A dual role for Integrin α6β4 in modulating Hereditary Neuropathy with liability to Pressure Palsies

Abstract

Peripheral myelin protein 22 (PMP22) is a component of compact myelin in the peripheral nervous system. The amount of PMP22 in myelin is tightly regulated, and PMP22 over or under-expression cause Charcot-Marie-Tooth 1A (CMT1A) and Hereditary Neuropathy with Pressure Palsies (HNPP). Despite the importance of PMP22, its function remains largely unknown. It was reported that PMP22 interacts with the β4 subunit of the laminin receptor α6β4 integrin, suggesting that α6β4 integrin and laminins may contribute to the pathogenesis of CMT1A or HNPP. Here we asked if the lack of α6β4 integrin in Schwann cells influences myelin stability in the HNPP mouse model. Our data indicate that PMP22 and β4 integrin may not interact directly in myelinating Schwann cells, however, ablating β4 integrin delays the formation of tomacula, a characteristic feature of HNPP. In contrast, ablation of integrin β4 worsens nerve conduction velocities and non-compact myelin organization in HNPP animals. This study demonstrates that indirect interactions between an extracellular matrix receptor and a myelin protein influence the stability and function of myelinated fibers.

In This Issue Caption

Laminin receptor integrin α6β4 modulates peripheral myelin protein 22 (PMP22) neuropathy. In the current study, we employed biochemical and genetic studies to shed light on the relationship between PMP22 and signalling from the Schwann cell extracellular matrix. We demonstrate that ablation of β4 integrin delays the progression of Hereditary Neuropathy with Pressure Palsy (HNPP), but also causes a reduction in the speed of action potential propagation in HNPP animals. These findings suggest that in HNPP, integrin α6β4 has opposing effects on two distinct pathomechanisms, i.e.: the formation of tomacula and the propagation of action potentials.

Introduction

Peripheral myelin protein 22 (PMP22) is a constituent of peripheral myelin (Welcher et al. 1991). Defects in PMP22 gene dosage cause the most common inherited diseases of peripheral nerves, which includes Hereditary Neuropathy with liability to Pressure Palsies (HNPP), due to deletion of a genetic region containing the PMP22 allele (Chance et al. 1993). Mice haploinsufficient for PMP22 recapitulate human HNPP pathology, that includes the formation of tomacula (smooth thickening of myelin typically at paranodes), progressive demyelination and failure of action potential propagation (Adlkofer et al. 1995, Adlkofer et al. 1997, Bai et al. 2010). Although PMP22 was identified 25 years ago, the function of PMP22 in myelin remains mysterious (Snipes et al. 1992). PMP22 may participate in actin-mediated cellular functions and in establishing lipid rafts (Lee et al. 2014). Other hypotheses regarding its function include compact myelin stabilization (D’Urso et al. 1999, Hasse et al. 2004), the promotion of Schwann cell differentiation, proliferation or apoptosis (Fabbretti et al. 1995, Amici et al. 2007, Sancho et al. 2001), calcium homeostasis (Nobbio et al. 2009) and maintenance of myelin permeability (Guo et al. 2014, Hu et al. 2016). β4 integrin interacts with the α6 integrin subunit to form a laminin receptor that provides stability and mechanical support to myelinated fibers (Feltri et al. 1994, Nodari et al. 2008, Quattrini et al. 1996),

During early stages of myelination, PMP22 and α6β4 integrin are co-expressed in Schwann cells and can be co-immunoprecipitated (Amici et al. 2006), but while PMP22 is present in compact myelin, α6β4 integrin is on the outer layer of the Schwann cell membrane. Therefore, it is uncertain how and where the interaction between PMP22 and α6β4 integrin may occur in nerves. Finally, deletion of Pmp22 or β4 integrin causes similar pathological features, but with drastically different age of onset, as Pmp22 haploinsufficient animals develop tomacula between 10 and 20 days of age (Adlkofer et al. 1995, Adlkofer et al. 1997) and mice with β4 integrin deletion in Schwann cells develop myelin folding and demyelination at 12 months (Nodari et al. 2008, Quattrini et al. 1996).

To elucidate the significance of the interaction between PMP22 and α6β4 integrin, we performed careful co-localization and biochemical analysis and identified novel interactors of the β4 integrin subunit. To ask if the lack of α6β4 integrin in Schwann cells influences myelin stability in the HNPP model, we generated mice lacking both PMP22 and β4 integrin in Schwann cells. We find that ablation of β4 integrin in HNPP does not modify myelin thickness or internodal length and does not worsen abnormal myelin folding, but worsens nerve conduction velocities in HNPP animals, and this correlates with worsening of the paranodal and internodal disorganization. These data indicate that integrin α6β4 and PMP22 cooperate to organize the internode and to allow proper propagation of the axonal signal along the axons.

Methods

Animal models and genotyping

All animal procedures were performed in accordance with San Raffaele Institute and University at Buffalo animal care committee regulations (n. 363 and UB1188M respectively). Itgb4 floxed mice (RRID:MGI:3804178), Pmp22 knock-out mice (RRID:MGI:3794448) and P0-Cre transgenic mice (RRID:IMSR_JAX:017927) have been described previously (Feltri et al. 1999, Adlkofer et al. 1995, Nodari et al. 2008). Animal’s references can be retrieved with RRIDs at http://scicrunch.org/resources. Animals used as controls are Pmp22 +/+; Itgb4 fl/fl; P0-cre negative (Pmp22 +/+; Itgb4 SC +/+). Pmp22 +/− animals are Pmp22 +/−; Itgb4 fl/fl; P0-Cre negative (Pmp22 +/−; Itgb4 SC +/+). Itgb4 SC −/− animals are Itgb4 fl/fl; P0-Cre positive (Pmp22 +/+; Itgb4 SC −/−). Finally, Pmp22 +/−; Itgb4 SC −/− animals are Pmp22 +/−; Itgb4 fl/fl; P0-Cre positive (Pmp22 +/−; Itgb4 SC −/−). All mice were backcrossed to C57BL/6 to reach congenic background. Both males and females were included in the study. Mutant and control littermates were sacrificed at the indicated ages, and sciatic nerves were dissected. No animals were excluded from the study. Animals were housed in cages of no more than 5 animals in 12 h light/dark cycles. Most experiments were conducted with 3 animals per age and per genotype. A flow-chart of the experimental procedures is available in Fig. S5. Genotyping of mutant mice was performed by PCR on tail genomic DNA, as described previously (Feltri et al. 1999, Adlkofer et al. 1995, Nodari et al. 2008).

Electrophysiology

Animals were analysed at 30 days of age as described previously (Poitelon et al. 2015). Briefly, mice were anesthetized with tribromoethanol, 0.02 ml g⁻¹ of body weight, and placed under a heating lamp to avoid hypothermia. Sciatic nerve motor conduction velocity and amplitude were obtained with subdermal steel monopolar needle electrodes: a pair of stimulating electrodes was inserted subcutaneously near the nerve at the ankle, then at the sciatic notch, and finally at the paraspinal region at the level of the iliac crest to obtain three distinct sites of stimulation, proximal and distal, along the nerve. Compound motor action potential were recorded with an active electrode inserted in muscles in the middle of the paw and a reference needle in the skin between the first and second digits.

Morphological assessments

Mutant and control littermates were euthanized at the indicated ages, and sciatic nerves were dissected. Tissues were fixed in 2% buffered glutaraldehyde and post fixed in 1% osmium tetroxide. After alcohol dehydration, the samples were embedded in Epon. Transverse sections (0.5 – 1 nm thick) were stained with toluidine blue and examined by light microscopy. Morphological measurements were quantified from anatomically comparable whole cross-section of sciatic nerve cut halfway between the sciatic notch and the knee. For G-ratio analysis of sciatic nerves (axon diameter/fiber diameter), 4 images per semithin section were acquired at the 100x objective. Axon and fiber diameters were quantified using the Leica QWin software (Leica Microsystem). Myelinated fibers and tomacula were quantified on the whole cross-section of sciatic nerve using ImageJ (imagej.nih.gov/ij). Data were analysed using GraphPad Prism 6.01. Images were quantified blindly.

Teasing and osmication of nerve fibers

Sciatic nerves were fixed in 2% glutaraldehyde overnight at 4 °C and washed 3 times in phosphate buffer (79 mM Na₂HPO₄, 21 mM NaH₂PO_4, pH 7.4). Sciatic nerves were stained in 1 % osmium, washed 4 times in phosphate buffer and incubated at 55 °C during 12 hours in 30% glycerol followed by 12 hours in 60% glycerol and 12 hours in 100% glycerol.

Immunoelectron microscopy

Immuno-electron microscopy analyses were performed as in (Quattrini et al. 1996) but with embedding modified as described in (Yin et al. 2000). Primary antibody to PMP22 (Abcam, ab61220, RRID:AB_944897), and 5 nM of gold conjugated secondary antibody were used. Immuno-electron microscopy analysis was performed once.

Cell culture

Rat 804G (RRID:CVCL_J122) and hamster CHO (RRID:CVCL_0213) cells were maintained in DMEM (Gibco), 10% FCS, penicillin and streptomycin. 804G permanently transfected with β4 integrin mutants were a gift from Dr. Giancotti (Spinardi et al. 1993). Transfection of HA-Pmp22 and β4 integrin were performed using Lipofectamine 2000 (Invitrogen), according to the manufacturer’s instructions. Cells were analyzed 72h after transfection to insure efficient expression of the transfected plasmids. Each transfection was repeated at least three times.

Western blot

Sciatic nerves were dissected, removed from their epineurium, frozen in liquid nitrogen then pulverized. Transfected hamster CHO cells (RRID:CVCL_0213) and sciatic nerves were resuspended in lysis buffer (95 mM NaCl, 25 mM Tris-HCl pH 7.4, 10 mM EDTA, 2% SDS, 1 mM Na₃VO₄, 1 mM NaF and 1:100 Protease Inhibitor Cocktail, Roche). Protein lysates were incubated at 4 °C for 30 min then centrifuged at 16,000 rpm for 30 min at 4°C. The concentration of the protein supernatants was determined by BCA protein assay (Thermo Scientific) according to manufacturer’s instructions. Equal amounts of homogenates were diluted 3:1 in 4 X Laemmli (250 mM Tris-HCl pH 6.8, 8% SDS, 8% β-Mercaptoethanol, 40% Glycerol, 0.02% Bromophenol Blue), denatured 5 min at 100 °C, resolved on SDS-polyacrylamide gel, and electroblotted onto PVDF membrane. Blots were then blocked with 5% BSA in 1 X PBS, 0.05% Tween-20 and incubated over night with the appropriate antibody. Abcam anti-PMP22 (ab61220, RRID:AB_944897), Cell signaling anti-PAK1 (2602, RRID:AB_330222), GenWay anti-p-PAK1 (GWB-961E2C, RRID:AB_10275529), Millipore anti-integrin β1 (MAB1997, RRID:AB_2128202), anti-Itpr3 (AB9076, RRID:AB_571029), Novocastra anti-β-dystroglycan (NCL-b-DG, RRID:AB_442043), Roche anti-HA (11867423001, RRID:AB_390918), Santacruz anti-integrin β4 human (sc-9090, RRID:AB_2129021), anti-integrin α6 (sc-6597, RRID:AB_2128041) and Sigma anti-β-tubulin (T4026, RRID:AB_477577), anti-calnexin (C4731, RRID:AB_476845). Antibodies references can be retrieved with RRIDs at http://scicrunch.org/resources. Anti-integrin β4 mouse/rat (lot 2213) was gift of Dr. Brophy, Centre for Neuroregeneration, Univ. of Edinburgh). Membranes were rinsed in 1 X PBS and incubated for 1h with secondary antibodies. Blots were developed using ECL, ECL plus (GE Healthcare) or Odyssey CLx infrared imaging system (Li-Cor). Western blots were quantified using Image J software (http://imagej.nih.gov/ij). Each Western blot was repeated at least three times.

Immunoprecipitation

Rat 804G cells and sciatic nerves fibers were resuspended in lysis buffer (150 mM NaCl, 50 mM Tris-HCl pH 7.4, 0.5% Sodium deoxycholate, 1% NP-40, 1:100 Protease Inhibitor Cocktail, Roche). For each immunoprecipitation, 500 μg of proteins were incubated with 50 μl of Protein G Sepharose for 1 hour at 4 °C on a rotating wheel. After a short centrifugation, supernatants were transferred to a new vial, incubated overnight at 4 °C with 1 μg of antibody (Abcam anti-Integrin β4 (ab25254, RRID:AB_2129042) or Roche anti-HA (11867423001, RRID:AB_390918)) on a rotating wheel. 50 μl of Protein G Sepharose were added to the mix and incubated for an additional 3 hours. After centrifugation, supernatants were collected as unbound fraction (Ub), pellets were washed three time in lysis buffer, resuspended in Laemmli and processed as Western blot as the immunoprecipitated fraction (IP). Equal amounts of proteins were incubated with the target antibody, precipitated, separated by SDS/PAGE and probed. Each immunoprecipitation was repeated at least three times.

Immunohistochemistry

We tested various fixation and anti-PMP22 antibodies against nerves from Pmp22 −/− animals at various ages. Surprisingly, numerous antibodies anti-PMP22 (Abcam ab126769, RRID:AB_11129961; Assay Biotech C0306, RRID:AB_10686127; LifeSpan, LS-C118559-100, RRID:AB_10796545; LS-C122324-100, RRID:AB_10799898; Santacruz sc-58572, RRID:AB_785236; sc-18535, RRID:AB_2167000; sc-71911, RRID:AB_2167001; sc-65739, RRID:AB_2167002; Sigma-Aldrich SAB4502217, RRID:AB_10746275) gave positive stainings in all conditions tested even in null tissues, and were not further pursued. Only one anti-PMP22 (ab61220, RRID:AB_944897) showed no PMP22 signal in SCs of PMP22 −/− animals (Fig. S4). The antibodies references can be retrieved with RRIDs at http://scicrunch.org/resources. The following conditions gave no background staining for PMP22 in Pmp22 −/− null animals (data not shown). Unfixed sciatic nerve sections and sciatic nerve teased fibers were fixed with cold methanol for 1 min or cold acetone for 10 min, washed in 1 X PBS, blocked for 1 h in blocking solution (5% Fish skin gelatin, 0.5% Triton X-100, 1 X PBS), then incubated overnight with the following antibodies: Abcam anti-Integrin β4 (ab25254, RRID:AB_2129042), anti-PMP22 (ab61220, RRID:AB_944897), Alomone labs anti-Kv1.1 (APC-009, RRID:AB_2040144), Covance anti-Neurofilament M (PCK-593P, RRID:AB_663194), Millipore anti-Integrin β1 (MAB1997, RRID:AB_2128202), Invitrogen Rhodamine Phalloidin (R415, RRID:AB_2572408), Santacruz anti-Integrin α6 (sc-6597, RRID:AB_2128041), Sigma anti-Itpr3 (AB9076, RRID:AB_571029), anti-Claudin-19 was a gift by Dr. Shoichiro Tsukita and Dr. Mikio Furuse. Slides were rinsed in PBS, incubated 1 h with Jackson DyLight 488 or 549-conjugated secondary antibodies, stained with DAPI, and mounted with Vectashield (Vector Laboratories). Images were acquired with an upright microscope Leica DM5000, a Zeiss Apotome or a confocal microscope Leica SP5II. Each immunochemistry was repeated at least three times.

Proximity Ligation Assay

Unfixed sciatic nerve sections were fixed in cold methanol for 1 min, washed in in 1 X PBS then blocked for 1 h in blocking solution (5% Fish skin gelatin, 0.5% Triton X-100, 1 X PBS). Slides were then incubated overnight with the following antibodies: Abcam anti-Integrin β4 (ab25254, RRID:AB_2129042) and anti-PMP22 (ab61220, RRID:AB_944897) or Millipore anti-Neurofilament M (MAB-10651, RRID:AB_2150076) and anti-Neurofilament H (AB1989, RRID:AB_91202). MINUS probe was coupled to anti-Integrin β4 using In Situ Probemaker (Sigma). The following procedures were carried out using the Duolink starter kit (Sigma) products. PLA probes PLUS (rabbit) and MINUS (mouse) were diluted 1:5 in blocking solution and incubated 1 h at 37 °C. Slides were washed twice in Wash Buffer A for 2 min. Ligase and ligation buffer were diluted 1:40 and 1:5 were diluted in 1 X PBS, and incubated on slides at 37°C for 30 min. Slides were washed twice in Wash Buffer A for 2 min. Polymerase and amplification buffer were diluted 1:80 and 1:5 in 1 X PBS, and incubated on slides at 37°C for 120 min. Slides were washed three time in Wash Buffer B for 10 min. Cells were stained with DAPI, and mounted with media supplied by the kit. Images were acquired with an upright microscope Leica DM5000. The proximity ligation assay was performed twice.

Quantitative real-time PCR

For quantitative real-time PCR (qPCR), we sampled sciatic nerves from control and mutant mice at P5, P10, P30 and P90. Sciatic nerves were frozen in liquid nitrogen, crushed with a metallic pestle and total RNAs were isolated from sciatic nerve using TRIzol (Life Technologies). First strand cDNA was prepared from 1 μg or RNA using Superscript II, 50 μM oligo(dT)₂₀ and 50 ng random hexamers, according to the manufacturer’s instructions. qPCR was performed using Sybr Green (Roche) in a LightCycler 480 System (Roche) according to standard protocols. Normalization was performed using 18S rRNA as a reference gene. The relative standard curve method was applied using wild-type mice as reference. At least 3 animals were used for each genotype at each age. Primers were designed with the Roche Primer Design Center (http://qpcr.probefinder.com/organism.jsp) and their efficiency was assessed by standard curve. Only primers 90% efficient or above were selected. Primer dimerization was tested in vitro by Amplify X (http://engels.genetics.wisc.edu/amplify) and the homogeneity of the PCR products was assessed by a dissociation curve. We used the following primers to amplify 18S: 5′ ctcaacacgggaaacctcac 3′ and 3′ cgctccaccaactaagaacg 5′, Egr2: 5′ ctacccggtggaagacctc 3′ and 3′ aatgttgatcatgccatctcc 5′ Itga6: 5′ cctgaaagaaaataccagactctca 3′ and 3′ ggaacgaagaacgagagagg 5′, Itga7: 5′ agaaggtggagcctagcaca 3′ and 3′ gctgaacaccacacacttgg 5′, Itgb1: 5′ caaccacaacagctgcttctaa 3′ and 3′ tcagccctcttgaattttaatgt 5′, Itgb4: 5′ cttggtcgccgtctggta 3′ and 3′ tcgaaggacactaccccact 5′, Mpz: 5′ gctgccctgctcttctctt 3′ and 3′ tttccctgtccgtgtaaacc 5′, Pmp22: 5′ ccgtccaacactgctactcc 3′ and 3′ cgctgaagatgacagacagg 5′. Each quantitative real-time PCR was repeated twice.

Mass spectrometry and data analysis

Immunoprecipitated eluates were separated by 4–12% SDS–PAGE, stained with Coomassie Brilliant Blue (Bio-Rad) and excised in eight slices for LC-MS/MS analysis. Mass spectrometry analysis was performed by LC-MS/MS using an LTQ-Orbitrap mass spectrometer (ThermoScientific, Bremen, Germany). Tryptic digests for each band were first cleaned using Stage Tips as described previously (Rappsilber et al. 2007) and then injected in a capillary chromatographic system (EasyLC, Proxeon Biosystems, Odense, Denmark). Peptide separations occurred on a homemade column obtained with a 10-cm fused silica capillary (75 μm inner diameter and 360 μm outer diameter; Proxeon Biosystems) filled with Reprosil-Pur C18 3 μm resin (Dr Maisch GmbH, Ammerbuch-Entringen, Germany) using a pressurized ‘packing bomb’. A gradient of eluents A [distilled water with 2 % (v v⁻¹) acetonitrile, 0.1% (v v⁻¹) formic acid] and B [acetonitrile, 2% (v v⁻¹) distilled water with 0.1% (v v⁻¹) formic acid] was used to achieve separation from 8% B (at 0 min, 0.2 ml/min flow rate) to 50% B (at 80 min, 0.2 ml min⁻¹ flow rate). The LC system was connected to the orbitrap equipped with a nanoelectrospray ion source (Proxeon Biosystems). Full-scan mass spectra were acquired in the LTQ-Orbitrap mass spectrometer in the mass range m z⁻¹ 350–1500 Da and with the resolution set to 60000. The ‘lock-mass’ option was used for accurate mass measurements. The 10 most intense doubly and triply charged ions were automatically selected and fragmented in the ion trap. Target ions already selected for the MS/MS were dynamically excluded for 60 s (Olsen et al. 2005). Protein identification and quantification were achieved using the MaxQuant software version 1.3.0.5 (Cox & Mann 2008). Cysteine carbamidomethylation was searched as a fixed modification, whereas N-acetyl protein and oxidized methionine were searched as variable modifications. Mass spectra were analyzed by Andromeda plugin in MaxQuant using UniProt complete proteome Mus musculus 2013 database. Protein quantification was based on LFQ intensities. Peptides and proteins were accepted with a false-discovery rate of 0.01, two minimum peptides identified per protein of which one unique. The experiments were done in biological duplicate performing two technical replicates.

Gene ontology analysis

Gene ontology (GO) clustering analysis was performed using the Cytoscape (http://chianti.ucsd.edu/cyto_web/plugins/index.php) plugin Biological Network Gene Ontology (BiNGO) (http://www.psb.ugent.be/cbd/papers/BiNGO/index.html). The degree of functional enrichment for a given cluster and category was assessed quantitatively (P value) by hypergeometric distribution, a multiple test correction was applied using the false discovery rate (FDR) algorithm, fully implemented in BiNGO software. Overrepresented biological process categories were generated after FDR correction, with a significance level of 0.05. The full analysis is listed in Data S3.

Statistical analyses

Experiments were not randomized, but data collection and analysis were performed blindly to the conditions of the experiments. Researchers blinded to the genotype performed morphological analyses, nerve conduction velocities and morphometric analyses. The data obtained are presented as mean ± s.e.m. No statistical methods were used to predetermine sample sizes, but our sample sizes are similar those generally employed in the field. t-test and One-way ANOVA with Bonferroni’s multiple comparisons test were used for statistical analysis of the differences among multiples groups according to the number of samples. Values of P ≤ 0.05 were considered to represent a significant difference. This study was not preregistered.

Results

Ablation of integrin α6β4 has both beneficial and detrimental effects on the HNPP mouse model

To test for functional interactions between α6β4 integrin and PMP22, we asked whether α6β4 integrin in Schwann cells modifies the timing or severity of myelin abnormalities associated with HNPP, exploiting an authentic model of Pmp22 loss-of-function (Adlkofer et al. 1995, Bai et al. 2010) and the conditional, Schwann cell specific, knock-out for integrin β4 (Nodari et al. 2008, Feltri et al. 1999). We generated double mutant (Pmp22 +/− and Itgb4 SC −/−) animals and analyzed myelination in the peripheral nervous system.

Analysis of Pmp22 +/− nerves by semithin sections from 20 to 90 days of age confirmed previous reports that haploinsufficiency of Pmp22 does not cause axonal loss or early demyelination (Fig. 1a–d), but promotes the formation of tomacula (Fig. 1a arrows, Fig. 1g open arrows) and a reduction in the length of myelin internodes (Fig. 1g–i) (Adlkofer et al. 1995, Amici et al. 2007, Adlkofer et al. 1997, Hu et al. 2016). Surprisingly, we observed that ablation of integrin α6β4 slightly delays the appearance of tomacula (Fig.1e, f). A significant reduction in the number of tomacula can be observed at 20 days of age (P20) while myelination is ongoing, but disappears at P200 once myelin is mature (Fig. 1e). We also noticed a moderate increase in the length of myelin segments (internodes) in double mutants as compared to Pmp22 +/− animals at P30 (Fig. 1g–i). The increase is significant for internodes of small caliber fibers (Fig. 1i). Together these data indicate that ablation of integrin α6β4 delays the onset of morphological defects caused by Pmp22 haploinsufficiency.

Figure 1. Ablation of β4 integrin in Schwann cells delays the morphological defects of HNPP.

Schwann cell-specific deletion of β4 integrin delays the formation of tomacula and restores internodal length in HNPP mice at 30 days of age (P30). (a) Semithin cross-sections of sciatic nerves stained with toluidine blue from the indicated genotypes 30 days of age. Tomacula can be distinguished in Pmp22 +/− and double mutants by their myelin thickening (arrows). Scale bar, 50 μm. (b) G ratio was decreased by 0.035 at P20 and 0.038 at P90 only in Pmp22 +/− and not further modified by ablation of β4 integrin in Schwann cells. (c–e) Number of myelinated fibers at P20, P30 and P90 (c), axonal size distribution at P30 (d) and number of tomacula at P20, P30, P90 and P200 (e). Fibers were quantified from a whole semithin section of sciatic nerve. (f–g) Examples of paranodal tomacula in Pmp22 +/− and double mutant animals are shown in (g), and their quantification in (f). The appearance of tomacula is slightly delayed in Pmp22 +/−; Itgb4 SC −/− sciatic nerves as compared to Pmp22 +/+; Itgb4 SC +/+ nerves. Internodal length (h–i) quantified from osmicated teased sciatic nerve fibers from at 30 days of age. Note that ablation of β4 integrin improves the shorter internodal length of Pmp22 +/− animals (h), especially in small caliber fibers (i). All graphs indicate mean ± s.e.m. n (animals) ≥ 3 per genotype (Data S1). Statistical analyses were performed using t-test with Bonferroni correction (b, d, f, h) and two-way ANOVA with Bonferroni post hoc test (i). * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.001. n.s. indicates non-significant.

In contrast, electrophysiological analyses in 1-month old animals showed that ablation of α6β4 integrin worsens the nerve conduction velocities (NCV) of HNPP animals (Table 1). This was not due to a change in internodal length, which was increased in double mutants only in small myelinated fibers which do not contribute to NCVs (Li 2015). These data indicate that integrin α6β4 and PMP22 collaborate to allow proper propagation of the nervous signal along axons, in a way that is independent of myelin thickness (Fig. 1b) and of internodal length.

Table 1. Ablation of integrin β4 decreases nerve conduction velocities of HNPP animals.

Electrophysiological analysis of animals at 30 days of age. The analysis reveals a decrease of NCV (43.85 m/s vs. 33.06 m/s) and an increase of latency (1.28 ms vs. 1.75 ms) in Pmp22 +/− animals. Additional ablation of β4 integrin further decreased the NCV (24.54 m/s). ± indicate s.e.m. Statistical analysis was performed using one-way ANOVA. n.s. indicates non-significant. n ≥ 7 nerves per genotype (Data S1).

	Pmp22 +/+; Itgb4 SC +/+	Pmp22 +/+; Itgb4 SC −/−	Pmp22 +/−; Itgb4 SC +/+	Pmp22 +/−; Itgb4 SC −/−
Conduction velocity (m/s)	43.38±1.55	40.30±2.62	33.06±1.01	24.35±0.5
	n.s.		P = 0.0017	P = 0.01
cMAP amplitude (mV)	3.53±0.38	2.84±0.22	2.38±0.12	2.88±0.14
	n.s.		n.s.	n.s.
cMAP latency (ms)	1.28±0.04	1.50±0.15	1.75±0.05	1.84±0.02
	n.s.		P = 0.0058	n.s.

Overall, ablation of integrin α6β4 affects positively the HNPP phenotype by reducing the number of tomacula and restoring internodal length in small calibre fibers, but it also reduces NCV. While these observations are conflicting, they likely represent different pathogenic mechanisms of Pmp22 haploinsufficiency, which are modulated differently by integrin α6β4.

Integrin α6β4-PMP22 do not colocalize in myelinating Schwann cells

Because integrin α6β4 is present at the outer Schwann cell membrane (Feltri et al. 1994, Einheber et al. 1993), while PMP22 is localized in compact myelin (Haney et al. 1996) (see also Fig. S1), we hypothesized that the two proteins could interact in mature myelinated fibers in trans. We confirmed that integrin α6β4 and PMP22 can be co-immunoprecipitated from lysates of sciatic nerves at P30, as reported (Amici et al. 2006), and that this interaction is lost in PMP22-null or β4 integrin SC-null mice (Fig. 2a). However, using cell lines expressing various truncated forms of β4 integrin and transfected with an HA-PMP22 plasmid, we found that all truncated forms of β4 integrin appear to bind PMP22 (Fig. 2c). These data indicate that the two proteins do not interact in trans, but they could interact in cis via the only domain of β4 integrin that was not deleted, namely the transmembrane domain (i.e. amino acid 661–853) (Fig. 2d). From three distinct immunohistochemical approaches (i.e. proximity ligation assay, immunohistochemistry on sciatic nerve cross sections and on teased fibers), we showed that integrin α6β4 does not colocalize with PMP22 in the outer layer of the Schwann cell nor in compact myelin at P30 or as early as P3 (Fig. 2e–h, Fig S2, Data S2).

Figure 2. β4 integrin and PMP22 co-immunoprecipitate, but do not colocalize in myelinating Schwann cells.

(a) Immunoprecipitation of β4 integrin in wild type, Itgb4 SC −/− and Pmp22 −/− sciatic nerves at 30 days of age. The data show that PMP22 immunoprecipitates with β4 integrin in wild type sciatic nerves, but not in nerves lacking either β4 integrin or PMP22. I = input, IP = immunoprecipitation and Ub = unbound fraction. (b) Post-lysis immunoprecipitation of PMP22 and β4 integrin. Human β4 integrin or HA-PMP22 were transfected independently into separate populations of CHO cells. The left panel is a western blot of the CHO cells transfected either with β4 integrin or with PMP22. The right panel is an immunoprecipitation performed from lysates of cells transfected only with integrin β4 subunit, only with HA-PMP22, or from a mix of the two lysates. The data show the interaction of PMP22 and β4 integrin detected in vitro after lysis of the cells. (c–d) Immunoprecipitation of PMP22 with various truncated forms of β4 integrin. 804G cell lines that stably express the truncated forms of human β4 integrin (cl. A–F) were transfected with HA-PMP22. The various truncated forms of β4 integrin (red) are depicted in (d) while interacting with the α6 integrin subunit (blue). All the truncated forms of β4 integrin interact with PMP22. cl.Ø does not express β4 integrin. (e) Proximity ligation assay (PLA) of β4 integrin and PMP22. Proximity ligation assay generates a signal (red dots) only if two proteins are in close vicinity (< 30 nm). The analysis does not detect close localization between β4 integrin and PMP22. (f) Immunolocalization of α6β4 integrin and PMP22 in teased fibers (f) and cross-sections (g–h) of sciatic nerve at 3 days (h) and 30 (f–g) days of age. Sections were stained for PMP22 (green), α6β4 integrin (red) and DAPI (blue). The data show that α6β4 integrin does not colocalize with PMP22 in adult myelinated Schwann cells. (Enlarged figures are displayed in Figure S2. A z-stack movie of a teased fiber is displayed in Data S2). Images were acquired with an apotome or a confocal microscope. Scale bar, 50 μm (e), 20 μm (f), 5 μm (g and h).

The discordance between the presence of co-immunoprecipitation from sciatic nerve lysates, the absence of colocalization and the persistence of co-immunoprecipitation in clones lacking most of the β4 integrin domains, may indicate that the interaction between integrin α6β4 and PMP22 is artefactual, and occurs in cell lysates, but not in vivo. This alternative explanation is supported by the observation that the interaction between β4 integrin and PMP22 can occur in vitro, after solubilization and mixing of the two proteins from two independent lysates, as shown in Fig. 2b. Overall, because we could not identify a specific β4 integrin domain of interaction with PMP22, nor could co-localize the two proteins in myelinating Schwann cells, we conclude that the two molecules probably do not physically interact in myelinating Schwann cells in vivo.

Ablation of integrin α6β4 increases PMP22 protein levels

Even if PMP22 and integrin α6β4 do not interact directly, the modification of the HNPP phenotype observed in Pmp22 +/−; Itgb4 SC −/− mice supports the existence of a functional interaction between the two proteins. We thus asked if ablation of β4 integrin affects the expression of Pmp22. We did not observe an effect on the mRNA levels of Pmp22 or the myelin related genes Egr2 and Mpz (Fig. 3a). However, the protein levels of PMP22 were modestly increased by 25% in Pmp22 +/+; Itgb4 SC −/− and in double mutant animals (Fig. 3b). Conversely, it was shown that absence of PMP22 could affect the expression and organization specifically of the α6β4 integrin basal lamina receptor (Amici et al. 2006, Amici et al. 2007). Thus, we asked if Pmp22 haploinsufficiency could affect the expression of α6β4 integrin or other basal lamina receptors. However, we did not observe any difference in the mRNA, protein levels, or localization of any laminin receptors in HNPP sciatic nerves (Fig. S3).

Figure 3. PMP22 protein levels are increased in Pmp22 +/−; Itgb4 SC −/− sciatic nerves.

(a–b) mRNA (a) and protein (b) levels for Pmp22, Egr2 and Mpz were assessed in sciatic nerves of the indicated genotypes at 30 days of age. Pmp22 mRNA and protein levels are reduced by 50% in Pmp22 +/−. Ablation of Itgb4 does not affect Pmp22, Egr2, Mpz mRNA levels, but causes an increase in the levels of PMP22 protein. Also, ablation of Pmp22 does not affect the protein levels or the localization of laminin receptors (Fig. S3). Graphs indicate mean ± s.e.m. n (animals) = 3 per genotype (Data S1). Statistical analyses were performed using t-test (a) and one-way ANOVA with Bonferroni post hoc test (b). * P < 0.05, ** P < 0.01.

NCV defects can be due to thinner myelin, shorter internodes, smaller axons, alterations in the electrical properties of myelin due to imbalanced lipid composition or abnormal paranodal axoglial junctions (Coetzee et al. 1996, Miyamoto et al. 2005, Dupree et al. 1998, Sherman et al. 2005, Court et al. 2004). We showed that there is no overall decrease in the number of myelinated fibers, internodal length or myelin thickness in Pmp22 +/−; Itgb4 SC −/− mice (Fig. 1). Therefore, it is possible that modifications of the NCVs are associated with other morphological features such as the electrical properties of the myelin sheath. Recently studies have reported that deficiency of Pmp22 in HNPP causes abnormal myelin permeability, also called “functional demyelination” (Guo et al. 2014, Hu et al. 2016). According to this model, the absence of Pmp22 affects the formation of tight and adherens autotypic junctions during development, and causes an increase of p21-activated kinase (PAK1) (Hu et al. 2016, Guo et al. 2014). The combined haploinsufficiency of PMP22 with increased PAK1 levels then leads to a further disruption of autotypic junctions and of F-actin organization in animals older than 3-months (Hu et al. 2016). It is possible that the absence of α6β4 integrin aggravates or accelerates the defects associated with myelin permeability in HNPP nerves. To test this hypothesis, we analyzed the localization of F-Actin at Schmidt-Lanterman incisures, Claudin-19 at paranodes and Kv1.1 at juxtaparanodes. In 1-month old sciatic nerves from Itgb4 SC −/−; Pmp22 +/− animals, we did not observe mislocalization of these proteins (Fig. 4a–b, d). In 3-month old animals, we confirmed that staining for F-Actin was increased in Schmidt-Lanterman incisures of PMP22 mutants (Fig. 4c, e), and ablation of β4 integrin alone did not affect the F-Actin localization. However, F-actin staining was increased in Schmidt-Lanterman incisures of PMP22 and double mutants at 3-months (Fig. 4e) and F-actin was more disorganized in numerous internodes of double mutants, both at 1-month and 3-months of age (Fig. 4c, white lines, Fig. 4f). We also analyzed the activation of PAK1 at 15 days and 3 months of age. We confirmed that p-PAK1 levels are increased in Pmp22 +/− mutants, by 30% at P15 and by 60% at P30 (Hu et al. 2016), and this appeared to be amplified in double mutants (Fig. 4g). Thus, ablation of Itgb4 may aggravate some of the defects in compact myelin organization observed in Pmp22 mutants.

Figure 4. Some aspects of internode architecture are altered by ablation of PMP22 and β4 integrin.

(a–f) Immunolocalization of Claudin-19 (a), Kv1.1 (b) and F-Actin (a–f) in teased fibers from nerves of mice of the indicated genotypes at 30 days (a, b, d, f) and 90 days (c, e) of age. Localization of Claudin-19 at paranodes and mesaxons (a, empty arrows) and of Kv1.1 at juxtaparanodes and mesaxons (b, empty arrows) is preserved in Pmp22 +/−, Itgb4 SC −/− and double mutant mice at this age. While F-Actin does not accumulate in Schmidt-Lanterman incisures of animals at 1 month (d), it increases in incisures of Pmp22 and double mutants at 3 months (c, e). In addition, F-Actin is frequently disorganized in the internodes of double mutants (c, white line). Accumulation of F-Actin in Schmidt-Lanterman incisures was analyzed by measuring the fluorescence intensity of rhodamin-labeled phalloidin, using a (d) 6 × 6 μm² or (e) 9 × 9 μm² rectangle including the entire area of every incisures. n ≥ 50 Schmidt-Lanterman incisures per genotype (Data S1). Fluorescence was increased in Pmp22 +/− and double mutant mice at P90. (f). Schmidt-Lanterman incisures were quantified from at least 30 internodes per genotype (Data S1). There are less Schmidt-Lanterman incisures in nerves from double mutants than in those from Pmp22 +/− Error bars indicate s.d. (g) Western blot analyses for p-PAK1 and PAK1 in sciatic nerves from mice at 15 and 90 days of age. p-PAK1 increase in Pmp22 +/− mutants. Combined ablation of Pmp22 and Itgb4 increases the level of p-PAK1 when compared to Pmp22 +/−. Graphs indicate mean ± s.d. Statistical analyses were performed using t-test (e) and One-way ANOVA with Bonferroni post hoc test (f). ** P < 0.01, **** P < 0.0001.

Identification of novel partners of α6β4 integrin to further elucidate the functional interaction with PMP22

To obtain more insights in the functional interaction between α6β4 integrin and PMP22, we used mass spectrometry to identify novel proteins that immunoprecipitate with β4 integrin from sciatic nerve lysates at 30 days of age (Fig. 5a). From two independent experiments, we identified 73 candidate partners for β4 integrin. All these partners were significantly enriched in immunoprecipitates of wild-type as compared to Itgb4 SC −/− nerves (Data S3). Among the 73 proteins, we identified several known partners of β4 integrin, including α6 integrin, vimentin, plectin and the laminin β1 subunit (Rezniczek et al. 1998, Homan et al. 2002, Lotz et al. 1990, Sonnenberg et al. 1988) (Data S3). Interestingly, we identified periaxin as a novel interactor of β4 integrin (Data S3). Both Prx −/− and Pmp22 +/− sciatic nerves develop similar tomacula and manifest decreased NCVs (Court et al. 2004, Gillespie et al. 2000), making it a candidate to functionally link β4 integrin and PMP22. However, the protein levels and localization of periaxin were normal in Pmp22 +/−; Itgb4 SC −/− animals (data not shown).

Figure 5. β4 integrin interacts with inositol triphosphate gated calcium channel ITPR3.

(a) Coomassie blue staining shows the difference between Pmp22 +/+; Itgb4 SC +/+ and Pmp22 +/+; Itgb4 SC −/− nerves after immunoprecipitation of the β4 Integrin subunit. (b) Molecular function Gene Ontology categories enriched among the partners of Integrin β4. The list of the proteins identified is shown in Data S3. Dashed lines indicated an indirect connection between two categories. (c) Immunoprecipitation of β4 integrin in Pmp22 +/+; Itgb4 SC +/+ and Pmp22 +/+; Itgb4 SC −/− sciatic nerves. The data confirm the interaction between β4 integrin and ITPR3 and the co-immunoprecipitation of the α6 integrin subunit. I = Input, IP = Immunoprecipitation and Ub = Unbound fraction. (d–f) Immunolocalization of ITPR3 in teased fibers from sciatic nerves at 30 days. Fibers were stained for β4 integrin, ITPR3, NaV, Caspr and Phalloidin. The data show that ITPR3 is enriched in the paranodal region (d). We also noticed a reduction in ITPR3 staining in the paranodes of β4 integrin mutants (e–f). Accumulation of ITPR3 at the paranode was analyzed by measuring the fluorescence intensity of ITPR3, using a 10 × 10 μm² rectangle including the entire area of every paranode. Scale bar, 15 μm. Nodes with adjacent tomacula were excluded from the analysis. n = 44–104 paranodes per genotype (Data S1). Graphs indicate mean ± s.d. One-way ANOVA with Bonferroni post hoc test (f). *** P < 0.001, **** P < 0.0001.

For an overview of the molecular functions of the β4 integrin network based on this list of proteins, we used BiNGO, a software which recognizes network linkage by gene ontology (GO) hierarchy. The most highly represented terms in the molecular function GO category were binding and protein binding (Fig. 5b and Data S3). Consistent with previous reports, β4 integrin interacts with several cytoskeletal and actin-binding proteins (Fig. 5b). This could potentially explain the slight disruption of F-actin organization that we observed in Itgb4 SC −/−; Pmp22 +/− internodes.

A careful examination of the map shows a statistically significant enrichment of interactors linked to calcium binding and calcium transport GO categories (Fig. 5b, red underlines). Specifically, we identified two novel Ca²⁺ transporters as β4 integrin partners, namely the plasma membrane calcium-transporting ATPase 4 (Atp2b4) and inositol 1,4,5-trisphosphate receptor type 3 (Itpr3) (Fig. 5b). ITPR3 is an inositol triphosphate receptor that controls intracellular Ca²⁺ concentrations. ITPR3 is localized in the endoplasmic reticulum, but has also been detected at the plasma membrane (Kuno & Gardner 1987) and in the nucleus (Nicotera et al. 1990). In Schwann cells, ITPR3 is localized at the nodal/paranodal region and in Cajal bands (Toews et al. 2007, Martinez-Gomez & Dent 2007). Of note, PMP22 may regulate intracellular Ca²⁺ concentration, and Nobbio et al. have proposed that PMP22 neuropathies could be caused by calcium dysregulation (Nobbio et al. 2009). To confirm that ITPR3 interacts with α6β4 integrin, we immunoprecipitated β4 integrin from wild-type nerves, and verified that ITPR3 and α6 integrin co-immunoprecipitation by western blot (Fig. 5c, left panel). In addition, the ITPR3 band was almost completely absent when the immunoprecipitation was performed from Itgb4 SC-null nerves, which retained very low levels of β4 integrin protein, likely deriving form perineurial cells of vessels (Fig. 5c, right panel). In contrast, some Atp2b4 could still be immunoprecipitated with β4 integrin from Itgb4 SC-null nerves (not shown) and we did not pursue Atp2b4 any further. We next examined the localization of ITPR3 in our mutants. In our hands, ITPR3 localized mainly to paranodal regions of wild-type nerves (Fig. 5d). α6β4 is present along all the outer Schwan cell membrane, including above the paranodal regions (Fig. 5d), suggesting that the two proteins co-localize and could interact in this location. Interestingly, we observed a reduction of the paranodal staining for ITPR3 in Itgb4 SC-null nerves that became more evident in Pmp22 +/−; Itgb4 SC −/− double mutants (Fig. 5e–f). Thus, ablation of Itgb4 alters the paranodal localization of ITPR3. Finally, we asked if this mislocalization of ITPR3 affects the concentration of calcium in Schwann cells. We performed a fluorometric determination of the intracellular Ca²⁺ concentration in Schwann cells co-cultured with DRG neurons from Pmp22 +/+; Itgb4 SC +/+ and Pmp22 +/−; Itgb4 SC −/− animals, but we did not observe any difference (data not shown).

Discussion

There is evidence that PMP22 maintains myelin integrity, and part of PMP22 function may be mediated through its physical interaction with β4 integrin (Amici et al. 2006). To address the role of α6β4 integrin in PMP22 neuropathies, we genetically ablated β4 integrin in Schwann cells together with PMP22. We demonstrate that integrin α6β4 influences nerve conduction velocity in HNPP animals without affecting myelin thickness but possibly by enhancing the abnormalities in the architecture of myelinated fibers. Furthermore, a6β4 integrin delays the progression of myelin abnormalities in HNPP. Finally, we provide several lines of evidence that the effects of β4 integrin on HNPP pathophysiology are probably indirect, because integrin β4 and PMP22 do not colocalize in myelinating Schwann cells.

The role of laminin-integrin signaling in Schwann cell development has been defined in a large body of work–for review see (Feltri et al. 2016, Monk et al. 2015). Among laminin receptors, the function of α6β4 integrin is the most elusive and subtle. α6β4 integrin is localized at the abaxonal surface of myelinating SCs, where it regulates myelin maintenance, as its absence cause an anticipated appearance of myelin outfoldings at paranodes and juxtaparanodes, that is significantly worsened by additional ablation of another laminin receptor, dystroglycan (Nodari et al. 2008). Ablation of α6β4 integrin in Schwann cells also causes disorganization of Kv1.1 in the internodal mesaxon (Nodari et al. 2008), and of F-actin in HNPP non-compact myelin (this paper). This, together with the identification of a new candidate partner, ITPR3, enriched in the paranodal region of Schwann cells and with the worsened nerve conduction velocities of HNPP animals lacking β4 integrin, suggest a more general function for integrin α6β4 broadly related to the organization of the architecture of myelinated fibers. Similarly, laminins 211 and 511 are enriched in the nodal region (Occhi et al. 2005) and the laminin receptor dystroglycan is important for the organization of Schwann cell microvilli and the localization of axonal Na⁺ channels (Saito et al. 2003, Occhi et al. 2005, Colombelli et al. 2015). We postulate that the slight disorganization of non-compact myelin domains in internodes, Schmidt-Lanterman incisures and paranodes caused by ablation of Itgb4 in Schwann cells may exacerbate the functional defects of HNPP, where 35 – 45 % of the nodes have abnormal myelin foldings, leading to a worsening of NCVs.

Myelin instabilities such as outfoldings and tomacula may derive from a common mechanism (Goebbels et al. 2012). The results of our study on PMP22 tomacula and α6β4 integrin outfoldings could be viewed as conflicting with this hypothesis, as we do not observe a worsening of myelin instabilities in animals lacking both proteins. However, we detected a novel interaction of integrin α6β4 with an inositol triphosphate receptor, ITRP3. Phosphoinositol and AKT signaling have been associated with myelin instabilities (Pereira et al. 2012). Thus, further studies will need to focus on the potential role of phosphoinositol receptors at nodes of Ranvier, in the regulation of Ca²⁺ concentration at the paranodes, and in the maintenance of myelin stability.

Abbreviation

HNPP

Hereditary Neuropathy with Pressure Palsy

SC

Schwann cell

NCV

nerve conduction velocities