Abstract
Haploinsufficiency of peripheral myelin protein 22 (PMP22) causes hereditary neuropathy with liability to pressure palsies, a peripheral nerve lesion induced by minimal trauma or compression. As PMP22 is localized to cholesterol-enriched membrane domains that are closely linked with the underlying actin network, we asked whether the myelin instability associated with PMP22 deficiency could be mediated by involvement of the protein in actin-dependent cellular functions and/or lipid raft integrity. In peripheral nerves and cells from mice with PMP22 deletion, we assessed the organization of filamentous actin (F-actin), and actin-dependent cellular functions. Using in vitro models, we discovered that, in the absence of PMP22, the migration and adhesion capacity of Schwann cells and fibroblasts are similarly impaired. Furthermore, PMP22-deficient Schwann cells produce shortened myelin internodes, and display compressed axial cell length and collapsed lamellipodia. During early postnatal development, F-actin-enriched Schmidt-Lanterman incisures do not form properly in nerves from PMP22−/− mice, and the expression and localization of molecules associated with uncompacted myelin domains and lipid rafts, including flotillin-1, cholesterol, and GM1 ganglioside, are altered. In addition, we identified changes in the levels and distribution of cholesterol and ApoE when PMP22 is absent. Significantly, cholesterol supplementation of the culture medium corrects the elongation and migration deficits of PMP22−/− Schwann cells, suggesting that the observed functional impairments are directly linked with cholesterol deficiency of the plasma membrane. Our findings support a novel role for PMP22 in the linkage of the actin cytoskeleton with the plasma membrane, likely through regulating the cholesterol content of lipid rafts.
Introduction
Peripheral myelin protein 22 (PMP22) is a neuropathy-linked glycoprotein whose role in peripheral nerve myelination is unclear (Li et al., 2013). Hereditary neuropathy with liability to pressure palsies, a condition that in rare cases is associated with the PMP22-null genotype (Saporta et al., 2011; Li et al., 2013), is diagnosed in individuals with heterozygous deletion of the PMP22 gene. In transgenic mice, the absence of PMP22 causes delayed myelination, dysmyelination, and tomacula formation (Adlkofer et al., 1995; Amici et al., 2006, 2007). How reduced dosage of, or the absence of, PMP22 leads to myelination defects is not fully understood, but compromised signaling via the α6β4 integrin complex (Amici et al., 2006) and altered junctional permeability (Guo et al., 2014) are likely to contribute.
Peripheral nerve myelination involves a series of functionally distinctive phases, including longitudinal Schwann cell migration, association with specific axon segments, membrane wrapping, and compaction (Jessen and Mirsky, 2005). These dynamic alterations in cell shape and motility require constant rearrangement of the actin cytoskeleton. A role for filamentous actin (F-actin) in the linkage of glial membrane proteins has been long recognized (Trapp et al., 1989), and a number of actin regulatory proteins are now known to be critical for myelination. In oligodendrocytes, calcium/calmodulin-dependent kinase type II regulates myelination by stabilizing the actin cytoskeleton (Waggener et al., 2013), while in the PNS, profilin 1, an actin-binding protein, is essential for myelination through mediating cytoskeletal remodeling (Montani et al., 2014).
Independent studies in multiple cell types indicate that underexpression and overexpression of PMP22 have profound effects on cell membrane dynamics and motility (Brancolini et al., 1999, 2000; Roux et al., 2005; Zoltewicz et al., 2012), functions requiring alterations in the actin cytoskeleton. A potential link between PMP22 and actin is supported by the irregular distribution of F-actin in PMP22 mutant Trembler J mice (Kun et al., 2012). In normal nerves, PMP22 is localized to detergent-insoluble lipid rafts (Erne et al., 2002; Hasse et al., 2002), which are known to interact with F-actin in oligodendrocytes (Taguchi et al., 2005). Lipid rafts are enriched in cholesterol, which is essential for myelin expansion (Saher et al., 2009). Recently, we identified palmitoylated PMP22 (Zoltewicz et al., 2012), a finding that further supports the association of this hydrophobic tetraspan protein with lipid rafts (Linder and Deschenes, 2007).
Based on the cited discoveries concerning the role of F-actin in membrane dynamics, the linkage of F-actin to lipid rafts, and the localization of PMP22 in cholesterol-enriched microdomains, we tested the hypothesis that PMP22 is critical for actin-mediated cellular functions and for the integrity of cholesterol-enriched lipid rafts. Using cells and nerves from a PMP22-deficient mouse model (Amici et al., 2006, 2007), we identified impairments in cell adhesion, migration, and membrane expansion, activities that require actin remodeling. In the absence of PMP22, the distribution and levels of key lipid raft constituents, including cholesterol, is altered, while PMP22-linked cellular defects can be rescued by cholesterol supplementation.
Materials and Methods
Mouse colony.
The PMP22-deficient mice used in these studies are missing the first two coding exons of PMP22 (exons 2 and 3), which were replaced with lacZ (Amici et al., 2006). Mice were maintained under specific pathogen-free housing at the animal facility of the University of Florida. PMP22 heterozygote males and females were bred to generate WT (PMP22+/+), PMP+/−, and PMP-deficient (PMP−/−) littermates. Genotypes were determined by Southern blotting (Amici et al., 2006). All experiments in this study used cells and tissues from male and female PMP22+/+ and PMP22−/− littermates.
Serum cholesterol and apolipoprotein E measurements.
Fasting serum cholesterol and triglyceride concentrations were measured in 6-week-old mice (n = 3–4 mice per genotype) using a colorimetric method. The serum lipoprotein distribution was determined by size exclusion chromatography [fast protein liquid chromatography (FPLC)], as previously described (Tavori et al., 2013). Apolipoprotein E (ApoE) detection and quantification was performed using 2 μl of serum loaded onto NuPage 4–12% Bis-Tris precast gels for electrophoresis. Goat anti-ApoE (R&D Systems) and rabbit anti-goat horseradish peroxidase (HRP; Sigma) antibodies were used for murine ApoE detection. Intensity quantification of the bands was obtained using ImageJ software (Abramoff et al., 2004).
Cell culture studies.
Dorsal root ganglia (DRGs) explants were isolated from embryonic day 14 (E14) to E16 mouse pups and plated on collagen-coated coverslips (Rangaraju et al., 2008, 2010). After 6 d in vitro, cultures were induced to myelinate by the addition of 50 μg/ml ascorbic acid (Sigma) to the media, for up to 2 weeks. Primary mouse Schwann cells were cultured from postnatal day 5 (P5) to P18 sciatic nerves. Nerves from killed mice were dissected into DMEM supplemented with Ham's F12 (DMEM-F12; Invitrogen), 10% fetal bovine serum (FBS; Hyclone), and gentamicin (20 μg/ml; Invitrogen), and were kept at 37°C for up to 7 d. Nerves were dissociated by digestion with collagenase and dispase, followed by mechanical trituration and plating (Fortun et al., 2007). Fibroblasts were eliminated with 5 μm cytosine arabinoside (Sigma) supplementation for 4 d. Schwann cells were expanded in DMEM-F12 with 10% FBS and 10 μg/ml bovine pituitary extract (Biomedical Technologies Inc), and 2 μm forskolin (Calbiochem). For cholesterol depletion, Schwann cells were plated on poly-l-lysine (PL)-coated coverslips for 24–48 h using DMEM-F12 media supplemented with 10% lipoprotein-deficient FBS (Sigma), bovine pituitary extract (10 μg/ml; Biomedical Technologies Inc), and forskolin (2 μm; Calbiochem). Free cholesterol (Sigma) was dissolved at 10 mg/ml into 100% ethanol and added to the cells at a final concentration of 20 μg/ml. Cholesterol was added for 24 h, and control cultures were treated with equal volumes of ethanol without cholesterol. Mouse embryonic fibroblasts (MEFs) were harvested from E13 to E16 embryos (Takahashi and Yamanaka, 2006). All experiments were performed with Schwann cells passaged less than three times and up to six times for fibroblasts.
Peripheral nerve staining.
Sciatic nerves were either teased into individual fibers or frozen in liquid nitrogen cooled N-methyl butane and sectioned before staining. For F-actin alone, or for F-actin and flotillin-1 staining, freshly isolated nerves were immersed in 4% paraformaldehyde in 0.1 m PBS for 30 min at room temperature (RT), after which individual fibers were teased onto Superfrost slides, dried for 1 h, permeabilized with 0.2% Triton X-100 in 0.1 m PBS for 30 min, and blocked with 5% normal goat serum (NGS) for 30 min. Alexa Fluor 488-conjugated phalloidin (1:500; Invitrogen) and rabbit anti-flotillin 1 antibody (1:200; Abcam) were diluted in blocking buffer, and applied for 24 h at 4°C. For immunodetection of ApoE (Abcam) and S-100 β-subunit (Sigma), nerve sections were incubated with the primary antibodies overnight at 4°C. Secondary antibodies (Alexa Fluor 594 goat anti-rabbit IgG and Alexa Fluor 488 goat anti-mouse IgG; Invitrogen) were added for 1 h. The number of Schmidt-Lanterman incisures (SLIs) per unit length of nerve (100 μm), and SLI length/width measurements were obtained using established methodologies (Yin et al., 2008; Jung et al., 2011). For filipin labeling, longitudinal nerve cryosections (5 μm thickness) were fixed with 4% paraformaldehyde (Electron Microscopy Sciences) in PBS for 10 min, dried for 1 h, and rinsed twice in PBS. Next, tissues were incubated with 0.2 mg/ml filipin (Sigma) diluted in PBS (Strauss et al., 2010; Saher et al., 2012). Quantification of filipin-cholesterol fluorescence was performed using ImageJ software. For cholera toxin subunit β (Ctx-β) labeling, cryosections were fixed with 4% paraformaldehyde in PBS for 10 min and incubated in cold acetone for 10 min (Petr et al., 2010). Samples were rinsed, blocked with 5% BSA in PBS, and incubated with 20 μg/ml Alexa Fluor 594 Ctx-β (Invitrogen). Nuclei were stained with Hoechst dye (Invitrogen), and tissues were mounted using Prolong Antifade media (Invitrogen).
Cholesterol measurements in nerve extracts.
Between 5 and 9 mg of sciatic nerve tissue were extracted with 200 μl of chloroform/isopropanol/NP-40 (7:11:0.1) in a microhomogenizer. Organic phase was dried under nitrogen stream, and the dried lipid extraction was dissolved in 150 μl of cholesterol assay buffer (Abcam) and analyzed using the colorimetric method.
Biochemical studies.
Sciatic nerves from WT and PMP22−/− mice (n = 3–6 animals per group, per experiment) were processed for Western blotting (Amici et al., 2006). Cultured cells were lysed in SDS gel sample buffer (62.5 mm Tris, pH 6.8, 10% glycerol, and 3% SDS), supplemented with complete protease (Roche) and phosphatase (Sigma) inhibitors. Samples were separated on 7.5% or 10% acrylamide gels under reducing conditions and transferred to nitrocellulose membranes (Bio-Rad). Blots were probed with the following antibodies: rabbit anti-PMP22 (Pareek et al., 1997; Amici et al., 2006); rabbit anti-β-galactosidase (β-gal; Abcam); rat anti-E-cadherin (Sigma); goat anti-ApoE (Millipore); mouse anti-myelin associated glycoprotein (MAG, Millipore); anti-α-tubulin and anti-actin (Sigma); anti-glyceraldehyde-3-phosphate dehydrogenase (GAPDH; EnCor Biotechnology); vinculin (Sigma); and flotillin-1 (Abcam). For analyses of the carbohydrate modification of PMP22, total cell lysates (30 μg) were treated with N-glycosidase F or endoglycosidase H (endoH; New England Biolabs), according to the manufacturer's instructions. Proteins were separated on 12.5% acrylamide gels, then transferred to PVDF membranes and probed with a polyclonal anti-PMP22 antibody (Pareek et al., 1997; Amici et al., 2006). Appropriate horseradish peroxidase-conjugated secondary antibodies (Cell Signaling Technology) were applied for 1 h, and bound antibodies were visualized with ECL (PerkinElmer). Films were digitally imaged using a GS-710 densitometer (Bio-Rad) and formatted for printing by using Adobe Photoshop CS2.
Immunocytochemistry.
Schwann cells or MEFs were fixed in 4% paraformaldehyde and permeabilized in 0.2% Triton X-100 for 15 min at RT, or in 100% methanol (for DRG explants), and were then blocked in 5% NGS in PBS for 1 h at RT. Following incubation with rat anti-myelin basic protein (MBP; Millipore), rabbit anti-p75NTR (Millipore Bioscience Research Reagents), rabbit anti-ApoE (Abcam), and mouse anti-S-100 β-subunit (Sigma) or mouse anti-vinculin (Sigma) antibodies overnight at 4°C, the appropriate Alexa Fluor-conjugated secondary antibodies were applied for 2 h. Nuclei were stained with Hoechst dye. For surface Ctx-β binding, cells were rinsed twice in cold 0.1 m PBS and incubated with 20 μg/ml Ctx-β (Invitrogen) diluted in 0.1 m PBS with 0.1 m Mg2SO4 for 30 min on ice (Lee et al., 2013). Samples were rinsed three times in PBS, fixed in 4% paraformaldehyde for 10 min at RT, and were then counterstained for 24 h at 4°C with Alexa Fluor 488-conjugated phalloidin (10 μg/ml; Invitrogen) diluted in PBS. For filipin staining, cells were fixed with 4% paraformaldehyde in PBS for 10 min at RT, and incubated with 2 mg/ml filipin (Sigma) diluted in PBS. For cell surface labeling, cultures were incubated with 1 μm FM143-X (Invitrogen) for 30 min in normal media, then rinsed twice in PBS and fixed with 4% paraformaldehyde for 5 min. Coverslips were mounted using the ProLong Antifade media (Invitrogen). Samples were imaged within 24 h using a Spot camera attached to a Nikon Eclipse E800 or Leica spinning disk confocal microscope. All images were acquired with identical settings, and postacquisition formatting was performed using Adobe Photoshop CS5.
Morphometrics.
Schwann cells immunostained with anti-P75NTR antibodies or labeled with FM143 were used to analyze cell length, and anti-MBP antibody reactive myelin segments were measured for evaluating internode lengths in the myelination assays. The axial length was defined as the longest axis of the cell and was measured using ImageJ (http://rsbweb.nih.gov/ij/index.html). Values were exported to Microsoft Excel to calculate means and SEMs. Measurements were performed blindly on three independent samples for each genotype and treatment, and were repeated two to three times.
Cell adhesion assays.
Ninety-six-well microtiter plates were coated for 2 h at 37°C with 10 μg/ml laminin (Promega), 3% (w/v) bovine serum albumin (BSA), or 5 μg/ml fibronectin (Sigma) in DMEM. Plates were then blocked with 1% BSA in DMEM for 1 h at 37°C. Freshly dissociated cells (10,000 cells/well) in normal culture media were allowed to adhere at 37°C for 30 min for Schwann cells or 10 min for fibroblasts, and then fixed in 1% glutaraldehyde for 15 min at RT. Wells were stained with 0.1% crystal violet (Sigma) in water for 30 min at RT, dried, and permeabilized in 0.3% Triton X-100 in PBS. Absorbance at 550 nm was measured in a microplate reader (Bio-Rad), and data were graphed and analyzed using Microsoft Excel. A positive correlation between the number of adhered cells and the absorbance value obtained has been established (Aumailley et al., 1989). Assays were repeated at least three times to verify results, with each assay containing four wells per condition. Statistical significance was determined using an unpaired, two-tailed Student's t test (Excel, Microsoft).
Cell migration assays.
Cell migration was measured using modifications from Meintanis et al. (2001). Cells (5 × 104) were plated in 40 μl drops on PL or laminin-coated glass coverslips in normal growth medium. Twenty-four hours later, the medium was changed to 2% FCS/DMEM, and cells were incubated for 12 h. For testing cholesterol-depleted conditions, cells were plated in media with 10% lipoprotein-deficient serum and kept for 24–48 h. The cell monolayer was gently scratched with a sterile micropipette tip to generate a ∼1-mm-wide gap. Immediately after scratching (t = 0 h), 10% FCS/DMEM was added, and migration of the cells within the gap was monitored microscopically at t = 0 h and t = 8 h using a Nikon T1-SM inverted microscope equipped with a Nikon DS-L1 camera. The gaps were asymmetrically marked with a needle along the scratches, and images at t = 0 h and t = 8 h were overlapped and aligned in Photoshop CS2. Over a period of 8 h, the area migrated by the cells was measured using ImageJ software (National Institutes of Health). The difference between the areas [(initial area at t = 0 h) − (area at t = 8 h)] was then calculated. Assays were repeated at least three times to verify the results. Statistical significance was determined using an unpaired, two-tailed Student's t test (Excel, Microsoft). Migration was observed for only 8 h to ensure that the presence of cells within the gap was due to movement and not proliferation. Parallel cultures were processed for immunostaining.
Statistical analyses.
To calculate the mean, SD, and SEM, raw values were exported into Microsoft Excel. At minimum, data from three independent experiments were quantified, and graphs were plotted representing the mean ± SEM. Statistical significance was determined based on p values obtained using an unpaired, two-tailed Student's t test.
Results
PMP22 is required for the organization of F-actin at Schmidt-Lanterman incisures
PMP22 is known to regulate cell shape and membrane expansion (Brancolini et al., 1999; Roux et al., 2005), functions that require changes in the actin cytoskeleton. We assessed the localization of F-actin during postnatal myelination in nerve fibers from WT (PMP22+/+) and PMP22−/− mice (Fig. 1A). In normal nerves, by 2 weeks of postnatal development, F-actin is prominent within intermittently dispersed conical-shaped funnels that are representative of SLIs (Fig. 1A, arrowheads), as well as in paranodal myelin loops, periaxonal membranes, and inner and outer mesaxons (Trapp et al., 1989). In contrast to the organized distribution of F-actin at SLIs of WT nerves, samples from age-matched PMP22-deficient mice show severely altered F-actin localization across all three ages examined. Notable pathological features include small and irregularly shaped SLIs, perinuclear clusters of F-actin (Fig. 1A, arrows), and intense phalloidin labeling at tomacula (Fig. 1A, circles). Quantification of the number of SLIs per unit length of nerve (100 μm) reveals a highly significant decrease in funnel-shaped incisures, while irregular phalloidin-positive actin clusters are increased in nerves of affected mice (Fig. 1B). As an additional analysis of SLI irregularities, we measured the shape of SLIs and determined the length/width index (Yin et al., 2008; Jung et al., 2011). As shown in Figure 1C, there is a significant decrease in the length/width index of SLI in nerves of PMP22-deficient mice. These results indicate that the structured distribution of filamentous actin at SLIs and other noncompacted myelin domains require PMP22.
By immunoblotting, we confirmed the expression of PMP22 in nerves from WT mice and its absence from the transgenic PMP22−/− samples, which are positive for β-gal (Fig. 1D; Amici et al., 2006). Based on the irregular profile of F-actin in SLIs, we examined the expression of proteins that are found in uncompacted myelin (Fig. 1E). Compared with age-matched WT mice, the levels of MAG, an adhesion molecule that mediates adaxonal contact during early development (Inuzuka et al., 1991), remain elevated at P60 in nerves from PMP22−/− mice (Fig. 1E). Similarly, the expression of E-cadherin in WT nerve lysates declines by P60 (Menichella et al., 2001; Crawford et al., 2008), but remains atypically high across all ages in PMP22-deficient samples. Despite these differences in the levels of MAG and E-cadherin, the abundance of total actin is comparable between the two genotypes (Fig. 1E). These data are in agreement with previous studies during the first 3 months of development in this model (Amici et al., 2006), and indicate that the regulation of actin filament organization and the timely expression of uncompact myelin proteins are affected by the absence of PMP22.
The absence of PMP22 has pronounced impact on Schwann cell elongation, lamellipodial morphology, and cell adhesion and migration
In addition to delayed developmental myelin synthesis, previous studies reported shortened myelin internodes in nerves from PMP22-deficient mice (Neuberg et al., 1999; Amici et al., 2007). Consistent with these reports, the majority of MBP-immunoreactive myelin segments produced by PMP22-deficient Schwann cells are significantly shorter compared with WT cells (Fig. 2A–C). Myelin produced by PMP22-deficient Schwann cells also display pathological features, including beading (Fig. 2B′) and bending (Fig. 2B″) within internodes. Comparison of frequency distribution of MBP-positive myelin segments by length between the genotypes reveals a significantly (***p < 0.001) higher proportion of internodes in the range of ≤75 μm, and a significantly (***p < 0.001) smaller proportion >100 μm, in cultures from PMP22−/− mice (Fig. 2C).
To examine whether the shortened cellular phenotype is inherent to PMP22-deficient glia even in the absence of axonal contact, we analyzed morphological features of nonmyelinating Schwann cells, after immunolabeling with an anti-P75 antibody (Fig. 2D,E). While WT Schwann cells display a polarized and elongated morphology with prominent lamellipodia (Fig. 2D,D′), cells cultured from PMP22−/− mice frequently appear flattened and short (Fig. 2E,E′). Quantification of the cell length along the longest axis (axial length) indicates that the absence of PMP22 causes a shortened phenotype, regardless of the plating substrate of either PL or laminin (Fig. 2F). We also detected morphological differences in the lamellipodial ruffles, which appear jagged and collapsed in the absence of PMP22 (Fig. 2E′), compared with the smooth and rounded morphology in nearly all WT cells (Fig. 2D′).
While it is known that the amount of PMP22 expression increases with myelination, the presence and maturation of the protein in nonmyelinating rat Schwann cells has been documented (Pareek et al., 1997). Similarly, in WT mouse Schwann cells lysates, immunoblotting with anti-PMP22 antibody identifies an ∼22 kDa band (Fig. 2G, lane C) that shifts to the 18 kDa core peptide upon treatment with N-glycosidase (Fig. 2G, lane N). In the fully mature and plasma membrane-associated form, PMP22 acquires a complex carbohydrate modification and is resistant to endoH (Pareek et al., 1993). In accord, a fraction of PMP22 in nonmyelinating WT mouse Schwann cells is not cleaved by endoH (Fig. 2G, lane H), indicating association with the plasma membrane. Together, these results show that in normal mouse Schwann cells PMP22 regulates the elongation and extension of plasma membrane in a cell-autonomous mechanism.
Next, we tested whether the lack of PMP22 influences Schwann cell migration, a function that requires filamentous actin dynamics (Insall and Machesky, 2009). The migratory efficiency of nonmyelinating Schwann cells was analyzed in the scratch wound assay (Meintanis et al., 2001), which promotes cells to undergo directional movement. After an 8 h period, PMP22-deficient cells show reduced mobility compared with WT (Fig. 3A,B). This impairment in migration by PMP22-deficient cells is significant for both laminin and fibronectin substrate (Fig. 3C). As migration by adherent cells requires dynamic assembly and disassembly of adhesive contacts with the extracellular matrix (ECM), we compared the adhesive properties of WT and PMP22-deficient cells. Using BSA or laminin as a substrate, PMP22−/− Schwann cells adhered ∼60% less to laminin than WT cells, attaching only slightly more to laminin than to BSA (Fig. 3D). This finding supports a role for PMP22 in mediating the interaction of Schwann cells with the ECM, likely through association with the α6β4 integrin complex (Amici et al., 2006).
In addition to altered adhesion, membrane dynamics at the leading edge may also account for the delayed migration of PMP22−/− Schwann cells. Therefore, we examined lamellipodia morphology after the scratch migration by immunostaining the cultures with an anti-P75 antibody (Fig. 4A,B). While WT cells display many broad, fan-shaped lamellipodia that face the scratch (Fig. 4A, arrows), PMP22−/− Schwann cells show a >20% reduction (mean ± SD, 100 ± 12 vs 75 ± 20; *p < 0.05) in properly formed lamellipodia (Fig. 4B, arrow). A number of thin, collapsed cell projections are visible at the leading edge of the migrating PMP22−/− cultures (Fig. 4B, arrowheads). The decrease in lamellipodia ruffles in PMP22−/− Schwann cells was confirmed by double immunostaining with antibodies to actin and the focal adhesion protein vinculin, which links adhesion molecules to the actin cytoskeleton (Mierke, 2009; Fig. 4C,D). The localization of vinculin-like staining is collapsed in lamellipodia of PMP22−/− Schwann cells (Fig. 4C), compared with the fan-shaped pattern in WT cells (Fig. 4C). We found a similar distribution of F-actin at the lamellipodia of migrating cells, by labeling with phalloidin (Fig. 4E,F).
As PMP22 was originally identified in fibroblasts (Schneider et al., 1988) and its expression levels are known to affect the morphology of NIH3T3 cells (Fabbretti et al., 1995; Brancolini et al., 1999; Roux et al., 2005), we asked whether the observed migratory deficits of PMP22-deficient Schwann cells would be replicated in MEFs. Consistent with the upregulation of PMP22 transcripts under growth-arrested conditions (Brancolini et al., 1999), β-gal is expressed in PMP22-deficient MEFs after 24 h culture in 0.5% FCS (Fig. 5A). Within the same cell lysates, the steady-state levels of vinculin and actin are not affected by the absence of PMP22 (Fig. 5A). Next, we assessed the adhesive properties of WT and PMP22−/− MEFs on BSA, poly-l-lysine, and fibronectin-coated tissue culture plates. As shown in the quantification of independent experiments, on all three substrata tested, the absence of PMP22 negatively impacts the adherence of the cells (Fig. 5B).
Similar to working with Schwann cells, we used the scratch wound assay to compare the migratory capacity of WT and PMP22−/− MEF cultures (Fig. 5C–D′). Analogous to the findings in glia (Fig. 3), fibroblasts deficient in PMP22 are impaired in their ability to close the wound, compared with WT cells (Fig. 5C–D′). Quantification of the percentage of the area migrated 8 h post-scratch wounding reveals that the mobility of PMP22−/− fibroblasts on fibronectin-coated plates is ∼30% less than that in WT cells (mean ± SEM, 52.5 ± 2.4 vs 36.6 ± 2.9; p = 0.00006). Using phalloidin, we examined the distribution of F-actin at the leading edge of the MEF cultures (Fig. 5E,F). Compared with PMP22−/− Schwann cells, both WT and PMP22-deficient MEFs extended fan-shaped filopodia; however, prominent stress fibers are observed in PMP22−/− cultures (Fig. 5F). It is possible that in MEFs the absence of PMP22 interferes with actin remodeling, thereby thwarting the migratory ability of the cells. Together, these results support our hypothesis that PMP22 facilitates basic cellular functions that require dynamic rearrangements of the F-actin cytoskeleton, including cell adhesion and motility.
PMP22 is critical for the establishment of lipid rafts
Actin filaments are known to interact with cholesterol-enriched lipid rafts in oligodendrocytes (Taguchi et al., 2005), as well as in epithelia (Head et al., 2014). Since PMP22 is a known constituent of lipid rafts in peripheral nerve myelin and in transfected HeLa cells (Hasse et al., 2002), we examined the localization of lipid raft proteins in the presence and absence of PMP22 (Fig. 6). In WT nerves, the scaffold protein flotillin-1 is detected along the outer Schwann cell membrane, and is colocalized with F-actin at SLIs (Fig. 6A, arrowheads). In nerves from PMP22−/− mice, flotillin-1 is similarly identified on the outer glial membrane; however, it is not enriched along the SLIs (Fig. 6B, arrows), where F-actin is detected. Indeed, analyses of independent teased nerve fibers from WT and PMP22-deficient mice reveal a significant decrease in SLIs with surrounding flotillin and a corresponding increase in actin clusters without flotillin in the knock-out samples (Fig. 6C). These results suggest a disruption in the organization of lipid rafts, and associated actin filaments (Fig. 1), in the absence of PMP22.
Next, we analyzed whole-nerve lysates from WT and PMP22−/− mice between the ages of P9 and P60 with antibodies to flotillin and ApoE (Fig. 6D). Compared with the steady levels of flotillin-1 in WT nerves, samples from PMP22−/− mice show an increase in flotillin-1 expression at P60 (Fig. 6D). As cholesterol is the main lipid component of rafts (Simons and Ikonen, 1997) and cholesterol is critical for myelination (Saher et al., 2005; Saher et al., 2011), we determined the levels of ApoE, a cholesterol transport molecule, within the same nerve lysates. While the level of ApoE in normal nerves diminishes at approximately P23, it drastically increases by 2 months of age in the transgenic samples (Fig. 6D). This observed atypical pattern in ApoE expression is reminiscent of what has been reported in squalene synthase (SQS)-deficient mice (Saher et al., 2009). As ApoE plays a major role in serum lipid metabolism (Davignon et al., 1999; Fazio and Linton, 2005), we analyzed cholesterol, triglyceride, and ApoE levels in the serum of 6-week-old WT and PMP22−/− mice. No significant differences were found between serum cholesterol levels of fasting PMP22−/− (66 ± 16 mg/dl) and WT (75 ± 20 mg/dl) mice. However, levels of serum triglycerides in PMP22−/− mice were significantly lower compared with WT mice (40 ± 3 and 65 ± 13 mg/dl, respectively; p < 0.01). Serum lipoprotein fractionation via FPLC showed that PMP22−/− mice carry 35 ± 6% less triglyceride in their very low-density lipoprotein compared with WT mice. In line with this observation, PMP22−/− mice also showed a 36% reduction in serum ApoE levels (p < 0.05) compared with WT mice. These data indicate that the increase in ApoE in affected nerve samples is due to changes within the nerve tissue, rather than the vasculature.
To examine the cellular source for the elevated ApoE levels detected in the whole-nerve lysates (Fig. 6D), we performed double immunolabeling on nerve sections with anti-ApoE and S-100β antibodies (Fig. 6E,F). In nerve sections from 4-week-old WT mice, there are few ApoE-reactive cells that also colabel with the Schwann cell marker S-100β (Fig. 6E, arrows). In comparison, ApoE-reactive, S-100β-positive cells are frequent in nerve sections from PMP22−/− mice (Fig. 6F, arrows), suggesting an increase in ApoE synthesis by Schwann cells. To further substantiate the increase in ApoE expression in Schwann cells in the absence of PMP22, we performed Western blots on cell lysates and immunostained cultured glia (Fig. 6G–I). As shown in Figure 6G, there is a notable increase in ApoE in whole Schwann cells, as well as MEF lysates, when PMP22 is deleted. The biochemical studies are corroborated by the reactivity of S-100β-positive cultured Schwann cells with ApoE antibodies (Fig. 6H,I). In agreement with the Western blots, the intensity of the ApoE-like immunofluorescence is elevated in cells from PMP22−/− mice, compared with WT mice, when the images are acquired with the same exposure settings (Fig. 6H,I).
To further examine the linkage between the absence of PMP22 and alterations in lipid rafts and cholesterol homeostasis (Fig. 6), we investigated potential changes in raft lipids in PMP22−/− nerves (Fig. 7). The ganglioside GM1 marker, Ctx-β (Heyningen, 1974) shows abundant granular distribution along myelinated fibers in WT nerves (Fig. 7A,A′). In comparison, nerves from PMP22−/− mice display an overall reduced labeling with Ctx-β, and some areas of the nerves are nearly devoid of signal (Fig. 7B,B′, arrows). Concurrently, a subset of PMP22−/− fibers displays clusters of GM1 (Fig. 7B′, arrowheads), indicating irregularities in the organization of lipid rafts. As a follow-up to the detection of Ctx-β clusters in nerves of PMP22-deficient mice (Fig. 7B,B′), we analyzed the distribution of GM1 on the cell surface of nonpermeabilized Schwann cells. In WT cells, Ctx-β fluorescence is prominent on the surface, with occasional round patches (Fig. 7C). The distribution of GM1 is notably different on the plasma membrane of PMP22-deficient cells, where the overall fluorescence is reduced, and fewer Ctx-β clusters are detected (Fig. 7D).
We also analyzed cholesterol distribution using filipin, a polyene antibiotic that binds specifically to cholesterol (Norman et al., 1972; Lintner and Dyer, 2000). In nerves from WT mice, the distribution of filipin-cholesterol is nearly homogenous, yielding a strong signal along myelinated membranes (Fig. 7E,E′). Reminiscent of the GM1 pattern, in PMP22-deficient nerves, there is an overall reduction in the filipin-cholesterol fluorescence (Fig. 7F,F′). Concomitantly, in certain areas, we detect clusters of filipin labeling (Fig. 7D′, arrows), which may suggest altered trafficking of cholesterol in Schwann cells. Therefore, we next examined the distribution of cholesterol in Schwann cells cultured from WT and PMP22−/− mice (Fig. 7G,H). As filipin permeabilizes the membrane of cells (Norman et al., 1972), this technique detects mostly intracellular stores of cholesterol. In the absence of PMP22, the distribution of cholesterol-containing vesicles appear larger and clustered (Fig. 7H, arrows) compared with the smaller, more uniform vesicular pattern in WT cells (Fig. 7G). Overall, there appears to be a preferential retention of cholesterol within the perinuclear region of the Schwann cell, when PMP22 is absent. To corroborate the apparent reduction in the overall filipin cholesterol labeling in affected nerves and Schwann cells, we quantified filipin-cholesterol intensity using ImageJ (Fig. 7I). As shown in the graphs, in both 3- and 7- to 9-week-old mouse nerves we detect an ∼30% decrease in filipin-cholesterol when PMP22 is absent. In agreement, analyses of total nerve cholesterol using a colorimetric method also detects a statistically significant 32 ± 4% (p < 0.01) reduction in cholesterol in the nerves of PMP22−/− mice, compared with WT mice. Together, these data indicate that the absence of PMP22 leads to alterations in the distribution of raft lipids and a significant decrease in total nerve cholesterol.
Exogenous cholesterol restores cell length and migration deficits in the absence of PMP22
To test whether altered levels of lipid raft cholesterol is a contributing factor in PMP22 deficiency-dependent cell elongation and migration defects, we assessed the effects of exogenous cholesterol supplementation on PMP22-null Schwann cells (Fig. 8). Control and PMP22-deficient cultures were maintained in 10% lipoprotein-deficient FCS before the addition of 20 μg/ml cholesterol, after which axial cell length and migratory capacity were tested. Cell morphology after staining with the plasma membrane marker FM-143 reveals no obvious difference between WT and PMP22-deficient cells after cholesterol supplementation (Fig. 8A). Indeed, quantification of axial cell length confirms that PMP22−/− cells significantly extend with cholesterol supplementation, and regain cell length similar to that in WT cells (Fig. 8B). We also performed the scratch-wound migration assay in the absence or presence of 20 μg/ml cholesterol (Fig. 8C). While migration of PMP22-deficient cells is significantly slower than that of WT cells in the absence of cholesterol, providing an exogenous source of cholesterol increases the overall migration speed of both cell types (WT, ***p < 0.001; PMP22−/−, ***p < 0.001). Quantification of data from independent experiments reveals that the percentage area migrated during an 8 h time period is similar between WT and PMP22-deficient Schwann cells (Fig. 8D). These results indicate that insufficient cholesterol in the plasma membrane is likely a factor contributing to the functional deficits in migration and cell extension by PMP22-deficient cells.
Discussion
The results shown in this article demonstrate that in peripheral nerves the absence of PMP22 causes severe disruption in the formation of F-actin-enriched SLIs and associated abnormalities in cholesterol-enriched lipid rafts (Fig. 9). We also provide evidence for reduced levels and abnormal localization of cholesterol in PMP22-deficient Schwann cells, an alteration that is associated with impaired cell adhesion, migration, lamellipodia extension, and myelination. With regard to cholesterol trafficking, by examining plasma lipids we ruled out a vascular source for the elevated nerve ApoE in affected mice and found alterations in the lipid balance (Fig. 9, model). How alterations in systemic lipids may impact neuropathy severity is currently unknown and requires further studies in additional neuropathic models. As exogenous supplementation of cholesterol reverses the PMP22-linked defects in cell shape and migration, the results indicate that either directly or through interactions with other glial proteins, PMP22 is crucial for stabilizing cholesterol in the plasma membrane. These findings suggest a novel disease mechanism by which reduced levels of PMP22 at the plasma membrane may lead to demyelination and developmental myelin abnormalities through alterations in the organization and stability of lipid rafts.
The most pronounced cellular defects in the complete absence of PMP22 involve the actin cytoskeletal network (Figs. 1, 2, 3, 4, 5). We found evidence for disruption in the F-actin network in peripheral nerves and in two different culture models, namely, Schwann cells and MEFs. While this finding is novel in the context of myelination, it is a phenotype that has been reported in a number of independent articles. Previous studies have linked the overexpression of PMP22 with alterations in actin-dependent cellular features, such as flattened morphology, altered tight junction permeability, and reduced motility in epithelia (Roux et al., 2005; Zoltewicz et al., 2012). In a number of other cell types, including Schwann cells, Cos7 cells, and NIH3T3 fibroblasts, increased cell spreading, membrane blebbing, and altered plasma membrane dynamics have been reported (Brancolini et al., 1999; Chies et al., 2003; Nobbio et al., 2004). Considering its tetraspan membrane topology, as well as BLAST searches, it is highly unlikely that PMP22 directly binds actin. Nonetheless, based on information from the literature, a number of mechanisms by which PMP22 indirectly affects actin dynamics can be proposed. First, using the same animal model as in the current study, we discovered that PMP22 interacts with α6β4 integrin (Amici et al., 2006), a protein complex with established function in linking laminin to the cytoskeleton (Lee et al., 1992). If the absence of PMP22 destabilizes the expression of β4 integrin, alternate integrin subunits with different binding properties may be expressed to compensate (Milner et al., 1997) and, therefore, may be responsible for the altered interactions with laminin and the subcellular actin network. PMP22 could also impact actin dynamics through its palmitoylation, which indeed has a pronounced influence on epithelial cell shape and migration (Zoltewicz et al., 2012). Even though palmitoylation-deficient PMP22 was detected at the plasma membrane, caveolin-enriched microdomains were abnormally clustered compared with cells expressing WT PMP22. PMP22 expression may also influence the Rho GTPases as PMP22 transfection in fibroblasts leads to Rho-mediated stress fiber assembly (Brancolini et al., 1999). An additional possibility implicates various scaffold proteins that through direct or indirect interactions with PMP22, or its binding partners such as protein 0, could affect the stabilization of F-actin.
Myelin has uniquely high cholesterol content and cholesterol-enriched membrane microdomains, or lipid rafts, which are functionally important for spatial regulation of structural and signaling proteins. PMP22 associates with lipid rafts in detergent-insoluble fractions (Erne et al., 2002; Hasse et al., 2002), which is consistent with the protein being localized at tight junctions (Notterpek et al., 2001). Lipid rafts serve important roles in the plasma membrane by anchoring signaling complexes, including tight junctions and adherens junctions, which interact with the F-actin cytoskeleton network (Head et al., 2014). Adherens junctions in SLIs contribute to the stabilization of myelin architecture (Fannon et al., 1995) and take an active role in the organization of the actin cytoskeleton (Tricaud et al., 2005). It was recently discovered that reduced PMP22 levels in heterozygous PMP22-deficient mice alter the permeability at tight junctions through interaction with MAG and the junctional adhesion molecule-C (Guo et al., 2014). Similarly, changes in tight junction permeability in response to PMP22 overexpression was described in epithelial cells (Roux et al., 2005). Additional support for the connection between lipid rafts and actin filamentous network in peripheral nerve myelin is the previously described interaction between lipid rafts and actin filaments in oligodendrocytes (Taguchi et al., 2005). The association between lipid rafts and F-actin in Schwann cells has not been studied in detail; however, when PMP22 is misexpressed, the organization of the F-actin cytoskeleton is disturbed (Kun et al., 2012).
As mentioned above, cholesterol is a known regulator for myelination in the CNS and the PNS (Saher et al., 2011). A number of elegant studies in mice show that when cholesterol synthesis is diminished, oligodendrocytes as well as Schwann cells are unable to form normal myelin (Saher et al., 2005, 2011). In the CNS, it is also known that proteolipid protein (PLP), a critical tetraspan myelin constituent in oligodendrocytes, directly binds cholesterol (Werner et al., 2013). Therefore, it is plausible that based on the structural similarities between PMP22 and PLP, in Schwann cells PMP22 binds cholesterol in the plasma membrane. In support of this hypothesis, PMP22 is an extremely hydrophobic tetraspan protein (Sedzik et al., 2002) that contains one highly conserved cholesterol recognition/interaction amino acid consensus (CRAC) motif within the fourth transmembrane domain (Sedzik et al., 2013). The similarities in abnormalities between PMP22−/− and cholesterol-deficient mice during early postnatal development, such as delayed myelination and reduction in compact myelin proteins (Saher et al., 2005; Amici et al., 2006), support a role for PMP22 in sequestering cholesterol in myelin. Finally, there are a number of early reports that described lipid abnormalities in Trembler and Trembler J mice (Larrouquere-Régnier et al., 1979; Bourre et al., 1984, 1989), which were later shown to carry mutations in PMP22 (Suter et al., 1992a,b). While myelin abnormalities likely involve alterations in the lipid content of nerves in general, the discovery of specific delays in cholesterol buildup in the Trembler nerve supports the idea that PMP22 is critical for sequestering cholesterol in myelin.
The functional deficits of elongation and migration in PMP22-deficient Schwann cells that are restored upon the addition of exogenous cholesterol may reflect a secondary impairment in cholesterol biosynthesis or transport. The pronounced upregulation of ApoE in nerves from PMP22−/− mice (Fig. 6G) may suggest that, in the absence of PMP22, the Schwann cells are more dependent on the uptake of cholesterol from neighboring cells, as was observed in the SQS-deficient model (Saher et al., 2011). The high levels of ApoE in affected nerves did not originate from the circulation, as PMP22−/− mice actually exhibited lower serum triglycerides and ApoE levels compared with WT mice. In support of an intracellular cholesterol transport defect, we detected lower levels of total cholesterol and reduced intensity of filipin-cholesterol on the plasma membrane of the mutant cells, while clumps of cholesterol were visible in the perinuclear region (Fig. 7H). How PMP22 expression specifically affects the transport and/or membrane association of cholesterol will require further studies as a number of Schwann cell proteins, including protein 0 contain CRAC motifs (Sedzik et al., 2002). PMP22 is known to interact with protein 0, whose export from the endoplasmic reticulum is dependent on cholesterol (Saher et al., 2009, 2011). Therefore, it is possible that the interaction between PMP22 and protein 0 is required for fully functional cholesterol transport.
The current study provides the rationale for alternative approaches in modulating myelination and myelin stability in PMP22-linked neuropathies through diet. While exogenous supplementation with lipids will not correct the genetic abnormality in PMP22 neuropathies, it may be effective in slowing disease progression. In support of such an approach is the improved phenotype of PLP-overexpressing transgenic mice, in which feeding a cholesterol-enriched diet prevented disease progression and supported myelin membrane expansion (Saher et al., 2012). It is conceivable that effective therapies for hereditary neuropathies will require a combinatorial therapy that will target the intracellular abnormalities in protein homeostasis and the associated pathology due to reduced levels of PMP22 in the plasma membrane.
Footnotes
These studies were supported in part by grants from the National Institutes of Health, the Facial Pain Research Foundation, and Tom and Susie Wasdin (L.N.); and the University of Florida University Scholars Program (S.Freeland., W.M.Z.). We thank Dr. Klaus Nave for insightful comments concerning this project, and Dr. Kassidy Chauncey for technical assistance. We also acknowledge past and current Notterpek laboratory members for help with tissue collection and genotyping.
The authors declare no competing financial interests.
- Correspondence should be addressed to Dr. Lucia Notterpek, 1149 South Newell Drive, P.O. Box 100244, Gainesville, FL 32611. Notterpek{at}ufl.edu