Beta1 integrin activates Rac1 in Schwann cells to generate radial lamellae during axonal sorting and myelination

Abstract

Myelin is a multispiraled extension of glial membrane that surrounds axons. How glia extend a surface many-fold larger than their body is poorly understood. Schwann cells are peripheral glia and insert radial cytoplasmic extensions into bundles of axons to sort, ensheath, and myelinate them. Laminins and beta1 integrins are required for axonal sorting, but the downstream signals are largely unknown. We show that Schwann cells devoid of beta1 integrin migrate to and elongate on axons but cannot extend radial lamellae of cytoplasm, similar to cells with low Rac1 activation. Accordingly, active Rac1 is decreased in beta1 integrin-null nerves, inhibiting Rac1 activity decreases radial lamellae in Schwann cells, and ablating Rac1 in Schwann cells of transgenic mice delays axonal sorting and impairs myelination. Finally, expressing active Rac1 in beta1 integrin-null nerves improves sorting. Thus, increased activation of Rac1 by beta1 integrins allows Schwann cells to switch from migration/elongation to the extension of radial membranes required for axonal sorting and myelination.

Figure 1.

DRG cultures from β1 integrin–null mice manifest impaired myelination. (A, A′, and A″) DRG cultures from mice with SC deletion of β1 integrin (phase in A) contain a mixture of β1 integrin–positive (asterisks) and –negative (arrows) SCs. Cells were stained with anti–β1 integrin (red) and anti-neurofilament (NF; green) antibodies, and with DAPI (blue) to visualize nuclei. β1 integrin–null SCs are found at distance from the neuronal cell bodies (A″). (B and C) Staining with anti-MBP antibodies reveals that SCs in mutant cultures synthesize fewer myelin internodes (F; P < 0.05 by t test, three plates per genotype for each experiment, n = 4). (D and E) Staining with anti-neurofilament antibodies (red) and DAPI (blue) shows that the numbers of axons and SC nuclei (small, cigar-shaped, and associated with axons) are not reduced in mutant cultures. (G) β1 integrin–positive and –negative cells (recognized using anti–β1 integrin antibodies; not depicted) were classified based on their relationship with axons. The number of β1 integrin–null SCs associated with axons was slightly reduced (mutant versus wt: P < 0.05 for red group, P < 0.01 for blue group, and P < 0.001 for yellow group by t test; n = 3,902 wt and 5,536 knockout [ko]). Red indicates not contacting, blue indicates not longitudinally oriented, and yellow indicates properly associated. Error bars indicate SEM. Bars: (A) 200 μm; (A′ and B–E) 100 μm; (A″) 50 μm.

Figure 2.

Reduced number of peripheral lamellipodia in β1 integrin–null SCs. (A) Staining with anti-talin (green) and anti–β1 integrin (red) antibodies reveals that β1 integrin–null SCs (arrowheads) have a smaller surface than wt cells (arrows) when spreading on laminin (Lam1), but not PLL or vitronectin (Vitr). (B) The surface of wt and β1 integrin–null SCs was measured on various substrates. Wt versus β1-null surface, by t test: P < 0.0001 on laminin, n = 86 wt, 40 knockout (ko); P = 0.64 on vitronectin, n = 85 wt, 69 knockout; P = 0.027 on PLL, n = 39 wt, 37 knockout. (C) The number of lamellipodia (asterisks in A) is reduced in mutant SCs plated on laminin, but not PLL or vitronectin (n = 674 wt, 278 knockout on laminin; n = 34 wt, 73 knockout on vitronectin; n = 26 wt, 39 knockout on PLL). (D–F) “Axial” lamellipodia are present at the end of the long axis of the cell, within of a 20° angle (arrows), whereas “peripheral/radial” lamellae are outside this zone (arrowheads). When spreading on laminin, β1 integrin–null SCs have a selective loss of peripheral lamellae. The number of axial lamellae was similar between wt and mutant SCs (1.9 ± 0.54 and 1.8 ± 0.07, respectively; P = 0.325 by t test), whereas the number of peripheral lamellae was significantly reduced in mutant SCs (3.5 ± 0.35 and 0.4 ± 0.11; P < 0.0001 by t test, n = 91 wt and 53 knockout). The length of extension of both axial and peripheral lamellipodia was reduced in β1 integrin–null SCs plated on laminin. Error bars indicate SEM. Bars, 40 μm.

Figure 3.

Radial lamellipodia in spreading SCs and myelination are dependent on high levels of active Rac1. Rat SCs spreading on laminin, PLL, or vitronectin were treated with different concentrations of the specific Rac1 inhibitor NSC23766. (A–C) Levels of GTP-bound Rac measured after Western blotting of PDB-GST pull downs using anti-Rac1 antibodies. Active Rac levels were decreased in a dose-dependent fashion after 30 (A and B; mean of three experiments) and 120 min of treatment (C; three experiments). (D–F, H–J, and L–N) Staining of cells using phalloidin after 30 (not depicted) or 120 min of treatment with 0 μm NSC23766 (D, H, and L; arrows indicate axial lamellipodia, and asterisks mark radial lamellipodia), 100 μM NSC23766 (E, I, and M; arrows indicate axial lamellipodia), and 300 μM NSC23766 (F, J, and N; arrows indicate axial processes lacking lamellipodia) on laminin (D–F), PLL (H–J), or vitronectin (L–N). Under these conditions, radial lamellipodia were present only on laminin. On laminin, we observed a dose-dependent decrease in the number of SC lamellipodia. Only radial lamellipodia were decreased at intermediate doses of inhibitor (100 μM; E and G) and active Rac1 (A and B; P < 0.0001 by t test), whereas axial lamellipodia were unchanged (G; P = 0.6 by t test). In contrast, both axial and radial lamellipodia were decreased at high doses of inhibitor (300 μM; F and G) and low levels of active Rac1 (A and B). n = 100 SCs. Error bars indicate SEM. (P–S) 50 μM of inhibitor were sufficient to inhibit myelination in mouse DRG. Staining for MBP is in green, neurofilament (NF) is in red, and DAPI is in blue. Bars, 100 μm.

Figure 4.

Rac1 activation and membrane targeting are deficient in β1 integrin–null SCs. (A–C) Expression and activation of small Rho GTPases during nerve development. (A) Western blot on sciatic nerves lysates for Rac1, Cdc42, and Rho. (B) Pull-down assay of PDB-GST for Rac and Cdc42, and GST-Rotekin for RhoA. Active proteins were normalized to total Rac1, Cdc42, and RhoA, and equal loading was verified by β-tubulin. (C) The relative levels of active Rho GTPases at P5 and in the adult were normalized to P1 levels. (D) Levels of active Rac1 in mutant nerves were normalized to total Rac1 levels and divided by relative levels of active Rac1 in wt nerves. Loading was verified by β-tubulin (not depicted). The levels of active Rac1 are reduced in β1 integrin–null nerves by about half. In similar assays, the levels of active Cdc42 were not reduced, whereas levels of active RhoA were reduced. Representative experiments from a minimum of three repetitions. Error bars indicate SEM. (E–J) SCs on laminin, treated with PBD-GST and saponin, and stained with anti-GST (green) and anti–β1 integrin (red) antibodies. The presence of PBD does not interfere with the capacity to form large lamellipodia (G). PBD is enriched at the leading edge of lamellipodia in β1 integrin–positive cells (E, arrows; enlarged in E′) but not in β1 integrin–null SCs (I; magnification in I′). (K–N) Staining of wt cells with anti-Cdc42 (K and M, red), and anti–β1 integrin antibodies (L, green; merge image in N) shows that under these conditions, Cdc42 does not contribute to the recruitment of PBD to the leading edge, as it is excluded from lamellipodia (asterisks show lamellipodia, and arrows show the restricted Cdc42 staining). (K and M) The localization of Cdc42 (green) in β1 integrin (red) negative cells is similar to that of wt cells. Bars: (E and F) 30 μm; (E′) 36 μm; (G and H) 41.4 μm; (I and J) 29 μm; (I′) 38.8 μm; (K and L) 35.8; (M and N) 50.6.

Figure 5.

Rac1 inactivation in SCs. (A) A conditional Rac1-null mouse (right) and a control littermate (left) are shown. Mutant mice show signs of neuropathy, such as clenching of the hind limbs (arrow). (B) PCR on genomic DNA shows that the Rac floxed allele was recombined at high efficiency in peripheral nerves (no rec, nonrecombined; rec, recombined), but not in liver or kidney. (C) Western blot analysis from Rac1-null mice and control sciatic nerves show reduction of Rac1 protein in mutant nerves (densitometric value of Rac/β-tubulin = 0.3 in mutant and 0.7 in wt). Nerves were stripped of perineurium (SN-P) to reduce contribution from perineurial cells. (D) Activation assays show that low Rac1 activity (possibly due to Rac3) remains. The level of active Cdc42 and RhoA are not different between wt and Rac1-null nerves. (E and F) Longitudinal sections of sciatic nerves from wt or Rac1-null nerves stained with DAPI shows that the density of nuclei was not decreased in mutant nerves (G; 378 ± 57 nuclei/μm² in mutant versus 207± 19 nuclei/μm² in wt nerves; P = 0.056 by t test). Error bars indicate SEM.

Figure 6.

Impaired lamellae and myelin formation by Rac1-deficient SCs in vitro. SCs isolated from P5 Rac-null (D–F) or wt (A–C) sciatic nerves were plated on laminin (A and D), PLL (B and E), or vitronectin (C and F) and stained using TRITC-conjugated phalloidin. Wt cells make more radial (asterisks) and axial (arrows) lamellipodia on laminin than on the other substrates. Both radial and axial lamellipodia formation are impaired in mutant SCs (G–I). n = 50 cell per condition. (J–N) DRG from mutant mice synthesize fewer myelin internodes per field (P < 0.01 by t test, n = 25 fields in both genotypes). Error bars indicate SEM. Bars: (A–F) 50 μm; (J–M) 100 μm.

Figure 7.

Inactivation of Rac1 in SCs delays sorting of axons and impairs myelination. Transverse semithin sections of wt (A–C) and mutant (D–F; Rac1 knockout [ko]) sciatic nerves at P5 (A and D), P10 (B and E), and P28 (C and F), and electron microscopy from mutant and wt sciatic nerves at P5 (G and H) and mutant nerves at P10 (I). In wt P5 nerves, few bundles of unsorted axons are still present (A and G, asterisks), whereas few fibers have SCs and axons in a one-to-one relationship (G, 1:1) and most large axons have thin myelin (A and G; a, axon; m, myelin). In contrast, P5 mutant nerves contain many large bundles of unsorted axons (D and H, asterisks), many fibers in a one-to-one relationship (D and H, 1:1), and no myelin (D and H). In P10 wt nerves, sorting is almost complete and myelin sheaths become thicker (B and not depicted). In contrast, P10 mutant nerves still contain frequent bundles of unsorted axons (E, arrows), including large caliber axons (e.g., double asterisks on an axon of 3 μm shown in I). By P10, many more one-to-one fibers are seen (E and I, 1:1) and myelination begins (E). Promyelinating SCs show aberrant processes directed away from the axon (I, arrowheads). At P28, myelination is complete in wt nerves (C). Mutant nerves show hypomyelination (F). Bar: (A–F) 20 μm; (G and H) 5 μm; (I) 1.5 μm.

Figure 8.

Infection with an adenovirus expressing CA-Rac1 improves the sorting phenotype of β1 integrin conditional null nerves. Adenoviruses expressing CA-Rac1 were injected into the endoneurium of β1 integrin–null or wt P10 sciatic nerves at the common peroneal/tibialis bifurcation. At P22, transverse semithin sections were examined 2 mm distal (+1; D and H) or 1.5 mm (−1; C, G, K, and L), 3 mm (−2; B, F, I, and J), and 4.5 mm (−3; A and E) proximal to the injection site. At this age, injection of adenovirus or saline produced minimal damage or inflammation in wt (I and J) and mutant (A–H, K, and L) nerves. In mutant nerves treated with CA-Rac1, the number and extension of bundles of unsorted axons was reduced as compared with saline-treated nerves (compare E–H to A–D, respectively). The effect was more obvious near the injection site (−1 and +1). Here, as shown in the enlarged inset of C and G, many axons in the bundles had been sorted, ensheathed, and even thinly myelinated (compare L with K). Asterisks in K and L indicate unsorted axon bundles, and arrowheads point to one-to-one promyelinating fibers. (M) Quantification of the rescue. The area of unsorted bundles significantly decreased in CA-Rac1–treated nerves (P < 0.0001 by t test, n = 278 [control] and 305 [CA-Rac1] bundles). Although the fraction of one-to-one fibers significantly increased in CA-Rac1–injected nerves (P < 0.02 by t test; n = 7 animals), the number of myelinated fibers did not. Error bars indicate SEM. (N) Expression of the HA-tagged CA-Rac1 virus confirmed by Western blot analysis. HA-Rac was expressed in nerves injected with CA-Rac at the expected size, but not in nerves injected with saline. Control HA shows the positive control for the HA antibodies (lysates from cells transfected with a construct coding for a different HA fusion protein). Bar: (A–J) 50 μm; (K and L) 9 μm.

Figure 9.

Model for β1 integrin–mediated Rac1 activation in peripheral nerve development. (A) Low levels of Rac generate axial lamellipodia during SC migration and elongation on axons. (B) Increase in Rac levels generates several radial/peripheral lamellipodia (curved arrows) that allow SCs to sort and myelinate axons. (C) Reconstruction of an SC during radial axonal sorting in newborn nerve (modified from Webster, H.D., R. Martin, and M.E. O'Connell. 1973. Dev. Biol. 32:401–416 with permission from Elsevier), showing the radial cytoplasmic processes (lamellae; curved arrows) between and around axons. We propose that, similar to radial lamellipodia in cultured cells, the switch from longitudinal elongation to radial process extension is mediated by β1 integrins via activation of Rac1.