM are both congenital, autosomal dominant and tissue-specific diseases, despite the fact that Dyn2 is ubiquitously expressed and required for embryonic development (16,54). Thus, how do dominant mutations in a ubiquitously expressed protein cause tissue-specific diseases? Given the relatively mild and clinically variable effects of these disease-causing mutants, they must be hypomorphic alleles with subtle perturbations in Dyn2 functions.
To answer these questions, we established stable cell lines from Dyn2 conditional KO mouse fibroblasts to express the disease-related mutant proteins in the presence and absence of endogenous Dyn2 to mimic both heterozygous and homozygous states. We also utilized quantitative membrane trafficking analyses to identify potentially subtle cellular defects associated with disease-related Dyn2 mutants in both the heterozygous and homozygous states.
Common cellular defects associated with Dyn2 mutations linked to two different human diseases
Through this analysis, we identified two membrane trafficking pathways that were perturbed in a dominant-negative manner by all the disease-linked Dyn2 mutants studied, thereby implying a common pathological mechanism. These included the clathrin-independent endocytosis of CTB and EGFR and the export of p75/neurotrophin receptor from the Golgi and its delivery to the PM. Importantly, the effects of the Dyn2 mutants on p75 and EGFR trafficking were not severe, perhaps consistent with the slow progression and tissue specificity (discussed below) of the diseases. However, we cannot rule out that other and/or more severe defects might be revealed if the mutants were studied in neuronal or muscle cells.
It was perhaps unexpected that mutants from two different diseases result in same cellular defects. However, subtle nerve abnormalities have been observed in some Dyn2-CNM patients (55,56), suggesting some commonality in pathogenesis. Interestingly, different classes of Dyn2 mutants might exert their dominant effects on p75 export by different mechanisms. The PH domain CMT-related mutations that colocalize at the TGN with WT Dyn2 may interfere directly with its function, whereas the middle domain CNM-related mutations, which are not localized to the TGN, appear to interfere with targeting of WT Dyn2 to the TGN. Both dominant-negative effects could occur through hetero-oligomerization with WT Dyn2 to alter its activity and/or localization. Further studies are needed to define the biochemical properties of these disease-related Dyn2 mutations and possible molecular mechanisms of Dyn2-mediated diseases.
Lipid rafts and potential disease pathogenesis
Lipid rafts are sphingolipid- and cholesterol-rich domains of the PM that contain a variety of signaling and transport proteins. Many diseases, including cancer, cardiovascular diseases, neurological diseases and muscle disorders, have been linked to defects in raft organization and/or function (57,58). In muscles, mutations in caveolin-3, which is required for normal, lipid raft-dependent T-tubule development, cause four distinct muscle diseases (59–61). Mutations in amphiphysin 2, a Dyn-interacting BAR-domain containing protein, have also been suggested to affect T-tubule remodeling (62,63) and/or the endosomal membrane (64) and can result in CNM (65). Another CNM-causing gene, myotubularin (19,66), is a lipid phosphatase that is important for regulating vesicle trafficking from the endosome to the lysosome (67,68). These studies indicate an important role for lipid rafts and membrane trafficking in the maintenance of normal muscle function. Thus, the partial inhibition of a clathrin-independent pathway for EGFR uptake and raft-dependent endocytosis of CTB in mutant Dyn2 cells may contribute to the pathogenesis of muscular weakness and wasting.
In addition to membrane trafficking and organization, lipid rafts are also important for cell signaling, either as organizing centers for signaling molecules or for receptor internalization (69,70). Recently, the effects of myopathy-associated caveolin-3 mutants on growth factor signaling have been observed (71). We also observed enhanced ERK activation upon EGF stimulation in mutant cell lines; however, this effect was very mild and only observed in cells expressing the Dyn2 mutants in the absence of endogenous Dyn2. Thus, its relevance to disease pathology remains to be determined.
In contrast to our findings, others have reported that the activation of ERK1 and ERK2 was impaired by transient overexpression of CNM-linked Dyn2 mutants (31). These discrepancies in results obtained using our reconstitution system compared to others' transient transfection systems may reflect either cell-type differences or differential levels of expression of the mutant proteins. Indeed, the same study reported no effect on ERK activation in human fibroblasts derived from CNM patients expressing the R465W mutant (31). However, given the small size of the sample and huge variation within healthy control fibroblasts in this study (31), a more comprehensive and extensive examination of the EGF signaling pathway in DNM2-CNM and CMT patients and healthy controls would be necessary.
Transport from the Golgi and potential disease pathogenesis
Others had identified a role for Dyn in the transport of the p75/neurotrophin receptor from the Golgi (10,37), and we showed that this was a Dyn2 isoform-specific function (14). Altered neurotrophin receptor transport is emerging as a strong candidate to influence neurodegeneration (72,73). The impaired ability of disease-linked Dyn2 mutants to support p75 export from the Golgi and the decreased Golgi localization of CNM-linked middle domain mutants suggest that defects in Dyn2-dependent post-Golgi trafficking of p75 or other proteins may contribute to the pathogenesis.
Given the long distance between the cell body and dendrite or axon terminals in peripheral sensory and motor neurons, their sensitivity to perturbations in the transport from the Golgi to the PM is expected. Indeed, a mutation in KIF1B, a plus end-directed microtubule motor protein, has been linked to CMT (74). As for CNM, in addition to the strong dependence of muscle fibers on the nervous system, the importance of Golgi function and its distribution in muscle fibers for skeletal muscle generation and maintenance have been recently addressed (75). Thus, the post-Golgi trafficking defect associated with disease-linked Dyn2 mutants may contribute to the neuronal and muscular pathology and our data imply that a weak perturbation could eventually cause disease over prolonged time.
Ubiquitously expressed protein and tissue-specific diseases
How can dominant mutations in a ubiquitously expressed and essential protein cause a tissue-specific disease phenotype? It has been suggested that human diseases can result from subtle mutations acting over prolonged time periods in tissues that do not generally regenerate. Furthermore, as discussed earlier, there may be a strong dependence for peripheral neurons and muscle cells on the two membrane trafficking pathways we have identified as defective in CMT- and CNM-linked Dyn2 mutants. The endocytic trafficking of other signaling receptors might also be affected leading to tissue-specific alterations in signaling.
CME deficiency is not the major disease-causing mechanism
Several reports have suggested that disease-linked Dyn2 mutants are defective in their ability to support CME (15,31), although not every disease-related Dyn2 mutant inhibited CME (15). In general, these conclusions were based on the effects of transient overexpression of Dyn2 mutants, and it is possible that the dominant-negative phenotype could be due to other mechanisms of interference [e.g. titration of other essential components or the cytotoxicity of Dyn2 (32)] that are dependent on high levels of overexpression. In our hands, constitutive CME is not inhibited by the expression of these disease-related Dyn2 mutants even up to fivefold over endogenous levels. Moreover, these mutant proteins fully support CME even in the absence of endogenous WT Dyn2.
The importance of defining cellular phenotypes under physiological conditions
Defects in several cellular processes, including CME, the Golgi morphology, microtubule dynamics and centrosome cohesion, have been reported to be the potential disease-causing mechanisms of Dyn2 mutants (15,17,18,20,31). There is some inconsistency between this study and previous reports. Again, based on the different experimental design, we hypothesize that these discrepancies may result from (1) different protein expression levels or (2) different cell types (14). Clearly, it will be important to analyze the cellular phenotype in neuronal and muscle cells expressing these disease-causing Dyn2 mutants at physiological expression levels. In this regard, while this manuscript was in preparation, Bitoun's group reported the generation and initial characterization of a Dyn2 R465W knock-in mouse model that recapitulated an age-related myopathy (76). The authors report a number of muscle defects in these animals, including intracellular dysfelin accumulation, an abnormal mitochondria and reticular network, increased calcium concentration and transcriptional upregulation of genes involved in the ubiquitin–proteosome and autophagy pathways (76). Interestingly, mouse embryo fibroblasts derived from these animals exhibited a slight defect in CME that was not detected in our cells expressing the same mutant. Further work will be needed to explain this discrepancy. Regardless, our study is complementary to this article as it broadens the potential cellular activities that may contribute to Dyn2-induced human diseases and illuminates more areas for phenotypic analysis in model- and patient-derived cell lines.
Materials and Methods
Generation of Dyn2 mutant cell lines
Conditional Dyn2 KO cells (Dyn2+/−) were generated and cultured as previously described (14). To generate stable cell lines expressing the disease-related Dyn2 mutants, C-terminal GFP-tagged WT or mutant human Dyn2 constructs were subcloned into the retrovirus vector pMIEG3 and used to generate retroviruses as described (14,77). Retroviral supernatant was prepared by using the Phoenix packaging cell system and subsequently used to infect the Dyn2+/− cells in the presence of 8 µg/mL polybrene (Sigma). Cells were harvested 48 h after infection and GFP-positive cells with low, middle and high levels of expression were isolated using FACS. Cells were then cultured and those lines expressing comparable levels of protein relative to WT Dyn2-GFP were selected for further study.
Anti-Dyn2 antibody (sc-6400) and anti-EGFR (sc-03-G) were from Santa Cruz Biotechnology, anti-β-tubulin antibody (T-4026) was from Sigma, anti-γ adaptin, AP1 subunit (610385) was from BD Transduction Laboratories, anti-acetylated tubulin (ab24610) was from Abcam and anti-actin antibody (MAB1501R) was from Chemicon International. Both anti-ERK1/2 (137F5) and phospho-ERK1/2 (Thr202/Tyr204) antibodies were purchased from Cell Signaling. The HA (12CA5) antibody was purified from hybridoma obtained from Ian Wilson (TSRI), and mouse monoclonal anti-α-adaptin, AP2 subunit was purified from hybridoma AP.6 obtained from Frances Brodsky (UCSF).
Immunofluorescence staining and microscopy
For indirect immunofluorescence staining, cells cultured overnight on coverslips were fixed for 40 min with 4% paraformaldehyde and permeabilized with 0.05% saponin. After blocking with 2% BSA, cells were stained with the indicated primary and secondary antibodies. To reduce cytosolic Dyn2 signals, cells were fixed and permeabilized simultaneously with 2% warm paraformaldehyde and 0.5% TX-100 for 2 min and further fixed with 4% paraformaldehyde for 40 min. After immunofluorescence staining, cells were observed under wide-field epifluorescence microscopy using an inverted Olympus IX-70 microscope with a 100×, 1.35 numerical aperture (NA) oil-immersion objective.
Quantitative analysis of p75 export
Site-specific monobiotinylated p75-GFP was generated by introducing a 15-amino acid acceptor peptide sequence (AP) (38) into p75-GFP just after its signal sequence (Figure 4A). The AP-p75-GFP and biotin ligase BirA were constructed on the same vector separated by an IRES and under the tet-regulatable promoter. Adenoviruses expressing this construct were generated as previously described (78) and used to infect tTA-HeLa cells. Alternatively, fibroblasts were coinfected with tTA-encoding adenoviruses.
To analyze p75 trafficking, 70% confluent cells were coinfected with adenovirus expressing the tTA and adenovirus encoding AP-p75-GFP under control of a tetracycline-regulatable promoter. After overnight infection in the presence of 1 µg/mL tetracycline, AP-p75-GFP expression was induced by washing out tetracycline and incubating at 37°C for 5 h. To accumulate AP-p75-GFP in the Golgi and deplete newly synthesized AP-p75-GFP from the early secretory pathway, the cells were incubated for 3 h at 20°C in bicarbonate-free medium containing 100 µg/mL cycloheximide. To monitor p75 exit from the Golgi, cells were shifted to 32°C for 40 min, as previously described (37), and the surface-biotinylated p75-GFP was stained with streptavidin–Alexa-568. After extensive ice-cold PBS wash, the fluorescence intensities of Alexa-568 (surface) and GFP (total) were measured in 1% TX-100 cell lysates. For each cell line, control samples were subjected to the same experimental procedure except that they were infected with only tTA adenovirus virus to determine background levels of fluorescence intensity, which were subtracted. The ratio of Alexa-568:GFP fluorescence was taken as a measure of efficiency of delivery of AP-p75-GFP to the cell surface. As this is not an absolute value, data were normalized to the ratio obtained in control cells expression of WT Dyn2-GFP.
Transferrin (Tfn) internalization was performed as described (44) using bTfn as a ligand and assessing its internalization into an avidin-inaccessible compartment. Internalization of bEGF was analyzed as Tfn uptake except that cells were serum-starved for 4 h prior to the uptake assay. For CTB uptake, cells on coverslips were incubated with 1 µg/mL Alexa Fluor-594 CTB (Life Technologies) for 30 min at 4°C and then shifted to 37°C medium to allow internalization. Cells were washed extensively with cold PBS and acid buffer (150 mm NaCl, 150 mm glycine, pH 2.0), and then fixed and mounted for fluorescence microscopy. After fixation and mounting, the cells were viewed under an epi-fluorescence microscope. To quantify CTB uptake efficiency, for each cell line, 106 cells were detached from the dish and incubated with CTB-594 on ice for 30 min, then divided into three aliquots. One aliquot was shifted to 37°C for 10 min and stopped on ice. For total surface binding, cells kept on ice were washed 3× with cold PBS. The other two samples, with or without 10-min uptake, were subjected to an acid wash to obtain values for internalized or background signals, respectively. After extensive washes, cells were lysed with 1% TX-100 and fluorescence intensity was read in a 96-well plate reader (Bio-Tek Instruments).
EGF-stimulated ERK activation and EGFR downregulation
After 4-h serum starvation, cells were treated with 10 ng/mL EGF (Molecular Probes) for 5 min (ERK activation) or 90 min (EGFR degradation). After washing with cold PBS, cells were lysed in lysis buffer [50 mm Tris pH 7.5, 150 mm NaCl, 1% TX-100, 1 mm Na3VO4, 0.1 µm okadaic acid and proteinase inhibitor (Roche)] and spun to remove debris. Equal amounts of lysate/sample were subjected to western blot analysis.
We thank James R. Lupski (Baylor University, Houston, TX) for valuable discussions and helpful suggestions. We thank Cody Fine and Uyen Ngo from the TSRI Core FACS facility for help with cell sorting and the members of the Schmid lab for helpful discussion and critically reading the manuscript. This work was supported by NIH grants (GM42455 and MH61345) to S. L. S., and a fellowship supported by the Muscular Dystrophy Association (MDA) to Y.-W. L. (MDA-114824). This is TSRI Manuscript No. 20980.