Dlg1-PTEN Interaction Regulates Myelin Thickness to Prevent Damaging Peripheral Nerve Overmyelination

See allHide authors and affiliations

Science  11 Jun 2010:
Vol. 328, Issue 5984, pp. 1415-1418
DOI: 10.1126/science.1187735


The thickness of the myelin sheath that insulates axons is fitted for optimal nerve conduction velocity. Here, we show that, in Schwann cells, mammalian disks large homolog 1 (Dlg1) interacts with PTEN (phosphatase and tensin homolog deleted on chromosome 10) to inhibit axonal stimulation of myelination. This mechanism limits myelin sheath thickness and prevents overmyelination in mouse sciatic nerves. Removing this brake results also in myelin outfoldings and demyelination, characteristics of some peripheral neuropathies. Indeed, the Dlg1 brake is no longer functional in a mouse model of Charcot-Marie-Tooth disease. Therefore, negative regulation of myelination appears to be essential for optimization of nerve conduction velocity and myelin maintenance.

The thickness of the myelin sheath is an essential parameter governing nerve conduction velocity (NCV) (1). This velocity falls from its optimum when myelin is too thick or too thin. Optimal conduction is obtained by adjusting the length and thickness of the myelin sheath to the axon diameter (2). To achieve this in peripheral nerves, myelinated axons express neuregulin-1 type III, which stimulates myelination by Schwann cells (SCs) (3, 4). Myelination slows down after the first two postnatal weeks in mice as optimal thickness is approached (5), which suggests regulatory mechanisms that curb myelination.

We have previously characterized a mouse model of Charcot-Marie-Tooth disease 4B (CMT4B), a peripheral neuropathy caused by mutations in MTMR2 (6). Schwann cell hypermyelination with focally folded myelin is the morphological disease hallmark (7), which indicates dysregulation of myelin production. MTMR2 interacts with mammalian disks large homolog 1 (Dlg1, also SAP-97) (8), a PDZ-containing membrane-associated guanylate kinase (MAGUK)–family protein involved in basic cellular events including cell polarization, protein trafficking, and tumorigenesis (9). Here, we describe Dlg1 function during SC myelin sheath formation.

Dlg1 expression was silenced in SCs cocultured with DRG neurons (10) (fig. S1A). We confirmed that Dlg1-silenced SCs failed to myelinate (fig. S1B) (11). These cells also showed migration defects (fig. S1, C and D) and expressed less polarity protein Par3 (fig. S1E), which is required for cell migration (12) and myelination (13). Occasionally, silenced SCs overcame their migration defect and myelinated (fig. S1F). The resulting myelin segments were thicker than those of controls (fig. S1G), which indicated a role for Dlg1 in regulating myelin sheath thickness.

We then silenced Dlg1 expression efficiently in myelinating SCs in mouse sciatic nerves at early stages of myelination (3 to 4 days postnatal, dPN) by injecting lentiviral vectors expressing short hairpin RNAs (shRNAs) (fig. S2) (10, 14). Two months after injection, we found a strikingly increased thickness of myelinating SCs silenced for Dlg1 compared with control cells (Fig. 1, A and B; 9.3 ± 0.2 μm versus 7.1 ± 0.3 μm, P < 0.001). Average cell lengths remained unaffected (Fig. 1B, 347 ± 11 μm versus 363 ± 19 μm, P = 0.49). The effects were SC autonomous, because Dlg1-silenced cells were thicker than neighboring cells on the same axon (Fig. 1C).

Fig. 1

Dlg1 silencing increases myelin sheath thickness. (A) SC infected with Dlg1 shRNA virus (bottom) shows a larger mean diameter than SC infected with control shRNA virus (top). Scale bars, 20 μm. (B) Diameters versus lengths of SCs infected with Dlg1 or control shRNAs viruses. Lines: Linear regressions for control (n = 43), slope line 1 and Dlg1 (n = 87), line 2 shRNAs. Two independent shRNAs against Dlg1 yielded similar results. (C) At a node of Ranvier, a SC infected with Dlg1 shRNA virus (green overlay) is larger than the next noninfected SC (NI, brown overlay). Scale bar, 5 μm. (Graph) Values for infected SCs are normalized to values for noninfected (NI) SCs (n = 13 nodes) (±SEM). (D) (Left) SC infected with Dlg1 shRNA virus shows PLAP staining at interface with axon (arrowheads). Scale bar, 1 μm. Note the thick myelin sheath of this SC in comparison with the next noninfected SC. (Insets) Details of SC with PLAP staining (top) and SC without staining (bottom). (Right) Mean g ratio (±SEM) of infected (Dlg1sh) cells (n = 111) and noninfected (NI) cells (n = 172).

Coexpression of placental alkaline phosphatase (PLAP) allowed detection of infected SCs by electron microscopy (Fig. 1D). Compact myelin of Dlg1-silenced cells appeared structurally normal (fig. S3), except that myelin was thicker (Fig. 1D). Consequently, the g ratio (axon diameter/total fiber diameter) was significantly reduced (Fig. 1D), with no change in average axon diameters (10.7 ± 0.4 μm versus 11 ± 0.2 μm, P = 0.5).

We also overexpressed Dlg1 fused with green fluorescent protein (GFP) in SCs of mouse sciatic nerves. Two months after birth, the majority of GFP-Dlg1–positive cells were non–myelin-forming cells; most cells expressing GFP alone had myelinated (fig. S4), which suggested that Dlg1 overexpression prevented SC myelination.

Our data indicated that Dlg1 acts as an inhibitor of SC myelination. Myelination is stimulated by axonal neuregulin-1 type III activating erbB2 and erbB3 receptors and stimulating the phosphatidylinositol 3-kinase (PI3K)–AKT pathway in SCs (15). Thus, we tested whether Dlg1 interferes with AKT activation. Indeed Dlg1 silencing induced a significant increase of AKT activation in cultured SCs (Fig. 2A and fig. S5A), as well as in myelinating SCs in vivo (Fig. 2B). Moreover, Dlg1 reduction potentiated AKT activation upon neuregulin-1 stimulation in cultured SCs (Fig. 2C), which demonstrated that Dlg1 inhibits neuregulin-driven AKT activation.

Fig. 2

Dlg1 inhibits AKT activation via interaction with PTEN. (A) Cultured SCs silenced for Dlg1 (Dlg1sh) have more phospho-AKT (P-AKT). For quantifications, see fig. S5A. (B) At a node of Ranvier (arrowhead), a myelinated SC infected with Dlg1 shRNA virus (GFP) has more P-AKT than the next noninfected SC. Scale bar, 5 μm. (C) Cultured Dlg1-silenced SCs (Dlg1sh) stimulated with neuregulin-1 (Nrg1) for 15 min generate more P-AKT than cells infected with control shRNA virus or noninfected cells (NI). Representative results of two independent experiments. (D) PTEN coimmunoprecipitates with Dlg1 (IP Dlg1) but not with nonimmune serum (NS) in rat sciatic nerves at 20 dPN. Representative results of two independent experiments. (E) Nrg1 stimulation of SCs increases Dlg1 and PTEN amounts after 15 min. Representative results of four independent experiments. (F) More PTEN coimmunoprecipitates with Dlg1 when SCs are stimulated with Nrg1. (Graph) Relative amount of coimmunoprecipitated PTEN in stimulated and nonstimulated SCs (±SEM; for four independent experiments). No PTEN is detected when nonimmune serum (NS) is used. (G) Cultured SCs, left untreated or treated with Nrg1 for 15 min, and vehicle dimethylsulfoxide (DMSO) or inhibitors of PI3K (Ly), AKT (AKTi), MAPK (U2), and mRNA translation (cycloheximide, Cx). Representative results of three experiments.

Dlg1 can interact directly with PTEN (phosphatase and tensin homolog deleted on chromosome 10) (16, 17), an inhibitor of AKT activation (18). This interaction may stabilize PTEN (19), which results in less AKT activation (17). We could coimmunoprecipitate PTEN with Dlg1 from both cultured SCs (fig. S5B) and sciatic nerves (Fig. 2D). Moreover, Dlg1 silencing reduced PTEN levels, whereas Dlg1 overexpression increased them (fig. S5C), which indicated that Dlg1 regulates PTEN amounts in SCs. We also observed that neuregulin-1 stimulation rapidly increased Dlg1 and PTEN (Fig. 2E), which resulted in more Dlg1-PTEN complexes (Fig. 2F).

Although neuregulin-1 activates PI3K-AKT and mitogen-activated protein kinase (MAPK) pathways in SCs (15, 20), inhibiting these pathways did not prevent the increase in Dlg1 and PTEN upon stimulation (Fig. 2G). This rapid increase was also not blocked by cycloheximide (Fig. 2G), which suggested that neuregulin-1 stimulation prevented Dlg1 and PTEN degradation. Indeed, blocking the SC proteasome increased Dlg1 and PTEN levels (fig. S5D). These proteins, and also AKT, were ubiquitinated in nonstimulated SCs and blocking the proteasome increased the amount of ubiquitinated proteins (fig. S5, E and F). Neuregulin-1 stimulation of these cells reduced the levels of ubiquitinated Dlg1 and PTEN (fig. S5E) but not of AKT (fig. S5F), which indicated that neuregulin-1 specifically prevents Dlg1 and PTEN ubiquitination and degradation, which leads to more Dlg1-PTEN complexes.

In mouse sciatic nerves, axonal neuregulin-1 type III stimulates myelination and the myelination rate [illustrated by the levels of myelin protein zero (P0) mRNA], and AKT activation (phospho-AKT levels) increases during the two first postnatal weeks (Fig. 3A and fig. S6). Dlg1 and PTEN levels also increased (Fig. 3A), whereas levels of polyubiquitinated Dlg1 and PTEN decreased (fig. S7), which suggests that neuregulin-1 stimulation also stabilizes Dlg1 and PTEN during myelination.

Fig. 3

PTEN, AKT1, and AKT2 participate in the regulation of myelin sheath thickness. (A) Relative amounts (±SEM) of Dlg1, PTEN, and phospho-AKT analyzed during myelination in mouse sciatic nerve for at least three mice. (For original data, see fig. S6.) P0 mRNA data are derived from (5). Maximal expressions used as reference. Lines: Extrapolated regression curves. (B) Diameters versus lengths of SCs infected with PTEN or control shRNAs viruses. Lines: Linear regressions for control (n = 42) line 1 and PTEN (n = 37) line 2 shRNAs. (C) Diameters versus lengths of SCs infected with Akt1, Akt2, or control shRNAs viruses. Lines: Linear regressions for values of control (line 1) and Akt1 or Akt2 (line 2) shRNAs. n = 39 (AKT1), 41 (AKT2), and 69 (control). (D) Mean cell diameter (±SEM), comparing silenced and nonsilenced SCs of the same length class (<300 μm) for n = 31 (AKT1), 34 (AKT2), and 22 (control). (E) Neuregulin-1 type III (Nrg1) on axons (green) stimulates erbB2 and erbB3 receptors in Schwann cells (brown). Receptor activation stimulates the PI3K-AKT pathway (blue arrows) promoting myelination. Neuregulin-1 stimulation also stabilizes Dlg1, which interacts with and stabilizes PTEN (red arrows). More PTEN is then available to inhibit AKT activation, which results in curbing of myelination.

Although murine myelination continues for several months, the myelination rate and AKT activation decrease around 2 weeks after birth (Fig. 3A) (5). This inhibition of myelination correlates with the plateau reached by Dlg1 and PTEN expression (Fig. 3A), consistent with the proposed inhibitory role of Dlg1 during myelination. To check the functional role of PTEN, we silenced its expression in SCs of sciatic nerves. As expected, reduction of PTEN expression correlated with an increase of phospho-AKT in infected cells (fig. S8). Myelinating SCs silenced for PTEN showed a thicker myelin sheath compared with control cells of similar length (Fig. 3B; 7.6 ± 0.3 μm versus 6.1 ± 0.3 μm; length < 405 μm; P < 0.01). We conclude that PTEN works in concert with Dlg1 as myelination inhibitor.

Transgenic expression of constitutively active AKT in SCs by using the proteolipid protein promoter has been reported not to affect myelination (21). However, low expression levels or compensatory mechanism may have overshadowed an effect. Thus, using our experimental setting, we tested whether AKT, a target of Dlg1 and PTEN, is required for SC myelination. Silencing of the more ubiquitously expressed isoforms Akt1 or Akt2 (22) (fig. S9) resulted in shorter myelin sheaths (Fig. 3C), which showed that these proteins are required for myelination and/or myelin sheath maintenance. Cells silenced for Akt2 had significantly reduced diameters compared with control cells of similar lengths, whereas Akt1-silenced cells did not (Fig. 3D), which suggests that Akt2 is also required for correct myelin sheath thickness.

Taken together, our data indicate that a molecular mechanism involving the critical functions of Dlg1 and PTEN prevents SC overmyelination when they are stimulated with neuregulin-1 (Fig. 3E). Myelination is also regulated, however, by the amount of neuregulin-1 expressed in axons (3, 4) and the availability of erbB2 and erbB3 receptors on SCs (23, 24). To test whether changes in these parameters are primarily responsible for curbing the myelination rate, we silenced Dlg1 in myelinating SCs at 20 dPN, after major myelination is stalling. Three weeks later, SCs silenced for Dlg1 had produced significantly larger myelin sheaths (Fig. 4A; 8.5 ± 0.3 μm versus 5.3 ± 0.2 μm; P < 0.001), which showed that mechanisms stimulating SC myelination were still active at the time myelination was curbed. Thus, the Dlg1 brake appears to be the main mechanism reducing myelination at this age. When overmyelination was allowed to proceed for longer (6 weeks after injection), demyelination features were abundant, and only the shortest myelinated cells remained (fig. S10). Thus, the cells with the most oversized myelin are unstable, consistent with observations in several peripheral neuropathies (7).

Fig. 4

Dlg1 is essential to curb myelination. (A) Dlg1 or control shRNAs adenoviruses were injected in mice after myelination slowdown (20 dPN). (Graph) Diameters versus lengths of infected SCs 3 weeks after injection. Lines: Linear regressions for values of control (n = 41) (line 1) and Dlg1 (n = 43) (line 2) shRNAs. (B) Hypermyelinated fiber expressing Dlg1 shRNA and PLAP (arrowheads) displays a typical outfolding (arrow). Scale bar, 800 nm. (C) Myelinating SC expressing Dlg1 shRNA and PLAP (arrowheads) show excessive and aberrant myelin deposition with large outfoldings. Scale bar, 2 μm. (D) Dlg1 or control shRNAs viruses were injected in sciatic nerves of MTMR2 mutant pups. (Graph) Diameters versus lengths of infected SCs. Lines: Linear regressions for values of control (n = 80) (line 1) and Dlg1 (n = 45) (line 2) shRNAs viruses.

Dlg1 interacts with the CMT4B-disease protein MTMR2 (8). We found that some Dlg1-silenced cells displayed small outfoldings (Fig. 4B) that occasionally expanded in large focally folded myelin sheaths, reminiscent of structural defects in CMT4B (Fig. 4C). This suggested that Dlg1 function might be altered in MTMR2 mutant SCs. Therefore, we silenced Dlg1 in SCs of the sciatic nerve of MTMR2 mutant mice. Dlg1-silenced cells were morphologically equal to control cells (Fig. 4D, diameter: 6.4 ± 0.4 μm versus 6.7 ± 0.3 μm, P = 0.47), showing that Dlg1 was not inhibiting myelination in these cells. We propose that the myelination brake triggered by Dlg1 is missing in CMT4B-diseased SCs, which results in overmyelination, defects in myelin deposition, and, finally, demyelination.

Over 50 years ago, Rushton showed in a pivotal paper (25) that myelin sheath diameter is optimally sized toward a certain NCV. The myelination brake involving Dlg1 and PTEN appears to be essential for this optimization, in addition to axonal stimulation.

Perturbation of this basic regulatory mechanism provides a mechanistic basis for the etiology of hereditary peripheral neuropathies characterized by hypermyelination, and the control of enzymatic activities involved in this mechanism, PTEN and AKT, may serve as therapeutic strategies for these diseases.

Supporting Online Material

Materials and Methods

Figs. S1 to S10


References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We thank J. Salzer for his help with cocultures, N. Mantei, C. Perrin-Tricaud for critically reading the manuscript, and the Light Microscopy Center and Electron Microscopy Center at ETH facilities of the ETHZ for technical help. Supported by the Swiss National Science Foundation (N.T. and U.S.), the National Centre of Competence in Research ”Neural Plasticity and Repair,” and ETH Zürich (ETHIIRA).
View Abstract

Stay Connected to Science

Navigate This Article