α_V integrins in Schwann cells promote attachment to axons, but are dispensable in vivo

Abstract

In the developing peripheral nervous system, Schwann cells (SCs) extend their processes to contact, sort, and myelinate axons. The mechanisms that contribute to the interaction between SCs and axons are just beginning to be elucidated. Using a SC-neuron coculture system, we demonstrate that Arg-Gly-Asp (RGD) peptides that inhibit α_V-containing integrins delay the extension of SCs elongating on axons. α_V integrins in SC localize to sites of contact with axons and are expressed early in development during radial sorting and myelination. Short interfering RNA-mediated knockdown of the α_V integrin subunit also delays SC extension along axons in vitro, suggesting that α_V-containing integrins participate in axo-glial interactions. However, mice lacking the α_V subunit in SCs, alone or in combination with the potentially compensating α₅ subunit, or the α_V partners β₃ or β₈, myelinate normally during development and remyelinate normally after nerve crush, indicating that overlapping or compensatory mechanisms may hide the in vivo role of RGD-binding integrins.

1 |. INTRODUCTION

Successful myelination in the peripheral nervous system (PNS) depends on the ability of Schwann cell (SC) glia to engage in cell-to-cell contact with neurons. This contact is important because some of the molecular cues that instruct immature SCs to differentiate into promyelinating SCs derive from axons, and SCs gain access to these cues by direct axon contact. Though several ligand-receptor pairs at the axo-glial interface have been identified (reviewed in Feltri, Poitelon, & Previtali, 2016; Monk, Feltri, & Taveggia, 2015), the molecular composition of this junction is far from complete.

Initially, precursor SCs receive survival and mitogenic signals from axons as they contact and migrate along them at the start of development (Jessen & Mirsky, 2005). The interaction with groups of axons then triggers immature SCs to deposit and organize basal lamina components such as laminins onto their ab-axonal surface (away from the axon). This basal lamina then becomes necessary to initiate radial sorting, a key premyelination step in the PNS (Feltri & Wrabetz, 2005). During radial sorting, immature SCs extend lamellipodia-like processes toward bundled axons to sort and separate larger axons destined for myelination from smaller axons destined for ensheathment by nonmyelinating Remak SCs (reviewed in Feltri et al., 2016). Finally, as a promyelinating SC attains 1:1 relationship with an axon, it extends its processes concentrically around as well as laterally along the axon to elaborate myelin (reviewed in Salzer, 2015). Fundamentally, SC development and the myelination of axons are contact-mediated phenomena.

The failure of SCs to generate the lamellipodia-like processes necessary to engage axons underlie many radial sorting and myelination defects in mice (Feltri et al., 2016). SC-specific deletion of small RhoGTPases and other cytoskeletal regulators, for instance, leads to myelination abnormalities (Benninger et al., 2007; Elbaz et al., 2016; Jin et al., 2011; Montani et al., 2014; Nodari et al., 2007; Novak et al., 2011). Also, ablating SC laminins or their cognate receptors impairs process extension; this impedes axon-SC interactions and results in profound myelination defects (Feltri et al., 2002; Yu, 2005; Yu, Chen, North, & Strickland, 2009). Many laminin receptors are integrins, Type-1 transmembrane molecules composed of α/β heterodimers that mediate cell–cell or cell–extracellular matrix (ECM) adhesion (Hynes, 1987). They link intracellular cytoskeletal components with extracellular components and govern numerous morphogenetic signaling pathways (Hynes, 2002b). We showed previously that conditionally deleting β₁-containing integrins in SCs (β₁ cKO) produces severe radial sorting defects in vivo (Feltri et al., 2002), yet deleting α₃β₁, α₆β₁, and α₇β₁ (three major laminin-binding integrins expressed by SCs) in single or double combinations only mildly affects radial sorting (Pellegatta et al., 2013; Previtali et al., 2003). These suggested that perhaps the inactivation of β₁ heterodimers that do not primarily bind laminins contributed to the severity of the β₁ cKO phenotype. Furthermore, β₁ heterodimers, which become strictly localized to the ab-axonal SC surface as myelination progresses (Feltri et al., 2002; Pellegatta et al., 2013), were also localized on SC processes ad-axonally (juxtaposed to the axon) while radial sorting was being performed (Berti et al., 2011). This suggested some β₁ heterodimers might function at the axo-glial interface independently of laminins.

These investigations made us focus on the largely uncharacterized roles of integrins that may bind other ligands besides laminins and led us to search for new SC molecules at the axo-glial junction (Poitelon et al., 2015). Ligands that bear an accessible RGD (Arg-Gly-Asp) or RGD-related tripeptide motif, which are quite prevalent, are recognized by some β₁-containing integrins as well as by all members of the second largest subfamily of integrins, the α_V-containing integrins (Supplemental Information Figure S1a), which are expressed by cultured SCs (Aumailley, Gerl, Sonnenberg, Deutzmann, & Timpl, 1990; Chernousov & Carey, 2003; Humphries, 2006; Milner, Wilby, et al., 1997). RGD-binding integrins like those containing α_V were long thought to be important for developing SCs and oligodendrocytes, the myelinating glia of the central nervous system (CNS). For instance, RGD peptides (decoys that block RGD receptors) hamper SC migration in vitro, suggesting RGD-binding integrins may participate in SC migration early in development (Milner, Wilby, et al., 1997). Moreover, oligodendrocyte myelination seems RGD dependent, as adding RGD peptides inhibits myelin synthesis in vitro (Cardwell & Rome, 1988). Indeed, several α_V heterodimers are postulated to participate in oligodendrocyte development (Blaschuk, Frost, & Ffrench-Constant, 2000; Milner, Frost, et al., 1997; Milner, Edwards, Streuli, & Ffrench-Constant, 1996; Milner & Ffrench-Constant, 1994; Shaw, Milner, Compston, & Ffrench-Constant, 1996). In vivo ablation of the integrin α_V subunit in the CNS and PNS using Nestin-Cre results in axonal degeneration in the CNS (McCarty et al., 2005; Mobley, Tchaicha, Shin, Hossain, & McCarty, 2009; Zimmerman et al., 1994). Interestingly, these mutant mice also develop limp paresis, which could be due to axonal Nestin-Cre expression, but it is also characteristic of impaired PNS myelination (Michelson, Russell, & Harman, 1955).

While RGD-binding integrins canonically bind ECM molecules like fibronectin (FN), vitronectin, fibrin, and the TGFβ complex, to name a few, reports from other systems show that α_V-containing integrins also bind cell surface molecules like Neuregulin-1 (NRG1) and L1 cell adhesion molecule, which are important axon-derived molecules in the PNS (Haney et al., 1999; Ieguchi et al., 2010; Itoh et al., 2005; Montgomery et al., 1996; Taveggia et al., 2005). In the CNS, α_Vβ₃ on glial astrocytes binds axon-derived Thy1 to promote formation of astrocytic focal adhesions necessary for neuron-astrocyte communication (Leyton et al., 2001). Furthermore, axonal molecule ADAM22, which belongs to a family that binds RGD-binding integrins, is speculated to bind SC integrins in order to properly function in PNS myelination (Ozkaynak et al., 2010; Seals & Courtneidge, 2003). All together, we hypothesized that RGD-binding integrins could play a role in the PNS myelination process.

Here, we analyzed the role of RGD-binding integrins, particularly α_V-containing integrins, during SC development. Using an established in vitro coculture system that recapitulates axon-SC interactions in vivo, we demonstrate that α_V-containing integrins contribute to process extension of SCs. However, SC-specific ablation of α_V integrin in vivo has no detectable consequences.

2 |. MATERIALS AND METHODS

2.1 |. Animal models

All experiments involving animals followed experimental protocols approved by San Raffaele Hospital and Roswell Park Institute Animal Care and Use Committees. Itgav (α_V)-floxed mice: Itgav^f/+ mice of mixed C57BL6/129S4/FVB background (Lacy-Hulbert et al., 2007; McCarty et al., 2005) were mated to generate Itgav^f/f or crossed with Mpz-Cre (P₀-Cre) transgenic mice of congenic C57BL6 background (Feltri, D’Antonio, Previtali, et al., 1999; Feltri, D’Antonio, Quattrini, et al., 1999) to generate Itgav^f/+; P₀-Cre^tg/+. Itgav^f/f were then crossed with Itgav^f/+; P₀-Cre^tg/+ to produce Itgav^f/f; P₀-Cre^tg/+. Resulting offspring were of mixed background. Itgav^f/f or Itgav^f/+ were used as controls. Itga5/Itgav (α₅/α_V)-floxed mice: Itga5-floxed mice, provided by van der Flier et al. (2010), were maintained in congenic C57BL6 background in our lab. We crossed Itga5^f/f with Itgav^f/f; P₀-Cre^tg/+ to produce Itga5^f/+; Itgav^f/+; P₀-Cre^tg/+, which were then crossed with Itga5^f/f; Itgav^f/f to produce Itga5^f/f; Itgav^f/f; P₀-Cre^tg/+. Itga5^f/f and Itgav^f/f were used as controls. Resulting offspring were of mixed background. Itgb3 (β₃) mice: Male and female Itgb3^+/−- mice of congenic C57BL6 background were purchased from The Jackson Laboratory (B6.129S2-Itgb3^tm1Hyn/JSemJ, Stock No. 008819) and mated to produce Itgb3^−/− mutants and Itgb3^+/+ controls. Itgb8 (β₈) mice: Itgb8^−/− mutants and Itgb8^+/+ controls of mixed C57BL6/129S4/ICR/CD1 background were described previously (Mobley et al., 2009). In all analyses, males and females were included, and only littermates or mice with the same parents were compared. PCR genotyping was performed in-house with phenol/chloroform-extracted DNA from tails or sciatic nerves, as previously described (Feltri, D’Antonio, Previtali, et al., 1999; Lacy-Hulbert et al., 2007; Mobley et al., 2009; van der Flier et al., 2010).

2.2 |. Cell culture

All coculture experiments strictly contained only pure populations of SCs and dorsal root ganglia (DRG) neurons. Purified rat SCs: Purified populations of SCs (at >99.5% purity) were derived from Sprague–Dawley P3 rat sciatic nerves (Envigo or Taconic) using method in Brockes, Fields, and Raff (1979) and as performed in Feltri, Scherer, Wrabetz, Kamholz, and Shy (1992). Secondary SCs (stored in liquid N₂) were resuspended in SC media (DMEM, 10% fetal bovine serum [FBS], 2 μM forskolin, 2 mM L-glutamine, 2 ng/ml, Neuregulin-β1, 100 U/ml penicillin, and 100 μg/ml streptomycin [penn/strep]), plated on 100 mm culture dishes coated with 0.1 mg/ml poly-L-lysine (PLL) (Sigma-Aldrich), and maintained at 37°C in 5% CO₂ (37°/5%). Purified rat DRG neurons: Pure DRG neurons were achieved as described in Einheber, Milner, Giancotti, and Salzer (1993). DRGs were isolated from Sprague–Dawley embryonic day (E) 15.5 rat embryos, dissociated with 0.25% trypsin in DMEM for 45 min at 37°/5%, pelleted, then further disaggregated with 20 gentle triturations in C-media (MEM, 10% FBS, 2 mM L-glutamine, 4 g/L D-glucose, 50 ng/ml nerve growth factor [NGF] [Harlan Bioproducts for Science]) supplemented with penn/strep. DRG suspension was then plated on 12 mm coverslips coated with 0.5 mg/ml rat collagen Type I (R&D Systems) at a density of 1.5 DRGs/coverslip. Coverslips were incubated at 37°/5%. After 24 hr, C-media were replaced by NB media (Neurobasal media, 2 mM L-glutamine, 4 g/L D-glucose, B27 Supplement 1X [Gibco], 50 ng/ml NGF) supplemented with 10 μM FUDR (Sigma-Aldrich) (NBF media) to eliminate endogenous, nonneuronal cells. After 48 hr, NBF media were replaced by NB media. DRG explants were maintained in alternating NBF/NB media for 2–3 cycles. These treatments resulted in pure DRG neurons devoid of all other cell types. Mouse SC isolation (for SMAD4 analysis in α_V SCs and for real-time quantitative PCR [RT-qPCR] of α₅/α_V SCs): Sciatic nerves were sampled from Itgav^f/f and Itgav^f/f; P₀-Cre^tg/+ littermates or Itga5^f/f; Itgav^f/f and Itga5^f/f; Itgav^f/f; P₀-Cre^f/f littermates at P17 and placed in chilled Leibovitz’s L-15 supplemented with penn/strep. Nerves were desheathed and teased under stereodissecting microscope. Teased nerves were dissociated in 0.25% DispaseII, 0.05% Type I collagenase, high-glucose DMEM overnight (O/N) at 37°C in 9% CO₂ (37°/9%). Next day, dissociated nerves were gently triturated, then 20 volumes of DMEM were added to hamper digestion. Suspension was filtered through 70 μm cell strainer. Cells were pelleted, gently resuspended in SC media, then plated on 12 mm coverslips or 35 mm dishes (previously coated with 0.1 mg/ml PLL (30 min room temperature [RT]), followed by 30 μg/ml laminin (Sigma-Aldrich L2020) O/N at 4°C) and incubated O/N at 37°/9%. For next 2 days, cells were washed with PBS to remove debris and replaced with new SC media. On the third day, fibroblasts were eliminated by incubating cells with rat antimouseThy1.2 antibody (Ab) (AbD Serotec MCA1474) diluted in DMEM for 30 min at 37°/9%, followed by rabbit complement [1:500] (EMD Millipore 234400) for 40 min at 37/9%. Cells were washed, then maintained in SC media at 37°/9% for 2 days. Complement treatment was repeated to achieve maximal fibroblast removal.

2.3 |. Competitive antagonist assay

Purified cultured rat SCs were washed 3x with PBS, trypsinized, pelleted, then gently resuspended in C-media with 20 gentle triturations. Antagonist peptides isoDGR (Curnis et al., 2006; Ghitti et al., 2012; Spitaleri et al., 2008), ac-isoDGR (Curnis et al., 2010), or negative control ARA peptides (200 μg/ml) were added to the SC suspension and incubated for 13 min at RT. Meanwhile, coverslips of purified rat DRG neurons were washed 3x with PBS. After 13 min, SC suspension (with peptides) was seeded onto DRG coverslips at a density of 20,000 SC/coverslip (200 μl drop/coverslip), then incubated for 3.5 or 16 hr at 37°/5%. Cocultures were then processed for immunostaining (see below). Five experiments were performed over 2 days for all conditions. To quantify SC length in Fiji/ImageJ, a line was drawn from tip to tip of bipolar SC (always going through nucleus), then measured. For multipolar SC, a line was drawn between the two longest observable tips (going through nucleus).

2.4 |. Immunoblots

Cultured SCs, cultured DRG neurons, or SC/DRG cocultures were washed 3x with PBS to remove nonadherent entities, then lysed from the plate with lysis buffer (95 mM NaCl, 25 mM Tris–HCl pH 7.4, 10 mM EDTA, 2% SDS, 1 mM Na₃VO_4, 1 mM NaF, Protease Inhibitor Cocktail [1:100] [Sigma-Aldrich]). Mouse sciatic nerves were sampled at indicated age, stripped of their epineurium (desheathed), snapfrozen, pooled, pulverized, then resuspended in lysis buffer. All lysates were boiled at 100°C for 5 min, then centrifuged at 14,000 rpm for 15 min at 16°C. Protein concentration was determined using BCA protein assay (Thermo Scientific). Homogenates containing 20 μg of protein were diluted in 4x Laemmli sample buffer (250 mM Tris–HCl pH 6.8, 8% SDS, 8% β-mercaptoethanol, 40% glycerol, 0.02% bromophenol blue, H₂O). Samples were boiled at 100°C for 5 min, centrifuged at 14,000 rpm for 2 min at RT, resolved on SDS-PAGE, then transferred to PVDF membranes. Membranes were blocked with 5% dry milk in 0.1% Tween-20/TBS (TBS-T) or 5% BSA in TBS-T for 1 hr shaking at RT, then incubated with primary Ab diluted in blocking solution O/N shaking at 4°C. The following primary Abs were used: rabbit anti-α_V integrin [1:1,000] (Cell Signaling #4711), rabbit anti-β₃ integrin [1:1,000] (Cell Signaling #4702), rabbit anti-β₅ integrin [1:1,000] (D24A5, Cell Signaling #3629), rabbit anti-β₆ integrin [1:1,000] (Proteintech 19695–1-AP), rabbit anti-β₈ integrin [1:5,000] (McCarty et al., 2005), rabbit anti-α₅ integrin [1:1,000] (EMD Millipore AB1928), rabbit anti-Sox10 [1:1,000] (Cell Signaling #89356), rabbit anti-GAPDH (Sigma-Aldrich G9545), rabbit anti-Calnexin (Sigma-Aldrich C4731), mouse anti-β-tubulin. (Sigma-Aldrich T4026). Membranes were washed with TBS-T, then incubated with appropriate HRP-conjugated secondary Ab diluted in 1% dry milk/TBS-T or 1% BSA/TBS-T for 1 hr shaking at RT. ECL substrate (Pierce ECL, Thermo Fisher) was applied to membrane, then developed on autoradiography films (GE Healthcare). Immunoblots were quantified using Fiji/ImageJ. Specificity of the α_V and β₃ integrin Abs was tested and confirmed by short interfering RNA (siRNA)-mediated knockdown of α_V or β₃ in SCs (Supporting Information Figure S4c,e). Specificity of β₈ Ab was confirmed previously (McCarty et al., 2005; Mobley et al., 2009). For β₅ and β₆ integrin Abs, we exploited α_V’s absence in DRG neurons and purposed these cells as negative controls as a way to infer the specific β₅ and β₆ bands in our immunoblots.

2.5 |. siRNA-mediated knockdown of integrins in SCs

Purified cultured rat SCs were plated at a density of 500,000/well on a six-well plate coated with 0.1 mg/ml PLL and incubated in SC media at 37°/5% O/N. SCs were transfected when 40% confluent (next day) with 33 nM of four siRNA targets against integrin subunit α_V (Thermo Scientific J-091817–09, 10, 11, and 12) or β₃ (Thermo Scientific J-097040–05, 06, 07, and 08) using Lipofectamine 2000 (Thermo Fisher), then incubated for 6 hr at 37°/5%. Cells were thereafter maintained in SC media at 37°/5% for 72 hr. To verify protein knockdown, SCs were washed, lysed from the plate, then processed for immunoblot as described above. To seed SCs onto DRG neurons, SCs were washed 3x with PBS, trypsinized, pelleted, then resuspended in C-media with 20 gentle triturations. SC suspension was seeded onto pure rat DRG coverslips (400 μl/coverslip, at a density of 20,000 SC/coverslip), then incubated for 3.5 or 24 hr at 37°/5%. Cocultures were then processed for immunoblot or immunostaining. The following number of experiments was performed over 3 days: scrambled = 8, si-α_V^#2 = 8, si-α_V^#2 = 6, si-α_V^#2 = 4. siRNA-treated SCs on FN: 12 mm coverslips were coated with 10 μg/ml FN (Sigma-Aldrich F2006) diluted in PBS and stored O/N at 4°C. siRNA-treated SCs were washed 3x with PBS, trypsinized, pelleted, then resuspended in SC media with 20 gentle triturations. SC suspension was plated on dried FN coverslips at a density of 40,000 SC/coverslip (80 μl drop/coverslip), then incubated for 3.5 hr at 37°/5%. Cells were then processed for immunostaining (see below). Quantification of SC length in 3.5 hr cocultures: same as competitive antagonist assay. Quantification of cell number in 24 hr cocultures: Each field was imaged under ×10 objective of Leica DM6000B microscope. All SCs in the field were associated with axons. Only SC DAPI was quantified in Fiji/ImageJ.

2.6 |. Immunostaining

Immunostaining of cultured SC or sciatic nerves: OCT-embedded sciatic nerves were sliced into 8 μm-thick sections with Cryostat and stored in −20°C for no more than 48 hr prior to immunostaining. Sections were thawed for 5 min, then submerged in PBS for 5 min. Cultured cells were washed 3x with PBS. Cells or nerve sections were fixed with 4% paraformaldehyde (PFA)/PBS for 10 min RT, permeabilized with 0.2% Triton-X/PBS for 10 min RT, then incubated in blocking solution for 1 hr RT. Cells or nerves were incubated O/N at 4°C with primary Abs diluted in blocking solution. The following primary Abs were used: rabbit anti-S100 [1:150–1:200] (Dako Z0311), goat anti-SOX10 [1:400] (R&D Systems AF2864), rabbit anti-SMAD4 [1:150] (Proteintech 51069–2-AP), rabbit anti-Ki67 [1:400] (Cell Signaling #9129), rabbit anti-phospho-Histone H3 (PH3) [1:400] (EMD Millipore 06–570). Cells or nerves were washed, incubated with appropriate secondary Ab diluted in blocking solution for 1 hr at RT, washed, stained with DAPI [1:10,000] (Sigma-Aldrich D9542) for 3 min RT, washed, mounted with Vectashield, then sealed. Immunostaining of co−/cultured SC for α_V: Cultured SCs were dyed with CellTracker fluorescent green probe (Invitrogen C7025), following manufacturer’s protocol. Dyed SCs were then seeded on DRGs for experiment (as described above). Cocultures were washed to remove nonadherent entities, then permeabilized with chilled methanol for 1.5 min at RT. Methanol was thoroughly aspirated, then cells were washed, permeabilized with 0.2% Triton-X/PBS for 10 min RT, washed, incubated in blocking solution (20% FBS, 2% BSA, 0.1% Triton-X, PBS) for 2.5 hr at RT, then washed again. Cells were then incubated for 1.5 hr at RT with primary Ab diluted in blocking solution. The following Abs were used: rabbit anti-α_V integrin [1:150] (EMD Millipore AB1930), rat or chicken anti-Neurofilament [1:500–1:1,000] (EMD Millipore MAB5448, Biolegend 822701), mouse anti-tubulin [1:200] (Sigma-Aldrich T4026). Cells were washed, then incubated with secondary Ab diluted in PBS for 1 hr RT. Remaining steps were same as above. Teased fiber immunostaining: Fixed fibers were submerged in chilled acetone for 5 min, then incubated in blocking solution (5% fish skin gelatin, 0.5% Triton-X100, PBS) for 1 hr RT. Remaining steps were same as general immunostaining above. Primary Abs used: rabbit anti-K_v1.1 (Alomone Labs APC-009), mouse anti-Nav Pan (Sigma-Aldrich S8809). Ki67 and PH3 immunostaining: Two nerves per animal, two animals per genotype were analyzed. The whole length of sampled P3 nerve (~3 mm) was immunostained, imaged, and quantified. All immunostainings were imaged using Confocal Leica TCS SP5 II microscope and Leica software. All images are Z-series projection comprised of 0.5–1 μm slices.

2.7 |. SMAD4 analysis in SCs

TGFβ stimulation was performed as described (D’Antonio et al., 2006). Briefly, SCs isolated from Itgav^f/f or Itgav^f/f; P₀-Cre^tg/+ mice were treated with recombinant mouse TGFβ1 (10 ng/ml) (R&D Systems 7666-MB-005) for 30 min at 37°/5%, washed, fixed with 4% PFA, then immunostained and imaged as described above. For Itgav^f/f, 2,391 SCs were quantified from five animals. For Itgav^f/f and P₀-Cre^tg/+, 1,955 SCs were quantified from five animals. SCs from each animal were plated in duplicate.

2.8 |. Real-time quantitative PCR

RNA was isolated from purified SCs derived from desheathed sciatic nerves of Itga5^f/f; Itga5^f/f or Itga5^f/f; Itga5^f/f; P₀-Cre^tg/+ postnatal day (P)17 and P6 mice using RNeasy Plus Micro Kit (Qiagen) and TRIzol (Thermo Fisher), respectively. cDNA was synthesized using Super-Script III Reverse Transcriptase kit (Life Technologies). Expression levels of α₅ integrin were measured using Taqman Gene Expression Assay kit (Applied Biosystems, TaqMan Assay ID Mm00439797_m1). 18S rRNA was used as endogenous control (Applied Biosystems, TaqMan Assay ID Hs99999901_s1).

2.9 |. Morphological analysis

Mutant and control littermates were euthanized at the age indicated. The sciatic nerve segment was sampled from approximately the same position for every age cohort. Routine semithin and ultrathin (electron micrograph [EM]) analyses were performed as described (Quattrini et al., 1996). Briefly, sciatic nerves were fixed in 2% buffered glutaraldehyde, then postfixed in 1% osmium tetroxide. After alcohol dehydration, nerves were submerged in propylene oxide, then in a 1:1 mixture of Epon-propylene oxide. Nerves were embedded in 100% Epon, and resin was allowed to polymerize. Semithin transverse sections were sliced 0.5-μm-thick using Leica UC7, stained with 2% toluidine blue, then examined by light microscopy with Leica DM6000B. EM transverse sections were sliced 700–900-Å-thick using Leica UC7, stained with uranyl acetate and lead citrate, then examined with electron microscope (model FEI BioTwin). Analyzed sections were sliced from the distal end of embedded sciatic nerve. Images acquired from semithins and EMs were nonoverlapping and comprehensive. To determine myelin thickness, the G-ratio (axon diameter/fiber diameter, where fiber = axon + myelin) was quantitated manually from EMs. In Fiji/ImageJ, a segmented line was drawn along the border of each axon, then its circumference was measured. The process was repeated for the whole fiber (axon including myelin), drawing the segmented line along the outer myelin border. Diameter was extracted using d = C/π. Axon diameter was divided by the fiber diameter to obtain G-ratio. Number of myelinated fibers was counted per EM field of view, then averaged. Any axon with surrounding myelin, regardless of thickness, was considered myelinated. Number of axon/SC units in 1:1 relationship was counted per EM field of view, then averaged. Any axon ensheathed (but not myelinated) by an individual SC was considered 1:1.

2.10 |. Internodal length

The length between two internodes was measured from teased sciatic nerve fibers sampled from adult α_V (P33) or α₅/α_V (P152) mice. Teased sciatic nerve fibers were prepared as previously described (Occhi et al., 2005). Briefly, control and mutant nerves were dissected, then immersed in chilled 4% PFA for 30 min. Fixed nerves were stored in PBS at 4°C until teasing (within 1 week). After epineurium removal, nerve fibers were gently teased apart with fine tungsten pins on 3-aminopropyltriethoxy-sylane-treated slides. All teasing steps were done under a stereodissecting microscope. Teased fibers were then allowed to dry on slides for at least 4 hr RT. Fibers were processed for K_v1.1 and PanNaV immunostaining and imaged as described above. Length of internodes was measured in Fiji/ImageJ. A segmented line was drawn from one (PanNaV) node to the next, then length was measured.

2.11 |. Nerve crush

Nerve crush was performed on adult mice using aseptic technique under a biobubble hood. Mice were anesthetized with isofluorane, and eyes were lubricated. While mouse laid flat on its dorsal side, the region of intended crush was shaved and cleaned with betadine and ethanol. On one leg, a small incision was made in the skin and muscle above the approximate position of sciatic nerve. Once nerve was exposed, any membranous ties to the muscle were severed using small scissors. A pair of straight serrated forcep clamps was dipped in liquid N₂ for 2 s, then nerves were tightly compressed with chilled clamps for 20 s. Crush site was marked with bromophenol blue powder, and process was repeated a second time on same crush site. Nerve was crushed, but not severed. Position of crush was approximately the same for every cohort. Other leg was unoperated. Injured skin was closed with surgical glue, then buprenorphine was administered subcutaneously as post-operatory analgesia. Animals were monitored daily after surgery and analgesics were administered as necessary. Then, 15 or 30 days postcrush, three segments of the operated nerve were sampled: (a) the segment proximal to the crush site, (b) the crush site, and (c) the segment distal to the crush site. Sciatic nerve from unoperated leg was also sampled. Nerves were prepared for semithin sectioning (as described above). Sections of all segments were analyzed, but only distal segments were quantified. Number of myelinated fibers was counted per semithin field of view, then averaged. Ages of animals used: α_V T15, 4 months; α_V T30, 3 months; α₅/α_V T15, 5 months; α₅/α_V T30, 4 months.

2.12 |. Image acquisition and statistics

Researcher was blind to the genotype during image acquisition and quantification. Images acquired were nonoverlapping. For antagonist and siRNA coculture experiments, areas of coverslip too congested with neurons were neither imaged nor included in quantification. All statistics were done in GraphPad Prism (ver 7.0). Specific analysis is stated in the legend. ****p < .0001, ***p < .001, **p < .01, *p < .05, ns = not significant (p > .05). All images are representative fields.

3 |. RESULTS

3.1 |. Antagonists of RGD-binding integrins hinder extension of SCs on axons in vitro

Previously it was shown that RGD peptides inhibit the process extension of SCs plated on FN (Milner, Wilby, et al., 1997). To determine if RGD-binding integrins are also involved in axo-glial interactions, we pretreated WT SCs with competitive antagonists against RGD-binding integrins, plated them on untreated WT DRG neurons, and assessed their interaction. SCs were pretreated with either (a) ARA, a cyclic Ala-Arg-Ala nonspecific peptide (negative control), (b) isoDGR, a cyclic isoAsp-Gly-Arg peptide with high affinity for α_Vβ₃, or (c) ac-isoDGR, a cyclic, acetylated form of isoDGR with similar affinities (thus less selective) for several RGD-binding integrins.

After 3.5 hr, ARA-treated SCs attached to axons and generated lengthy processes that aligned with one axon or contacted multiple axons (Figure 1a); they appeared like untreated SCs at 3.5 hr (Supporting Information Figure S1d). In contrast, treatment with isoDGR or ac-isoDGR hindered process extension: SCs attached to axons, but were visibly shorter and more stunted compared to ARA-treated SCs (Figure 1b–d). Those with little to no protrusions (stunted) and those with only short processes significantly increased in isoDGR and ac-isoDGR cohorts (Figure 1e). Furthermore, those with long processes significantly decreased. These demonstrate that RGD-binding integrins are involved early in axo-glial interactions and suggest the association of axons and SCs may be in part RGD dependent.

FIGURE 1.

Antagonist peptides that block RGD-binding integrins delay process extension of Schwann cells plated on axons. (a–c) WT cultured SCs pretreated with either (a) ARA (negative control), (b) isoDGR, or (c) ac-isoDGR antagonists. Peptides were seeded on untreated WT DRG neurons for 3.5 hr, then fixed and costained with anti-S100 Ab (green, labels SCs), anti-neurofilament (NF) Ab (red, labels neurons), and nuclear DAPI (blue). (a′–c′) Magnified area of (a–c). Scale bar is 50 μm in all panels. (d) Average (Avg) SC length in each cohort (student’s t test; ARA: n = 5 replicates, 1,880 total cells quantified; isoDGR: n = 5 replicates, 2,305 total cells quantified; ac-isoDGR: n = 5 replicates, 2,122 total cells quantified; SEM error bars). Compared to ARA, the average SC length decreased in isoDGR (−36%) and ac-isoDGR (−41%) cohorts. (e) SC lengths categorized into bins. In isoDGR and ac-isoDGR cohorts, significantly more SCs were stunted (<25 μm), more had short processes (25–50 μm), and fewer lengthened (50–100 μm) compared to ARA cohort (ordinary two-way analysis of variance (ANOVA), n = 5 replicates, SEM error bars). DRG, dorsal root ganglia; SC, Schwann cell

Since isoDGR- and ac-isoDGR-treated SCs in cocultures of 3.5 hr resembled untreated SCs in cocultures of 1 hr (Supporting Information Figure S1b), we asked whether the antagonist peptides simply delayed or completely arrested process extension. In prolonged cocultures, isoDGR and ac-isoDGR cohorts did lengthen, though not to the same span as ARA controls (Supporting Information Figure S1e–g), indicating delay, not blockade.

3.2 |. α_V-containing integrins are expressed during radial sorting in vivo

We next characterized the expression of α_V-containing integrins (α_Vβ₃, α_Vβ₅, α_Vβ₆, α_Vβ₈) in SCs. α_V-containing integrins constitute the majority of RGD-binding integrins (Supporting Information Figure S1a) in addition to α₅β₁ integrin, whose expression in peripheral nerves was characterized previously (Lefcort, Venstrom, McDonald, & Reichardt, 1992), and α₈β₁ integrin, which is not expressed in an RNA database from developing sciatic nerves (D’Antonio et al., 2013). Integrin β₅, β₆, and β₈ are known to dimerize exclusively with the α_V integrin subunit, whereas integrin β₃ can dimerize with the α_V or α_IIb subunits, the latter’s expression being limited to platelets and megakaryocytes (Phillips, Charo, & Scarborough, 1991).

By immunoblot, we detected α_V proteins in developing sciatic nerves (Figure 2a,b). α_V levels were high during radial sorting and the onset of myelination (P3–P5), progressively fell at the peak of myelination (P15) and after myelination (P28), and were relatively low in adults (2 months), which suggested α_V-containing integrins may have a role during radial sorting. We also detected the known partners of the integrin subunit α_V: β₃, β₅, β₆, and β₈. Except for subunit β_5, their levels generally declined throughout development. Because SCs are not the only cell type in the sciatic nerve, we also investigated if α_V and its β partners were expressed in purified SCs and DRG sensory neurons. By immunoblot, we detected integrins α_V, β3, β5, β6, and β8 in purified SCs, but only β5 integrin was also expressed in purified DRG neurons (Figure 2c). Coculturing DRGs and SCs did not increase expression of these integrins, with the possible exception of β8 (Figure 2c,d). These data provide evidence that SCs express RGD-binding integrins that contain subunit α_V and that they are present at relatively high levels during development.

FIGURE 2.

Integrin subunit α_V and its known β partners are expressed in sciatic nerves during early postnatal development. (a) Schematic: Radial sorting in mice begins around birth, is most active at P3–P5, and concludes around P10–P15. Myelination follows radial sorting and completes around P28. Immunoblot of sciatic nerves (s.nerve) from P3 to 2-month old WT mice probed with anti-integrin α_V, −β₃, −β₅, −β₆, and −β₈ Abs. β-tubulin (tub) was used as loading control. (b) Expression of each subunit as ratio of integrin/tubulin. Levels were normalized to P3 from the same immunoblot. See Supporting Information Figure S2 for full blots. (c) Immunoblot of cultured WT SCs (Lane 1), DRGs (Lane 2), and SCs + DRGs cocultured for 1.5, 3, 6, 10, 24, 48, and 72 hr (Lanes 3–9). Immunoblot is representative of three biological experiments. Error bars indicate SEM; n = 3. (d) Expression of each subunit as ratio of integrin/tubulin. Levels were normalized to Lane 1 from the same immunoblot. (e) RGD-binding integrins expressed in sciatic nerve, with references. SC, Schwann cell

3.3 |. SCs deficient in the α_V integrin subunit elongate less in vitro

To confirm that the effect of the antagonist peptide was mediated by α_V integrin heterodimers, we performed siRNA knockdown of the α_V integrin subunit (si-α_V^#2, -α_V^#3, or -α_V^#4) in SCs (Figure 3e), then plated them on WT DRG neurons. After 3.5 hr, α_V-silenced SCs elongated less on axons compared to SCs transfected with scrambled siRNA (negative control) (Figure 3a–d). Average SC length in all silenced cohorts decreased (Figure 3f), because significantly more SCs were stunted, more formed short processes, and fewer elaborated long processes (Figure 3g). α_V-silenced SCs also eventually lengthened in prolonged cocultures (Supporting Information Figure S3a–f), indicating elongation delay. In effect, depleting SCs of all α_V-containing integrins phenocopied isoDGR- and ac-isoDGR-treated SCs, suggesting that α_V heterodimers were the main mediators.

FIGURE 3.

Silencing the integrin subunit α_V in SCs phenocopies the effect of RGD antagonists in SCs. (a–d) SCs were transfected with scrambled siRNA (scram., negative control) or one of three sequences targeting the integrin α_V subunit (si-α_V^#2, -α_V^#3, -α_V^#4), then cocultured with untreated WT DRG neurons for 3.5 hr. Cultures were stained with S100 (green), neurofilament (NF) (red), and DAPI (blue) to label SCs, neurons, and nuclei, respectively. (a′–d′) Magnified area of (a–d). Scale bar is 50 μm in all panels. (e) Immunoblot of untransfected (untrans.), scrambled (scram), and si-α_V-treated SCs to confirm α_V protein knockdown. (f) Average (Avg) SC length in each cohort (student’s t test; scrambled: n = 8 replicates, 3,365 total cells quantified; si-α_V^#2: n = 8 replicates, 3,563 total cells quantified; si-α_V^#3: n = 6 replicates, 2,548 total cells quantified; si-α_V^#4: n = 4 replicates, 1,451 total cells quantified; SEM error bars). Compared to negative control, average length decreased in si-α_V^#2 (−31%), si-α_V^#3 (−44%), and si-α_V^#4 (−25%) cohorts. (g) SC lengths categorized into bins. Significantly more α_V-silenced SCs were stunted (<25 μm), more had short processes (25–50 μm), and fewer lengthened (50–100 μm and 100–150 μm) compared to negative control (ordinary two-way analysis of variance (ANOVA), n = ≥4 coverslips, SEM error bars)

We hypothesized α_Vβ₃ integrin to be the relevant receptor because isoDGR peptides (which have strong affinity for α_Vβ₃) and ac-isoDGR peptides (which do not discriminate between integrins α_Vβ₃, α_Vβ₅, α_Vβ₆, α_Vβ₈, or α₅β₁) elicited comparable effects. However, silencing subunit β₃ did not impact extension of SC on axons (Supporting Information Figure S4a,b), suggesting that β₃-associated integrins like α_Vβ₃ are nonessential for extension along axons. Moreover, α_V protein levels remained stable despite marked reduction of the subunit β₃ protein (Supporting Information Figure S4c–f), suggesting other α_V heterodimers may still be functioning in the absence of α_Vβ₃ integrin. Together, these suggest SCs can utilize multiple α_V-containing integrins for proper lamellipodia extension toward axons.

3.4 |. α_V-containing integrins localize to the tips of SC protrusions

We next determined the localization of α_V integrin when SCs contact axons by immunocytochemistry. We first confirmed the specificity of anti-α_V antibodies by plating scrambled and α_V-silenced SCs on FN-coated coverslips, then immunostained for α_V. After 3.5 hr on FN, α_V-silenced SCs spread and elaborated lamellipodia, with no obvious morphological dissimilarities from controls (Figure 4a,b), demonstrating that α_V integrin is nonessential for forming SC protrusions on FN. The tips of protruding lamellae were strongly immunoreactive in control SCs (Figure 4a); this specific signal and the diffused staining diminished in α_V-silenced SCs (Figure 4b). In 3.5 hr cocultures, α_V was enriched at the terminal ends of the SC that were contacting the axon (Figure 4c,d). Conversely, α_V-silenced SCs (Figure 4e), including the few able to elongate (Figure 4f), were not immunoreactive to α_V. These data illustrate that α_V integrin localizes at axon-associated SC extremities within few hours of coculture.

FIGURE 4.

α_V integrins are enriched at tips of Schwann cell (SC) protrusions. (a,b) Scrambled- and si-α_V^#2-treated SCs were labeled with cell tracker (CT) green, then plated on fibronectin (FN)-coated coverslips for 3.5 hr. Cells were costained with anti-α_V Ab (red), anti-tubulin Ab (Tub, cyan), and DAPI (blue). (a/a′/a′′) α_V integrin was detected at tips of protruding lamellae, particularly at membrane ruffles (arrowheads) and leading edges (arrows). (b/b′/b′′) α_V-silenced SCs lacked α_V integrin immunoreactivity at the tips. Also, the diffused cytoplasmic staining diminished. (c/c′/c′′) When seeded on neurons for 3.5 hr, α_V integrin concentrated at tips of the SC process and at sites of axon contact (arrows). Cells were costained with anti-neurofilament (NF) Ab (cyan) to label neurons. (d/d′) In contrast, α_V integrin was absent along the body of already elongated SCs on axons. (e/e′/e′′,f/f′) α_V-silenced SCs with short (e/e′/e′′) or long (f) processes had no α_V integrin immunoreactivity. All are Z-stack images acquired using identical confocal settings. Scale bar = 30 μm in all panels. Images are representative of at least four replicates. SCs were seeded on neurons at a density of 20,000

3.5 |. SC-specific deletion of subunit α_V does not affect myelination in vivo

To determine if α_V-containing integrins play a role in myelination in vivo, we deleted subunit α_V by crossing homozygous Itgav^f/f floxed mice with hemizygous mice carrying the P₀-Cre transgene. The P₀ promoter drives Cre expression primarily in SCs beginning at E13.5, producing conditional KO (cKO) mice (Itgav^f/f; P₀-Cre) with α_V deleted specifically in SCs (Figure 5a). Unlike embryonic and neonatal-lethal whole body Itgav^−/− mice (McCarty et al., 2002), SC cKOs were viable and survived to late adulthood without obvious behavioral defects. EM cross sections of P3, P5, and P10 sciatic nerves showed no evidence of improper radial sorting or myelination in cKOs (Figure 5b–g). Myelin thickness (G-ratio) (Figure 5j,k,m,n), the density of myelinated fibers (Figure 5l,o), the density of 1:1 axon/SC units (Figure 5p), basal lamina formation (Figure 5h), nodal organization (Figure 5q), and internodal length (Figure 5r) were comparable to controls. Furthermore, myelination in whole body Itgb3^−/− and Itgb8^−/− mice was normal (Figure 5s–u,v–y, respectively). Together, these data show that α_V-containing SC integrins are nonessential for in vivo myelination.

FIGURE 5.

Myelination in vivo proceeds normally in the absence of all α_V-containing integrins in Schwann cells (SCs). (a) The integrin subunit α_V was deleted specifically in SCs using P₀-Cre and Itgav-floxed (Itgav^f/f) mice. Immunoblot of sciatic nerves shows that α_V protein decreased in Itgav^f/f; P₀-Cre (cKO) compared to controls (Itgav^f/f) at P3 (−84% by densitometry analysis) and P32 (−96% by densitometry analysis). Bands below 100 kDa are likely nonspecific. (b–g) No obvious defects in electron micrograph (EM) cross sections of P3, P5, and P10 cKO sciatic nerves. Scale bar for (b–g) is 2 μm. (h–i) Basal lamina was continuous and appeared to form normally around axon bundles (h′,i′, red arrows) and individually myelinated fibers (h′′,i′′, red arrows) in cKO sciatic nerves at P3. Scale bar for (h,i) is 2 μm. (j–p) Quantification of several myelination parameters indicates normal myelination in cKO. Average G-ratio (≥120 fibers quantified per animal) (j,m), G-ratio distribution (k,n), and average number of myelinated fibers per field of view (fov, a 282 μm² area, ≥13 fields quantified per animal) (l,o) were comparable between control and cKO at P3 and P10 (student’s t test; n = 3 animals/genotype). Error bars represent SEM. (p) Average number of axon/SC units engaged in 1:1 relationship was not statistically different between Itgav^f/f and Itgav^f/f; P₀-Cre at P5 (student’s t test, n = 3 animals/genotype, ≥18 fields [282 μm² area] quantified per animal, SEM error bars). (q) To assess nodal organization, teased adult nerve fibers were immunostained for nodal marker PanNaV (red) to label Na⁺ channels and juxtaparanodal marker Kv1.1 (green) to label K⁺ channels. No defects were evident in cKO nerve fibers. Scale bar is 25 μm. (r) Average internodal length did not statistically differ between control and cKO nerves (student’s t test, n = 3 animals/genotype, ≥106 internodes quantified per genotype, SEM error bars). (s,t) Electron micrograph (EM) cross sections of P10 sciatic nerves from Itgb3^+/+ and Itgb3^−-/−- mice show no myelination defects. Scale bar is 2 μm. (u) Average G-ratios in Itgb3^+/+ and Itgb3^−-/−- at P10 are comparable (student’s t test, n = 3 animals/genotype, ≥50 fibers quantified per animal, SEM error bars). β₃ integrin protein absence was confirmed in Supporting Information Figure S4c. (v–y) Semithin cross sections of sciatic nerves from Itgb8^+/+ and Itgb8^−-/−- mice at P32 (v–w) and P120 (x,y) show no obvious morphological defects in integrin β₈ mutants. Scale bar for (v–y) is 12.5 μm

α_V integrin heterodimers bind to the RGD-bearing LAP (latent-associated peptide) fragment of the TGFβ complex to activate the TGFβ pathway (Henderson et al., 2013), and αVβ8 integrin in nonmyelin forming SCs maintain the bone marrow niche by regulating the activation of TGF-β receptors (Yamazaki et al., 2011). Furthermore, since disrupting TGFβ signaling in SCs can lead to reduced proliferation without affecting myelination (D’Antonio et al., 2006), we hypothesized that cell proliferation and TGFβ signaling may be altered in cKO SCs. However, cell proliferation was normal in cKO P3 sciatic nerves (Figure 6a–f), and mutant SCs isolated from cKO nerves properly shuttled SMAD4 into the nucleus upon TGFβ1 stimulation (Figure 6g–r).

FIGURE 6.

Schwann cells (SCs) lacking integrin α_V proliferate and respond to TFGβ1 normally. (a,b,d,e) Longitudinal sections of Itgav^f/f and Itgav^f/f; P₀-Cre sciatic nerves were immunostained for proliferation markers Ki67 (green) at P3 (a,b) and PH3 (green) at P5 (d,e). Both were costained with neurofilament (NF) (red) and DAPI (blue). (c,f) No appreciable difference in the percentage of Ki67-positive cells (student’s t test, n = 4 nerves) or of PH3-positive cells (student’s t test, n = 3 animals/genotype, SEM error bars) was detected between Itgav^f/f and Itgav^f/f; P₀-Cre. (g–r) SMAD4 was properly shuttled to the nucleus in α_V mutant cells. SCs isolated from Itgav^f/f or Itgav^f/f; P₀-Cre sciatic nerves were cultured, treated with TGFβ1, then immunostained for SOX10 (to identify SCs), SMAD4, and DAPI. SMAD4 localized more in the cytoplasm of untreated cells (m–r), but when treated with TFGβ1, SMAD4 localized more in the nucleus of both Itgav^f/f and Itgav^f/f; P₀-Cre SCs (g–l). Scale bar for (g-r) is 30 μm. (s) Average %nuclear-SMAD4/Sox10-positive cells were comparable between Itgav^f/f and Itgav^f/f; P₀-Cre (student’s t test, n = 5 animals/genotype, ≥1,370 Sox10-positive cells quantified per genotype, SEM error bars). Mean %Sox10-positive cells: Itgav^f/f 51%, Itgav^f/f; P₀-Cre 67%

3.6 |. SC-specific double deletion of the integrin subunits α₅ and α_V does not impact myelination in vivo

Compensation or functional overlap by other RGD-binding integrins could explain why α_V cKO had no phenotype. Numerous reports from other cell types demonstrating that α₅ and α_V integrins cooperate, compensate each other’s loss, or functionally overlap (Bharadwaj et al., 2017; Charo, Nannizzi, Smith, & Cheresh, 1990; Turner, Badu-Nkansah, Crowley, van der Flier, & Hynes, 2015; van der Flier et al., 2010; Yang et al., 1999; Yang & Hynes, 1996) led us to hypothesize that integrin α_V and α₅ may be redundant. Deletion of Itga5 specifically in SCs under the control of P₀-Cre, does not perturb myelination (data not shown). We therefore generated Itga5^f/f, Itgav^f/f; P₀-Cre conditional double knockout (cdKO) mice (Figure 7a–d, Supporting Information Figure S5a). As confirmed by immunoblot, integrin α_V proteins diminished in cdKO P10 nerves compared to controls (Figure 7a). We did not detect a decrease in α₅ protein in cdKO nerves by immunoblot (data not shown), despite genotypic confirmation that Itga5 recombined in sciatic nerves (Figure 7b). Unexpectedly, RT-qPCR analysis showed that the relative expression of integrin α₅ in cdKO nerves increased compared to controls (Figure 7c). This increase must be attributable to endoneurial cells other than SC, because we saw a marked reduction in relative expression of integrin α₅ in isolated cdKO SCs by RT-qPCR (Figure 7d).

FIGURE 7.

Myelination in vivo proceeds in the absence of α₅ and α_V integrins in SCs. (a-d) Assessment of α₅ and α_V expression in Itga5^f/f; Itgav^f/f (control) and Itga5^f/f; Itgav^f/f; P₀-Cre (conditional double knockout [cdKO]). (a) Immunoblot confirmed that α_V integrin protein decreased in P10 cdKO sciatic nerves compared to control. GAPDH was used as loading control. (b) Multiplex PCR (using HT030, −31, −32 primers) of genomic DNA from sciatic nerves showed that the Itga5-floxed gene was excised in cdKO nerves. See Supporting Information Figure S5a–c for details. (c) Real-time quantitative PCR (RT-qPCR) of P6 nerves showed that α₅ integrin mRNA increased in cdKO (+58%) compared to control. Results are from one experiment with three independent replicates: control ΔCt 19.3 ± 0.06 SD, cdKO ΔCt 18.6 ± 0.06 SD. (d) RT-qPCR of cultured SCs isolated from control and cdKO nerves showed that α₅ mRNA levels were markedly decreased in cdKO SC (−95%). Results are from one experiment with three independent replicates: control ΔCt 7.6 ± 0.4 SD, cdKO ΔCt 11.9 ± 0.3 SD. Relative expression for (c) and (d) shown as 2^−(ΔΔCt) value with SD errors bars. (e–j) Electron micrograph (EM) cross sections of P3 (e,h), P10 (f,i), and P28 (g,j) sciatic nerves from control and cdKO mice showed no obvious difference. Scale bar for (e–i, g–j) is 2 μm. (f′,i′) Magnified region of (f,i) shows that myelinated fibers (arrows) and axon bundles (arrowheads) at P10 in control and cdKO are surrounded by a continuous basal lamina. (k,m) Average G-ratios at P3 and P10 were similar between control and cdKO (student’s t test, n = ≥3 animals/genotype, ≥50 fibers quantified per animal, SEM error bars). (l,n) G-ratio distribution for P3 and P10. Difference in G-ratio distribution between control and cdKO at P10 was statistically significant (linear regression analysis of slopes, **p = .003 for P10, ns for P3). (o,p) Teased adult nerve fibers stained for PanNaV (red) and Kv1.1 (green) showed normal nodal organization in cdKO. Scale bar is 25 μm. (q) Average internodal length was similar between control and cdKO (student’s t test, n = 3 animals/genotype, ≥102 internodes quantified per genotype, SEM error bars)

cdKO mice survived to late adulthood and behaved normally like controls. EM cross sections of cdKO nerves did not display obvious defects in radial sorting or myelination at P3, P10, or P28 (Figure 7e–j). Average G-ratios (Figure 7k,m), nodal organization (Figure 7o,p), and internodal length (Figure 7q) at P3 and P10 were similar to controls. However, the G-ratio distribution (as a function of axon diameter) in P10 cdKO had a slight but significant slope downshift relative to control (Figure 7l,n), suggesting that large fibers may have thicker myelin in cdKO nerves. Nevertheless, these data overall show that α₅- and α_V-containing integrins are nonessential for in vivo myelination.

3.7 |. Remyelination after nerve crush proceeds in the absence of the integrin subunits α₅ and α_V

Several papers have proposed that RGD-based axo-glial interactions may be important for the remyelination process after nerve injury (Akassoglou, Akpinar, Murray, & Strickland, 2003; Akassoglou, Yu, Akpinar, & Strickland, 2002; Chernousov & Carey, 2003; Liu, Martinez, Durand, Wildering, & Zochodne, 2009; Qian et al., 2018; Rafiuddin Ahmed & Jayakumar, 2003; Zheng et al., 2015). To determine if α₅ or α_V integrins play a role in nerve degeneration, regeneration, or remyelination, we performed nerve crush on α_V integrin cKO and α₅/α_V integrin cdKO adult sciatic nerves, then examined axon regeneration and remyelination 15 and 30 days postcrush (T15 and T30, respectively). Injured nerves sampled from α_V integrin cKO (Figure 8a,c) or α₅/α_V integrin cdKO mice (Figure 8b,d) were morphologically similar to their respective controls, showing a similar number of axons and myelinated fibers and similar myelin thickness at both times, suggesting that α₅- and α_V-containing SC integrins are nonessential for nerve regeneration after crush.

FIGURE 8.

Schwann cells (SCs) lacking integrins α₅ and α_V in vivo can remyelinate and regenerate the nerve after injury. (a) Adult Itgav^f/f and Itgav^f/f; P₀-Cre sciatic nerves were crushed, then examined at 15 (T15) and 30 (T30) days postcrush. Shown are semithin cross sections of portion of nerve distal to the crush site. Remyelination and axon regeneration in α_V cKO appeared similar to controls. Scale bars are 12.5 μm. (b) Nerve crush was repeated in Itga5^f/f; Itgav^f/f and Itga5^f/f; Itgav^f/f; P₀-Cre adults. No morphological difference was detected between controls and α₅/α_V conditional double knockout (cdKO). Scale bars are 12.5 μm. (c,d) Average number of myelinated fibers per 140 cm² area (field of view [fov]) in α_V(c) and α₅/α_V (d) at T30. Differences were not statistically significant in both cohorts (student’s t test, n = 3 animals/genotype, ≥5 fields quantified per animal)

4 |. DISCUSSION

During development, SCs extend their processes to contact, sort, and myelinate axons. In vitro, we found that α_V integrins, a group that binds RGD-bearing ligands, localize in SC to specific sites of contact with axons and contribute to the extension of the process.

The integrin α_V subunit is expressed by western blot in early stages of nerve development when SCs are actively engaging axons to sort and myelinate them. We were unable to show the in vivo expression of α_V integrins by immunostaining because we could not validate the specificity of the anti-α_V antibody on tissue sections or teased fibers (data not shown). This is surprising, given that we could detect specific staining of α_V integrin in culture, and could be due to upregulation of α_V integrin on the more rigid and stiff environment of cultures. However, we believe that α_V integrin expression detected by western blot can be mainly attributable to SCs, because SC-specific ablation in vivo results in major reduction of the protein in whole nerve immunoblots, and we do not detect α_V integrin protein in cultured purified DRG neurons or axons. Our results are consistent with previous reports showing cultured SCs express the integrin subunits α_V, β₃, β₅, β₆, and β₈ (Chabas et al., 2013; Milner, Wilby, et al., 1997; Poitelon et al., 2015). Together with subunit β₁, a key SC integrin (Feltri et al., 2002), SCs can potentially form all five α_V integrin heterodimers. Of them, integrin α_Vβ₆ may be the least likely to participate in the myelination process because of its low expression in vivo. However, it might participate in injury or disease, as SC β₆ localizes to the injury zone of transected nerves (Liu et al., 2009) and is differentially upregulated in severely amyelinated nerves (Poitelon et al., 2016). Integrins α_Vβ₃ and α_Vβ₈, on the other hand, could have distinct roles in development. We found that expression of the integrin subunits β₃ and β₈ are dynamic and developmentally regulated. It echoes the dynamic expression of α_V-associated β subunits in oligodendrocytes, which is postulated to reflect an integrin switching (α_Vβ₁-α_Vβ₃-α_Vβ₈-α_Vβ₅) in oligodendrocytes during development (Blaschuk et al., 2000; Milner et al., 1996; Milner & Ffrench-Constant,1994; Milner, Frost, et al., 1997).

Blocking or silencing integrin α_V in vitro impairs process extension, and as a result, many SCs assume a rounded, stunted morphology. Diminishing only integrin α_Vβ₃ has no effect, but antagonistic inhibition of integrin α_Vβ₃, α_Vβ₅, α_Vβ₆, α_Vβ₈, and α₅β₁ by ac-isoDGR and knockdown of all possible α_V heterodimers do reveal a loss-of-function in vitro. It is possible that RGD-binding integrins not directly tested by our analyses participate in this process. Other β₁-containing integrins do not bind RGD primarily, but are capable of binding RGD or an RGD-related motif (Bazigou et al., 2009; Ruoslahti, 1996). Also, deletion of all β₁ integrins in SCs cause severe radial sorting defects (Feltri et al., 2002).

Surprisingly, conditional mutant mice lacking integrins α_V are normal, and α₅ integrin does not compensate for α_V integrin deficiency, despite the many papers postulating a possible role for these integrins in SC-axon interactions during myelination and remyelination (Akassoglou et al., 2002; Akassoglou et al., 2003; Chernousov & Carey, 2003; Liu et al., 2009; Qian et al., 2018; Rafiuddin Ahmed & Jayakumar, 2003; Zheng et al., 2015). We show that α_V and α₅ integrins are not required for radial sorting, myelination, and remyelination after injury in vivo. In our case, the coculture system may have been more permissive to reveal a phenotype because, unlike in vivo, seeded SCs are synchronized and associated with axons simultaneously. With a time-sensitive in vitro phenotype (delay), this becomes a major limitation in vivo because SCs in the nerve are asynchronized and associate with axons at different times. Similar discrepancies between the presence of phenotype in DRG coculture systems, but absence of phenotype in vivo has been observed previously in various reports (Eshed et al., 2005; Li et al., 1994; Owens, Boyd, Bunge, & Salzer, 1990; Owens & Bunge, 1989, 1990, 1991), and may be due to the lack of compensatory mechanisms in vitro, to the more simplified cellular system, or to the different 3D arrangements of DRGs versus peripheral nerves.

The possibility of compensation by other receptors in vivo likely explains the lack of phenotype. Many more cellular molecules are present in the nerve compared to our coculture environment, and any receptor capable of binding RGD motifs could potentially contribute; we propose β₁-containing integrins such as α₈β₁ (which binds to RGD motif) or α₄β₁ (which binds to RGD-related motif) (Humphries, 2006) as potential candidates. In endothelial cells lacking both α_V and α₅, α₄β₁ relocalizes to focal adhesions formerly occupied by α_V and α₅ (van der Flier et al., 2010). Assessing any integrin expression changes between control, α_V cKO, and α₅/α_V cdKO transcriptomes could help identify contributing integrins, and the coculture system would be a good and practical first step in which to test these all these potential integrins using peptidomimetics, next generation small molecule inhibitors (Reed et al., 2015), or gene knockdowns of central integrins. Nonintegrin molecules may also compensate, especially in light of integrins’ ability to cross talk and cosignal with growth factor receptors and several molecules along signal transduction pathways (Schwartz & Ginsberg, 2002). We show that RGD-binding integrins are involved in axo-glial interactions in vitro; identifying the potential axonal ligands of these integrins could also lead to the identification of the compensating molecules in question.

Perturbation of the same integrins between different systems and cell types have often resulted in discrepancies (Hynes, 2002a) perhaps because integrins are so versatile in their interactions, that their functions are context-specific and therefore somewhat unpredictable. in vitro testing coupled with in vivo validations should help define their context-specific roles. Additionally, recent papers show that a precise mechanism elicited by nonsense mutations in vivo triggers a complex compensatory response that upregulates many genes with sequence similarity to the deleted one (El-Brolosy et al., 2019; Ma et al., 2019). This mechanism masks defects in vivo. Many KO or Cremediated mutants are designed to generate frame-shift and stop codon which commonly elicit a nonsense-mediated decay response, and even if we have not formally documented it in our mutants, the design of the α_V and α₅ integrin animals both results in premature stop codons, and thus a nonsense mutation, indicating that this mechanism may be in place.

Interestingly, we recently found that α_V integrin, along with important integrin and promyelinating genes, is significantly downregulated in mice completely arrested in radial sorting (Poitelon et al., 2016). This could indicate α_V integrin is indeed a participant in radial sorting in vivo. These mice lack key mechanotransduction regulators YAP and TAZ in SCs, and it has been shown in other cell types that Itgav is a target of YAP (Nardone et al., 2017). Intriguingly, MDCK epithelial cells lacking α_V integrin assume a “roundish” morphology and fail to respond to substrate rigidity, which is indicative of impaired mechanotransduction (Teräväinen et al., 2013). It is thus conceivable that SC mechanosensing involves α_V integrins. The dissimilarity of the nerve and coculture mechanical environments may have also contributed to our in vitro and in vivo discrepancies in detecting α_V integrin expression and function. The ECM and the other cells surrounding SCs in vivo could impose a differential mechanical stimuli on SCs compared to the neurons and substrate in the coculture (Belin, Zuloaga, & Poitelon, 2017).

Despite possible overlap or compensation, we define the precise step in which SC α_V integrin exerts its function, at least in vitro. For one, α_V integrin is not required for axon recognition or adhesion, given that impaired SCs can still anchor themselves onto axons (not substrate) and that a comparable number of impaired and control SCs do adhere. Instead, α_V integrin seems to participate in the early phases of process extension, as many α_V-silenced or antagonist-treated SCs in short-term cocultures are stunted and many do not reach the lengths of their control counterparts. This aligns with Milner, Wilby, et al.’s (1997) findings, who demonstrated that SC migration is hindered by RGD peptides. Since process extension is an inaugural step of cell migration, it is possible that this precise step is perturbed.

Glial α_V integrin has mostly been studied within the context of ECM interactions, but here we show that SC α_V integrin may possibly function independently of ECM ligands and that its role could in part be axon dependent. SCs plated on FN appear unaffected by depletion of all α_V heterodimers, yet axon-related SCs become disturbed, suggesting axons have a unique impact on SC α_V integrin. We propose that SC α_V interacts with axonal molecules and functions ad-axonally. There are a few other examples of α_V integrins mediating cell–cell interaction. Interestingly, α_V integrins have been shown to bind directly to L1 (Montgomery et al., 1996), Neuregulins (Ieguchi et al., 2010), and ADAM molecules (Nath et al., 1999; Zhou, Graham, Russell, & Croucher, 2001). Nrg1-Type III and ADAM22 are essential for myelination (Ozkaynak et al., 2010; Taveggia et al., 2005) and α_V integrins could modulate the interaction between these molecules and their obligate receptors, Erb2/3 and Lgi4 (Bermingham et al., 2005), respectively.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.