Structural basis of myelin-associated glycoprotein adhesion and signalling

ABSTRACT Myelin-associated glycoprotein (MAG) is a myelin-expressed cell-adhesion and bi-directional signalling molecule. MAG maintains the myelin–axon spacing by interacting with specific

neuronal glycolipids (gangliosides), inhibits axon regeneration and controls myelin formation. The mechanisms underlying MAG adhesion and signalling are unresolved. We present crystal

structures of the MAG full ectodomain, which reveal an extended conformation of five Ig domains and a homodimeric arrangement involving membrane-proximal domains Ig4 and Ig5.

MAG-oligosaccharide complex structures and biophysical assays show how MAG engages axonal gangliosides at domain Ig1. Two post-translational modifications were identified—N-linked

glycosylation at the dimerization interface and tryptophan C-mannosylation proximal to the ganglioside binding site—that appear to have regulatory functions. Structure-guided mutations and

neurite outgrowth assays demonstrate MAG dimerization and carbohydrate recognition are essential for its regeneration-inhibiting properties. The combination of _trans_ ganglioside binding

and _cis_ homodimerization explains how MAG maintains the myelin–axon spacing and provides a mechanism for MAG-mediated bi-directional signalling. SIMILAR CONTENT BEING VIEWED BY OTHERS

STRUCTURAL INSIGHTS INTO THE CONTACTIN 1 – NEUROFASCIN 155 ADHESION COMPLEX Article Open access 03 November 2022 ARCHITECTURE AND REGULATION OF A GDNF-GFRΑ1 SYNAPTIC ADHESION ASSEMBLY

Article Open access 20 November 2023 THE DYNAMIC NATURE OF NETRIN-1 AND THE STRUCTURAL BASIS FOR GLYCOSAMINOGLYCAN FRAGMENT-INDUCED FILAMENT FORMATION Article Open access 03 March 2023

INTRODUCTION Myelination of axons enables enhanced conductance velocity in both the central and peripheral nervous system (CNS and PNS) of vertebrates. It also provides electrical insulation

and a decrease of the capacitance, as well as physical protection and metabolic support of long axons1. Myelin-associated glycoprotein (MAG) adhesion and signalling at the myelin–axon

interface regulates the formation and maintenance of myelinated axons, thus playing an important role in the development and function of the nervous system2,3. Aberrant MAG function, for

example from mutations that likely cause misfolding, or anti-MAG autoimmunity, has been associated with demyelination and neurodegenerative disorders, such as corticospinal motor neuron

disease also known as hereditary spastic paraplegias4, Pelizaeus–Merzbacher disease-like disorder5, demyelinating anti-MAG peripheral neuropathy6,7 and multiple sclerosis2,8. MAG is a type 1

single-pass transmembrane protein expressed on myelinating oligodendrocytes in the CNS and Schwann cells in the PNS2,3. MAG is the fifth highest expressed protein in myelin of the CNS9. It

is highly enriched at the innermost (adaxonal) myelin membrane along the internode, where it contacts the axon. MAG is also found on other myelin structures, such as the mesaxon,

Schmidt-Lanterman incisures and paranodal loops2,3. MAG adhesion maintains the myelin–axon spacing (periaxonal diameter) by interacting with specific neuronal gangliosides (glycolipids),

such as the major brain gangliosides GT1b and GD1a (refs 10, 11, 12, 13). More recently, the Nectin-like (Necl) proteins 1 and 4 have also been found to contribute to myelin–axon adhesion

along the internode14,15, although they are expressed less than MAG in mature myelin9 and knockout of Necl4 does not affect myelination16. MAG, also known as Siglec4a, is evolutionarily the

oldest member of the Siglec family17. Unlike all other Siglecs, MAG plays no role in the immune system and is exclusively expressed in the nervous system17. On the basis of the primary

sequence its extracellular region is predicted to consist of five Ig domains; an N-terminal V-type Ig domain that is typical for Siglecs and four C2-type Ig domains. This is followed by a

single membrane-spanning helix and an intracellular region predicted to be unstructured and of different length for two MAG isoforms, L-MAG and S-MAG. Like other Siglecs, MAG recognizes

sialic acid groups and the specificity of MAG has been established to be Neu5Ac-α2,3-Gal-β1,3-GalNAc (ref. 18). This trisaccharide is part of several neuronal gangliosides, most notably the

major brain gangliosides GT1b and GD1a, but also GM1b, GT1β and GQ1bα. MAG bridges the periaxonal space by interacting with these axonal gangliosides in _trans_ via the canonical Siglec site

at a conserved arginine (R118 in MAG) in the N-terminal domain19,20. MAG signalling is bidirectional, engaging in both axon-to-myelin as well as myelin-to-axon signalling. MAG has been

extensively studied as one of three classic myelin-associated inhibitors of central nervous system regeneration, the other ligands being Nogo66 and oligodendrocyte myelin glycoprotein2,3.

MAG inhibits neurite outgrowth and collapses axonal growth cones in a sialic acid binding-dependent manner. It does so as full-length transmembrane20,21, but also as a proteolytically shed

and soluble form called dMAG22. As a receptor, MAG controls myelin formation and integrity. How MAG transduces the extracellular signal into the myelinating cell is not well understood, but

it has been shown that the cytosolic domain of the L-MAG isoform binds to the cytoplasmic non-receptor tyrosine kinase Fyn23 and that antibody-induced crosslinking of L-MAG triggers its

localization to lipid rafts24 and activates Fyn in oligodendrocytes23. This activation of Fyn is essential for the initiation of myelination25. In contrast, the shorter MAG isoform S-MAG

binds to zinc and microtubules and this is postulated to have a structural function in mature myelin26,27. From earlier rotary-shadowed electron microscopy (EM) and sedimentation velocity

analytical ultracentrifugation (AUC) studies it was hypothesized that the extracellular segment of MAG has a back-folded Ig-horseshoe type structure, but the estimated maximum dimensions of

8.8 and 18.5 nm determined by AUC and EM, respectively, deviate substantially28,29. In the absence of any high-resolution structural data on MAG or its interaction with ganglioside ligands,

the conformation of the five Ig domains, the extracellular specificity-determining parameters and the mechanisms underlying MAG adhesion and bi-directional signalling are unresolved. Using a

combination of structural, biophysical and cellular techniques, we provide the structural basis of MAG-mediated adhesion and identify a dimerization-dependent mechanism that explains how

MAG regulates axon-to-myelin and myelin-to-axon signalling, and controls myelin–axon spacing. RESULTS MAG HAS AN EXTENDED CONFORMATION We determined crystal structures of the full

extracellular segment of mouse MAG (MAG1–5) in two different crystal forms that diffracted to a maximum resolution of 3.8 and 4.3 Å. These crystals were obtained by enzymatic deglycosylation

of MAG1–5 or reductive lysine methylation of glycosylated MAG1–5 (see ‘Methods’ section). In addition, crystals of a shorter construct, consisting of the three N-terminal domains (MAG1–3),

diffracted to a maximum resolution of 2.1 Å. The structures were solved by molecular replacement with individual Ig domains from homologous proteins. The exceptionally high-solvent content

of the two MAG1–5 crystal forms (91 and 85%, Supplementary Fig. 1) aided in obtaining phases of sufficient quality for initial model building (see also Table 1 and ‘Methods’ section for

details). In all three crystal forms MAG has an extended collinear conformation (Fig. 1). Only consecutive Ig domains interact with each other via hydrophobic interfaces (buried-surface area

ranging from 243 to 690 Å2) and short inter-domain linkers of up to two residues (Fig. 1a). Domains Ig1 and Ig2 form the largest interface in which the Ig2 loops at the N-terminal ‘head’

side interact with the A2-B (Ig domain β–strand numbering) side of Ig1 (Fig. 1a). The three other inter-domain interfaces are exclusively formed in a head-to-tail manner involving loops at

the N-terminal ‘head’ and C-terminal ‘tail’ side of the Ig domains (Fig. 1a). As predicted from the primary sequence the N-terminal Ig1 domain of MAG has a V-type Ig fold like other Siglec

family members30,31,32 and domains Ig3 and Ig4 are of the C2 type. Domains Ig2 and Ig5, however, have a C1-type Ig fold, contrary to the predicted C2-fold (Supplementary Fig. 2). The three

crystal structures of MAG are similar to each other with only small differences within the domains (r.m.s.d. ranging from 0.93 to 2.13 Å) and inter-domain angle rotation differences ranging

from 3.4 to 17.4°, the largest difference is in the domain Ig2–Ig3 angle (Fig. 1b). The combination of hydrophobic inter-domain interfaces, the lack of flexible linker residues and a

previously predicted inter-domain disulfide between Ig1 and Ig2 (C37–C165)33 explains the limited inter-domain flexibility observed between the three different crystal forms. MAG IS

POST-TRANSLATIONALLY MODIFIED The structures reveal MAG is post-translationally modified at several sites. MAG contains seven disulfides, five of which are canonical for Ig domains.

Cysteines 37 and 165 form an inter-domain disulfide between Ig1 and Ig2, and cysteines 421 and 430 form an additional intra-domain disulfide in Ig5, as shown previously33 (Supplementary Fig.

3). In addition, MAG carries N-linked glycans and previously eight N-linked glycosylation sites were determined in human MAG by mass spectrometry analysis34. We observe glycan electron

density for five of those equivalent sites in the mouse MAG structures (on asparagine 99, 223, 246, 315 and 406; Supplementary Fig. 4). The differences in these observations arise from one

N-linked glycosylation site that is not conserved (N106 in human MAG is a threonine in mouse) and from poorly resolved electron density for the other two sites (on N450 and N454) that are

situated in a flexible loop. One additional N-linked glycan is revealed by clear electron density on N332, in contrast to the previous study that did not find this residue in a glycopeptide

analysis34. Electron density at the N332-linked glycan suggest it is fucosylated in our recombinantly produced MAG (Supplementary Fig. 5). In native mass spectrometry experiments of intact

recombinant MAG we also observed internal mass shifts of +147 Da on MAG monomers (Supplementary Fig. 6), further indicating fucosylation as a post-translational modification. Thus the

combination of our data and that of others34 indicates that mouse MAG has eight N-linked glycosylation sites (on N99, N223, N246, N315, N323, N406, N450 and N454). In addition, MAG carries a

tryptophan C-mannosylation on W22. In all three crystal forms, electron density proximal to the side chain of W22 suggests this residue is C-mannosylated (Supplementary Fig. 6). Indeed,

analysis of the MAG primary sequence reveals that this tryptophan is part of the canonical WxxW motif (W22 is the first tryptophan) for C-mannosylation, a rare post-translational

modification present on several secreted proteins35. The W22-attached α-D-mannopyranosyl group has an unusual ring-flipped 1C4 chair conformation in the structures (regular mannose is 4C1,

Supplementary Fig. 6). This ring-flipped conformation is in agreement with previous nuclear magnetic resonance studies on mannosyl-tryptophan36 and can be explained by the preference of the

bulky tryptophan, covalently attached to the C1 of the mannose, to be in the equatorial position. We confirmed the C-mannosylation of W22 by in-gel trypsin digestion of MAG1–5 followed by

liquid chromatography-mass spectroscopy (LC-MS/MS) analysis of the (glyco)peptide fragments (Supplementary Fig. 7). In addition, native mass spectrometry of wild type and mutated MAG1–5, in

which the second tryptophan of the C-mannosylation motif is substituted for glutamine (W25Q, resulting in WxxQ), showed a mass shift of −221 Da in accordance with a loss of C-mannosylation

and confirming the importance of the WxxW motif (Supplementary Fig. 6). MAG1–5 CRYSTAL STRUCTURES REVEAL A DIMERIC ARRANGEMENT MAG1–5 forms a symmetry-related dimer at a crystallographic

two-fold rotation axis in both MAG1–5 crystal forms. They share the same interface (Fig. 2a) on domains Ig4 and Ig5, which buries a surface area of 2,037 Å2. The CC′FG face of Ig4 binds to

the ABDE face of Ig5 of the symmetry-related molecule and vice versa, thus forming two equivalent hemi-interfaces. The interface is mostly hydrophobic apart from the negatively charged E395

in Ig4, with hydrophilic residues lining the edges of the interface (Fig. 2b). To validate the interface we generated two interface mutations based on the structures, that we predicted to

either disrupt dimerization (I473 to E) or enhance it (N406 to Q). The hydrophobic I473 in the middle of the hydrophobic interface of Ig5 was mutated to a negatively charged glutamate, to

ensure disturbance of the hydrophobic effect as well as introducing electrostatic repulsion with the opposing E395 in Ig4 (Fig. 2c). N406 carries an N-linked glycan and in the glycosylated

lysine-methylated crystals of MAG1–5, the density for this glycan suggests that it sterically clashes with its symmetry partner in the dimer (Fig. 2d). We hypothesized that a dimer with

increased affinity would form, were this glycan absent. Remarkably, the N406Q mutant that lacks this glycan was the only MAG construct that showed a distinct dimer peak in size exclusion

chromatography (SEC; Supplementary Fig. 8). The other glycans are not expected to interfere with dimerization, also not when considering myelin-specific N-linked glycans37. We confirmed that

MAG dimerizes in solution via the Ig4–Ig5 interface by analysing glycosylated and deglycosylated forms of MAG1–5, MAG1–3 (that lacks the dimerization domains) and the aforementioned mutants

in small angle X-ray scattering (SAXS) and AUC experiments. SAXS CONFIRMS DIMERIZATION INTERFACE As predicted, SAXS analysis indicated disruption of dimerization for MAG1–3 and MAG1–5

I473E, whereas dimerization is enhanced for MAG1–5 N406Q and deglycosylated MAG1–5 compared with wt MAG1–5. The molecular mass (_M_m) based on the extrapolated intensity at zero scattering

angle (_I_0, scaled for concentration to bovine serum albumin (BSA)), the radius of gyration (_R_g), the maximum interatomic distance (_D_max) and the Porod volume all show these trends

(Fig. 2, Supplementary Fig. 9 and Table 2). The data show that at similar concentrations, MAG1–5 I473E appears smaller than wt MAG1–5 and MAG1–5 N406Q appears larger than wt MAG1–5. This can

be attributed to a shift in the monomer–dimer equilibrium; MAG1–5 I473E has a lesser and MAG1–5 N406Q a greater propensity to dimerize compared with wt MAG1–5. Furthermore, the MAG1–5 I473E

pair distance distribution function P(r) and derived _D_max, the _ab initio_ models as well as the Kratky plots confirm that MAG has an extended conformation and behaves as a semi-rigid rod

in solution (Fig. 2f,g,i). Whereas the SAXS data for glycosylated MAG1–3 and MAG1–5 I473E fit best to scattering curves calculated from single chains of the crystal structures of MAG1–3 and

MAG1–5, respectively (_χ_=2.87 and 2.95, Supplementary Fig. 10), the glycosylated MAG1–5 N406Q SAXS data fits best to scattering curves calculated from the dimer structure (_χ_=3.45,

Supplementary Fig. 10). Both glycosylated and deglycosylated MAG1–5 wt SAXS data fit best to scattering curves calculated from a combination of monomeric and dimeric crystal structures

(_χ_=4.05 for glycosylated and 3.72 for deglycosylated MAG1–5, Fig. 2j and Supplementary Fig. 11). Furthermore, _ab initio_ models based on the SAXS data of deglycosylated MAG1–5 I473E and

MAG1–5 N406Q agree remarkably well with the crystal structures of the monomer and dimer of MAG1–5, respectively (Fig. 2g, Supplementary Fig. 12, _χ_2 of the model-to-data fit are 1.05 and

1.33 for MAG1–5 I473E and MAG1–5 N406Q, respectively). These data confirm MAG dimerizes via domains Ig4 and Ig5 and that the MAG1–5 chains have an extended and relatively rigid conformation.

THE MAG1–5 DIMER IS WEAK IN SOLUTION WITH A _K_D OF 3.8 × 102 ΜM To quantify the affinity of MAG dimerization in solution, we performed sedimentation equilibrium AUC (SE-AUC) experiments

for wt MAG1–5, deglycosylated wt MAG1–5, MAG1–3, MAG1–5 I473E and MAG1–5 N406Q constructs. For each sample a global analysis was performed for different concentrations, centrifugal speeds

and wavelengths together (Supplementary Figs 13–16), except for MAG1–5 N406, which appeared to suffer from aggregation during the experiment. Similar to the SAXS analysis, the MAG1–5 I473E

and MAG1–3 SE-AUC data fit best to a single species that agrees with the _M_m of a monomer (Table 3). For wt MAG1–5 and deglycosylated MAG1–5, we could fit the data to a monomer–dimer

equilibrium, with _K_ds of 3.8 × 102 and 1.7 × 102 μM, respectively (Table 3). On the basis of a monomer–dimer equilibrium with these _K_ds, dimer fractions of 18% for glycosylated wt MAG1–5

and 24% for deglycosylated wt MAG1–5 are expected to be present in the SAXS experiments (calculated at 52.2 and 37.5 μM for glycosylated and deglycosylated MAG, respectively). Indeed, the

presence of a mix of monomers and dimers is observed in the SAXS data of both glycosylated and deglycosylated MAG1–5 (Fig. 2j, Supplementary Fig. 11). The lack of dimers in the AUC

experiments for MAG1–5 I473E and MAG1–3 and the higher affinity for deglycosylated MAG1–5 compared with glycosylated MAG1–5 further confirm that MAG forms dimers via domains Ig4 and Ig5

(Table 3). STRUCTURAL BASIS OF LIGAND RECOGNITION BY IG1 MAG binds to sialic acids of gangliosides with its N-terminal V-type Ig domain and has a preference for a Neu5Ac-α2,3-Gal-β1,3-GalNAc

moiety18. We observed unmodelled electron density in the 2Fo−Fc and Fo−Fc maps of the deglycosylated MAG1–5 structure close to R118 in the putative ligand-binding site20 (Fig. 3a). Native

mass spectrometry of purified MAG1–5 revealed a mixture of free and two ligand-bound MAG1–5 forms with mass differences of 835±2 and 854±2 Da compared with free MAG1–5 (Fig. 3b). The 835 Da

ligand possibly corresponds to a tetrasaccharide comprising the aforementioned Neu5Ac-α2,3-Gal-β1,3-GalNAc trisaccharide plus another hexose. The 854 Da ligand might correspond to a similar

tetrasaccharide where Neu5Ac is replaced by Neu5Gc, a mammalian sialic acid variant not produced by humans. These ligands are likely co-purified in complex with MAG from the expression

medium that contains beef digest (Primatone). We do not observe clear electron density for any ligands in the two other crystal forms (lysine-methylated MAG1–5 and MAG1–3). In these crystal

forms the unliganded MAG is apparently preferentially crystallized over the ligand-bound forms. The resolution of the MAG1–5–ligand complex (to 3.8 Å) is not sufficient to determine the

detailed structure of the bound oligosaccharide. However, the unmodelled electron density is compatible with either of the suggested compounds. To obtain more detailed information on

MAG–ligand interactions, the commercially available trisaccharide 3′-N-Acetylneuraminyl-N-acetyllactosamine (Neu5Ac-α2,3-Gal-β1,4-GlcNAc) was soaked into the MAG1–3 crystals, as these

provided higher resolution data. For the soaked crystals, diffraction data was collected to 2.3 Å resolution. The crystals were isomorphous to the unsoaked MAG1–3 crystals (Table 1) and a

Fo(soaked)−Fo(unsoaked) map revealed clear electron density in the oligosaccharide binding site for one of the two MAG1–3 chains in the asymmetric unit (Fig. 3c). Most likely the other chain

remained unliganded due to occlusion of the binding site by crystal packing. The ganglioside binding side is formed by the CC′ loop and the F and G β-strands of the N-terminal V-type Ig1

domain. In agreement with previous data20, the side chain of R118 in strand F forms a key salt bridge with the carboxylic acid group of the sialic acid in the trisaccharide ligand. Y65 in

the CC′ loop forms extensive Van der Waals’ contacts with the ligand as well as a hydrogen bond to the O9 of the Neu5Ac sialic acid. Other contributing interactions are made by: (1) the

backbone carbonyls of N126 and T128 in strand G that form hydrogen bonds with H-N5 and the H-O9 of the sialic acid respectively, (2) the backbone amide proton of T128 that forms a hydrogen

bond with the O8 of the sialic acid and (3) the side chain of Y127 that forms Van der Waals’ contacts with the glycerol group (C7–C9) of the sialic acid (Fig. 3d). We validated the

ganglioside binding properties of MAG by mutating ligand-binding residues to alanine in MAG1–5 and probed their interaction with GT1b ganglioside incorporated into liposomes. The set-up we

used, MAG1–5 coupled at the C-terminus to a streptavidin-coated surface plasmon resonance (SPR) chip and GT1b-containing liposomes in the mobile phase, enables avidity-enhanced interactions

that also occur in _trans_ between cells (Fig. 3f, see ‘Methods’ section). Indeed we observed specific binding of GT1b liposomes to wt MAG1–5, no interactions with the ligand-binding mutants

MAG1–5 R118A, T128A and Y127A and reduced interaction with MAG1–5 Y65A (Fig. 3f). In addition, the MAG1–5 W25Q mutant that lacks the tryptophan mannosylation on W22, still interacted with

GT1b liposomes in this assay. Remarkably, this W25Q mutant appeared to have higher affinity for the GT1b liposomes compared with wt MAG. This suggests that rather than contributing to the

interaction strength, this tryptophan mannosylation on W22 may play a regulatory role in binding (membrane-embedded) gangliosides. In summary, we have shown that MAG interacts with

membrane-bound gangliosides via the side chain of R118, the CC′-loop and the F and G β-strands of the N-terminal V-type Ig1 domain and that the W22 mannosylation does not enhance ganglioside

binding. Ligand interaction of MAG is similar to the sialic acid recognition of other Siglec family members (Fig. 3g)30,32. As in Siglec5 (ref. 32), the MAG Ig1 CC′-loop (residues 64–70)

seems to undergo conformational selection on ligand binding. This loop adopts a single conformation when ligand is bound, whereas it can have several conformations (including the

ligand-bound conformation) or is unstructured in the different unliganded MAG crystal forms (Fig. 3e). Furthermore, this loop adopts different conformations in the Siglec -1, -5 and -7

structures (both unliganded and ligand-bound forms, Fig. 3g)30,31,32. The combination of our structural and biophysical data on MAG–ganglioside interaction, with that of others on MAG’s

specificity for Neu5Ac-α2,3-Gal-β1,3-GalNAc (ref. 18) establishes the structural basis of ganglioside recognition by MAG. NEURITE OUTGROWTH INHIBITION DEPENDS ON MAG DIMERIZATION We tested

different MAG variants in neurite outgrowth assays to determine the role of MAG dimerization for neuronal plasticity inhibition (Fig. 4). In agreement with previous data20,38, MAG1–5 wt on

coverslips inhibited neurite outgrowth of hippocampal neurons compared with poly-L-lysine (PLL)-covered slips (Fig. 4a,b). Other dimeric variants (MAG1–5 N406Q and MAG1–5-Fc) inhibited

neurite outgrowth to a similar level (Fig. 4d,f). The monomeric MAG1–5 I473E and MAG1–3 wt on the other hand showed no significant neurite outgrowth inhibition (Fig. 4c,e). Interestingly,

MAG1–5-Fc R118A, which is dimeric but lacks sialic acid binding properties, showed neurite outgrowth stimulation instead of inhibition, compared with PLL (Fig. 4g). These data indicate that

dimerization through domains Ig4–Ig5 and the ability to bind sialic acid moieties on the neuronal surface are required for neurite outgrowth inhibition signalling by MAG for hippocampal

neurons. DISCUSSION MAG controls adhesion and signalling between myelinating cells and axons. In contrast to earlier studies28,29, we find that MAG does not fold back onto itself like an

Ig-horseshoe as in the L1CAM and axonin neuronal adhesion molecules. Instead, our data show that the extracellular region of MAG has an extended shape with limited inter-domain flexibility,

similar to several other cell adhesion molecules such as SYG, Cadherin and Nectin family members39,40,41. The structure of MAG is the first of a full extracellular portion of a Siglec family

member. Besides the common N-terminal V-type Ig domain for recognizing sialic acid moieties, Siglecs vary in the number of additional Ig domains; from 1 up to 16. Comparison of the

structures of MAG and Siglec5 reveals a different inter-domain orientation between domains Ig1 and Ig2, likely due to differences in amino acids at the interface (Supplementary Fig. 17 and

Supplementary notes). By binding to axonal gangliosides, MAG maintains a defined spacing between the innermost myelin surface and the axon surface10,11,12,42. This myelin–axon spacing has

been reported to be 12–14 nm based on electron micrographs of chemically fixed myelin tissue10,11,12. However, analysis of more recent electron micrographs of high-pressure frozen myelin

that does not suffer from fixation induced artefacts43 reveals an axon–myelin spacing of 9–12 nm. This periaxonal diameter matches well with a straightforward model that follows from our

structural data of the MAG dimer and the MAG–ganglioside interaction; two opposing membrane surfaces are spaced 10 nm apart when the membrane-proximal C-termini of the MAG dimer are

positioned on one membrane (the structures lack only two residues to the transmembrane helix) and the MAG dimer-bound gangliosides are positioned on the other membrane (Fig. 5). Although the

two crystal forms of the full extracellular segment of MAG have different inter-domain angles (3.4–17.4°), the overall arrangement and resulting structure-based axon–myelin spacing is

similar (see Fig. 2a). The agreement of intermembrane distance determined from high-pressure frozen EM on myelin tissue43 and here by structural and biophysical data on the extracellular

segment of MAG indicates that in the periaxonal space, MAG is dimerized _in cis_ via domains Ig4–Ig5 when bound to axonal gangliosides _in trans_. Intriguingly, in this model, the unusual

tryptophan C-mannosylation on W22 is positioned at the interface of MAG and the extracellular leaflet of the axonal membrane (Fig. 5, Supplementary Fig. 18). The WxxW motif is conserved

among MAG orthologues in vertebrates from fish to human, but is absent in all other Siglec paralogs (Supplementary Fig. 6). This suggests that tryptophan mannosylation is specific for the

function of MAG. We showed that the mannosyl group does not enhance the binding of MAG to GT1b ganglioside liposomes but may weaken it. Possibly, tryptophan mannosylation of MAG provides

specificity to sialylated ligand interactions. Alternatively, the close proximity of the mannosyl group to the axonal membrane during MAG–ganglioside interaction may indicate a regulatory

role in axonal membrane engagement. The buried-surface area of the MAG dimerization interface formed by Ig4 and Ig5 is large (2,037 Å2) and hydrophobic. We find, however, that the affinity

of MAG dimerization via this interface in solution is low (_K_d of 3.8 × 102 μM). The weak interaction is probably important in the native context, where MAG is expressed on the cell

surface, as the _cis_ dimer can be _trans_ stabilized by interaction with gangliosides on the opposing axonal membrane. For the N-Cadherin family of cell adhesion molecules it has been shown

that affinities as weak as 10 mM in solution are functionally important in the context of a _trans_-stabilized _cis_-dimer39. Dimerization of MAG may serve two purposes. It provides a

mechanism to restrain the intermembrane distance, since a _cis_–_trans_ stabilized MAG dimer would restrict angular freedom with respect to the membrane more than a _trans_ only stabilized

MAG monomer. In this sense, the MAG dimer could function as a ‘molecular leaf spring’ that maintains the well-defined spacing between the axonal membrane and the adaxonal myelin membrane

along the internode. In addition, dimerization of MAG could enable compaction of the periaxonal space. The weak _cis_-interaction of MAG, if not stabilized in _trans_, may ensure enough

monomer is available to bridge a wider periaxonal spacing (of up to 16 nm, based on the length of a MAG monomer) that may exist during myelin formation. Even greater distances could be

bridged if MAG binds to sialylated N- or O-linked glycans of axonal surface glycoproteins before reaching the gangliosides. When _trans_ interactions with gangliosides have been established

and possibly local concentrations are elevated due to the abundance of ganglioside ligands, formation of MAG _cis_-dimers is triggered concomitant with compaction of the periaxonal diameter

to its final spacing (Fig. 5). Dimerization of MAG can be regulated by modulating glycosylation on N406. Large and charged glycans on N406 can obstruct dimerization because of steric clashes

and coulombic repulsion. Trimming all MAG N-linked glycans down to single N-Acetylglucosamines by deglycosylation with Endo-Hf or preventing glycosylation on N406 by mutating it to

glutamine enhances dimerization (Fig. 2, Supplementary Figs 8–12 and 16, Tables 2 and 3). This glycosylation site might play a regulatory role during myelin development and

myelination-related pathologies. MAG glycosylation changes during development44 and abnormal glycosylation of MAG correlates with myelination deficiencies45,46,47. Possibly, modulation of

N406 glycosylation, either at the biosynthesis level or by extracellular trimming, affects MAG dimerization and thereby impacts on the myelin–axon interaction (see Supplementary notes for

details). Myelin-to-neuron signalling with MAG as a ligand can inhibit neurite outgrowth2,3,20,21,48. Studies based on MAG truncations, chimeras and mutant versions suggested that the

inhibitory properties reside in domains Ig4 and Ig5, and in the sialic acid binding site20,49,50. For example, a chimeric protein consisting of Siglec1 Ig domains 1–3 plus MAG Ig4-5, but not

Siglec1 domains 1–3 alone, has neurite outgrowth inhibition properties similar to MAG49. We confirmed that the sialic acid-binding properties of MAG are required for neurite outgrowth

inhibition in hippocampal neurons as the MAG1–5-Fc R118A mutant that lacks ganglioside binding properties does not inhibit, but surprisingly, stimulates neurite outgrowth (Fig. 4g). This has

not been reported before and may be an interesting new avenue for therapeutic intervention to enhance central nervous system regeneration. In addition we show that domains Ig4 and Ig5 are

essential for MAG dimerization and that neurite outgrowth inhibition is abrogated by the I473E point mutation that monomerizes MAG (Fig. 4e,c). Thus, it is the dimerization of MAG that is

required for neurite outgrowth inhibition, rather than direct interactions of domains Ig4 and Ig5 with neuronal receptors, as previously suggested49,50. Although other protein receptors have

been identified that mediate the neurite outgrowth-inhibiting signalling by MAG3, it has been shown that direct clustering of gangliosides by antibodies also leads to inhibition of neurite

outgrowth of hippocampal neurons20. The combination of our data and that of others indicates that MAG dimerization at domains Ig4 and Ig5 and sialic acid binding at domain Ig1 induces

neurite outgrowth inhibition for hippocampal neurons, by clustering of gangliosides. Axon-to-myelin signalling with MAG as a receptor controls myelin formation. Antibody-mediated

extracellular clustering of the L-MAG isoform activates Fyn kinase23 and Fyn activation is essential for the initiation of myelination25. Our structures show that the C-termini of MAG1–5 are

separated by 5.4 nm in the dimer, bringing the cytosolic regions into close proximity (Fig. 5). Probably, L-MAG dimerization as a result of _trans_ interaction with gangliosides on the axon

brings the cytosolic regions of MAG into close proximity to trigger activation of Fyn, similar to Fyn activation by signalling lymphocytic activation molecule clustering in immune cells51.

Whether MAG forms higher-order clusters that are triggered by dimerization needs to be established, but preference of both MAG and Fyn for lipid rafts24,52 suggests that both proteins can be

locally enriched in the membrane to assist clustering. METHODS GENERATION OF CONSTRUCTS AND MUTAGENESIS MAG constructs were generated by polymerase chain reaction (PCR) using mouse S-MAG

(IMAGE 40039200) as a template and primers to start at (UNIPROT) residue number 20 (after the signal peptide) and end at residue 325 for MAG1–3 and residue 508 for MAG1–5. Point mutants were

also created by PCR, either by a two-step PCR with overlapping primers (W22A, W25Q, Y65A, R118A, Y127A, T128A, I473E,) or by a single-step PCR using non-overlapping phosphorylated primers

(N406Q). All constructs were subcloned using BamHI/NotI sites in pUPE107.03 (cystatin secretion signal peptide, C-terminal His6-tag), unless indicated otherwise. LARGE-SCALE EXPRESSION AND

PURIFICATION Constructs were transiently expressed in N-acetylglucoaminyltransferase I-deficient (GnTI−) Epstein–Barr virus nuclear antigen I (EBNA1)-expressing HEK293 cells in suspension

(U-protein express). Medium was collected 6 days after transfection and cells were spun down by 10 min of centrifugation at 1,000 _g_. Supernatant was concentrated fivefold and diafiltered

against 500 mM NaCl, 25 mM HEPES pH 7.8 (IMAC A) using a Quixstand benchtop system (GE Healthcare) with a 10 kDa molecular weight cut-off (MWCO) membrane. Cellular debris was spun down for

10 min at 9,500 _g_ and the concentrate was filtered with a glass fibre prefilter (Minisart, Sartorius). Protein was purified by Nickel–nitrilotriacetic acid (Ni–NTA) affinity chromatography

followed by SEC on a Superdex200 Hiload 16/60 column (GE Healthcare) equilibrated in SEC buffer (150 mM NaCl, 20 mM HEPES pH 7.5). Protein was concentrated to 7–14 mg ml−1 using a 30 kDa

MWCO concentrator before plunge freezing in liquid nitrogen and storage at −80 °C. CRYSTALLIZATION AND DATA COLLECTION Since initial crystallization attempts did not yield

diffraction-quality crystals, several methods were used to enhance crystallization. Deglycosylation was performed by adding Endo-Hf (1.0 × 106 U ml−1, New England Biolabs) 1:100 (v per v)

directly to the concentrated protein and incubating overnight at 37 °C. Completeness of the reaction was analysed by SDS-PAGE and sample quality was assessed by SAXS (see Table 2). As an

alternative approach, reductive lysine methylation was performed on glycosylated MAG1–5 diluted to 1 mg ml−1, by two steps of 2 h incubation at 4 °C with 1 M dimethylamine–borane complex

(added 1:50, v per v), and 1 M formaldehyde (added 1:25, v per v)53. The reaction was completed by a final addition of 1 M dimethylamine–borane complex (added 1:100, v per v) and incubated

overnight at 4 °C, after which the reaction was quenched by performing a SEC run on a Superdex200 (GE Healthcare) column equilibrated in 20 mM Tris/HCl pH 7.5, 200 mM NaCl. Therefore, this

was the buffer used to set-up crystallization experiments of the lysine-methylated protein. Sitting-drop vapour diffusion at 18 °C was used for all crystallization trials, by mixing 150 nl

of protein solution with 150 nl of reservoir solution. Crystals of deglycosylated MAG1–5 (6.7 mg ml−1) appeared in a condition of 100 mM NaCl, 20 mM Tris/HCl pH 7.0, 7.7% PEG 4,000 (w per

v). Crystals of lysine-methylated MAG1–5 (8.4 mg ml−1) appeared in a condition containing 200 mM NaOAc, 20% PEG3350 (w per v). Crystals of MAG1–3 (12.2 mg ml−1) appeared in a condition

containing 0.05 M tri-sodium citrate, 1.2 M ammonium sulfate, 3% (w per v) isopropanol. MAG1–3 crystals were soaked overnight by addition of 1 μl reservoir solution containing 10 mM

3′-N-acetylneuraminyl-N-acetyllactosamine (Neu5Ac-α2,3-Gal-β1,4-GlcNAc, Sigma-Aldrich product A6936) to the drop. Crystals were cryo-protected with reservoir solution supplemented with 25%

of glycerol for deglycosylated MAG1–5 and MAG1–3 and with 25% ethylene glycol for lysine-methylated MAG1–5. After brief incubation in the cryo-protectant solution, crystals were

plunge-cooled in liquid nitrogen. Data sets were collected at 100 K at the Deutsches Elektronen-Synchotron PETRA III beamline P14 (lysine-methylated MAG1–5, _λ_=0.97553 Å), the European

Synchotron Radiation Facility (ESRF) beamline ID23-1 (MAG1–3, _λ_=0.97599 Å) and Swiss Light Source beamline PX (deglycosylated MAG1–5 and soaked MAG1–3, _λ_=0.99998 Å). STRUCTURE SOLUTION

AND REFINEMENT Data were integrated by IMOSFLM54 (MAG1–5 deglycosylated, MAG1–5 lysine-methylated and MAG1–3 ligand bound) or XDS55 (MAG1–3 unliganded) and scaled and merged by the AIMLESS

pipeline56. All structures were solved by molecular replacement using PHASER57. Initial search models were PDB IDs 1URL (ref. 58) for Ig1, 4FRW (ref. 41) residues 150–242 for Ig2, 1CS6 (ref.

59) residues 308–388 for domain Ig3, 3P3Y (ref. 60) residues 55-185 for Ig4 and 2YD6 (ref. 61) residues 132–221 for Ig5. First MAG1–3 was solved by searching for two copies of 1URL,

followed by two copies of 1CS6 and finally two copies of 4FRW. Search models were trimmed to polyalanine chains by CHAINSAW (ref. 62) to obtain better starting density. Next, deglycosylated

MAG1–5 was solved by searching for the first two (refined) domains of MAG1–3, followed by SCULPTOR (ref. 63)-trimmed versions of 3P3Y and 2YD6 respectively. Lastly, the Ig3 domain was

searched using the refined Ig3 of MAG from MAG1–3 as this domain had a much higher B-factor in the MAG1–5 crystals, likely because of the lack of any crystal packing contacts for this

domain. It was important to realize that the unit cell only contained a single copy of MAG, resulting in a solvent content of 91%, which was used in PHASER to estimate the total scattering.

Lysine-methylated MAG1–5 was solved by searching for the Ig1, Ig2, Ig4, Ig5 and Ig3 from the refined MAG1–3 and deglycosylated MAG1–5 structures, in that order. Again, Ig3 was searched last

because of the high B-factor. After molecular replacement, models were improved by iterative density modification by DM (CCP4)64, manual model building in COOT65 and refinement with

REFMAC66. Final refinement was performed with PHENIX67 and validation with Molprobity68. In both MAG1–5 structures, Ig3 from the higher resolution MAG1–3 structure was used as a reference

structure to restrain refinement, because of the high B-factor and the resulting poor density for that domain. Ramachandran statistics were (% Ramachandran favored/% allowed/% outliers):

93/7/0 (MAG1–5 deglycosylated), 91/9/0 (MAG1–5 lysine-methylated), 97/3/0 (MAG1–3 unliganded) and 96/4/0 (MAG1–3 ligand bound). The buried surface are for MAG dimerization was calculated by

PISA69. SMALL ANGLE X-RAY SCATTERING SAXS was performed at the ESRF BM29 BioSAXS beamline equipped with a 2D Pilatus 1 M detector (DECTRIS, Switzerland), operated at an energy of 12.5 keV.

MAG constructs were diluted with and dialyzed against SEC buffer using a 10 kDa MWCO membrane. The concentrations were determined by ultraviolet–visible spectroscopy on a nanodrop ND-1,000

spectrophotometer. Similar concentrations were selected for all samples to allow comparison (see Table 2). SAXS data were collected at 20 °C unless indicated otherwise. The data were

radially averaged, normalized to the intensity of the transmitted beam and exposure time and the scattering of the solvent-blank (SEC buffer) was subtracted, following standard procedures.

The curve was scaled using a BSA reference so that the _I_0 represents the _M_m of the sample. Radiation damage was monitored by comparing curves collected from the same sample, only curves

without radiation damage were merged. A single concentration was used for all measurements, no extrapolation to zero concentration was performed. Data were analysed by PRIMUS70, GNOM71,

DAMMIF72, CRYSOL73 and OLIGOMER70 of the ATSAS74 suite. ANALYTICAL ULTRACENTRIFUGATION Sedimentation equilibrium experiments were carried out in a Beckman Coulter Proteomelab XL-I and a

Beckman Optima XL-A analytical ultracentrifuge. Either 12 or 3 mm centerpieces with quartz windows were used, 12 mm for the lowest concentrations and 3 mm for the others. An-60 and An-50 Ti

rotors (Beckman) were used to carry out the measurements. MAG constructs were diluted with and dialyzed against SEC buffer using a 10 kDa MWCO membrane. Input concentrations of 3.8, 35.3 and

89.0 μM (MAG1–5 wt glycosylated), 16.8, 27.2, 35.5 and 144 μM (MAG1–5 wt deglycosylated), 3.6, 8.3, 41.4 and 102.8 μM (MAG1–5 I473E glycosylated), 7.2, 57.4 and 332 μM (MAG1–3 wt

glycosylated) were used. Sedimentation equilibrium runs were performed at 20 °C and at 7,500, 14,000 and 20,000 r.p.m. Absorbance was determined at 250, 280 and 300 nm with SEC buffer as

reference. Buffer density and viscosity were determined by SEDNTERP as 0.99823, g ml−1 and 0.001002 Pa s, respectively. IN-GEL DIGESTION AND LC-MS/MS MAG1–5 was separated by SDS-PAGE, and

cut from gel for digestion with trypsin (Promega). The gel band was cut to small pieces, washed in Milli-Q water and treated with acetonitrile to shrink the gel pieces. The sample was then

incubated in 1 g l−1 1,4-dithiothreitol for 60 min at 60 °C, treated with acetonitrile, alkylated with 10 g l−1 iodoacetamide for 30 min at room temperature in the dark and subsequently

washed with ammonium bicarbonate and treated with acetonitrile, twice. The gel pieces were then incubated on ice for 90 min with 30 mg l−1 trypsin. Excess trypsin was removed, the gel pieces

were covered in ammonium bicarbonate, and the samples were subsequently incubated overnight at 37 °C. The digested samples were collected and remaining sample was extracted from the gel

pieces by treatment with acetonitrile. The solution with the peptides was subsequently dried in a speedvac and the peptides resuspended in 10% formic acid, 5% dimethylsulfoxide in water.

Peptides were separated by reversed phase LC coupled on-line to an Orbitrap Elite for MS/MS analysis. The nano-LC consists of an Agilent 1200 series LC system equipped with a 20 mm ReproSil-

Pur C18-AQ (Dr Maisch GmbH) trapping column (packed in-house, i.d., 100 μm; resin, 5 μm) and a 400 mm ReproSil-Pur C18-AQ (Dr Maisch GmbH) analytical column (packed in-house, i.d., 50 μm;

resin, 3 μm) arranged in a vented-column configuration. The flow was passively split to 100 nl min−1. We used a standard 45 min gradient from 7–30% acetonitrile in aqueous 0.1% formic acid.

All precursors were fragmented by both ETcaD and HCD. Data were searched against a custom database of recombinant protein sequences, including the MAG constructs used here, with trypsin as

protease, allowing up to two missed cleavages. We used a 50 p.p.m. precursor mass window and 0.02 Da fragment mass window. The C-mannosylated peptide, with 3+ precursor charge, eluted after

33–34 min and was identified by both HCD and ETcaD MS/MS, with matched fragment ions supporting site localization for the C-mannosylation. NATIVE MASS SPECTROMETRY Purified protein samples

were deglycosylated with Endo-Hf as for crystallization, followed by buffer exchange to 150 mM ammonium acetate pH 7.5, using Vivaspin500 10 kDa MWCO centrifugal filter units. Samples were

loaded onto gold-coated borosilicate capillaries prepared in-house for nanoelectrospray ionization. Samples were analysed on a modified Orbitrap extended mass range (Thermo Fisher) for high

mass ions75. SURFACE PLASMON RESONANCE MAG1–5 wt and mutants cloned in-frame with an N-terminal cystatin secretion signal and a C-terminal biotin acceptor peptide and His6-tag (sequence

AAAGSGLNDIFEAQKIEWHEGRTKHHHHHH), were biotinylated in HEK293 cells by co-transfection with _E. coli_ BirA biotin ligase with a sub-optimal secretion signal (in a pUPE5.02 vector), using a

DNA ratio of 9:1 (MAG DNA: BirA DNA, m per m)). Additional sterile biotin (100 μl of 10 mg ml−1 Tris-buffered biotin per 4 ml HEK293 culture) was supplemented to the medium a few hours after

transfection. MAG mutants were purified from the medium by Ni–NTA affinity purification. Purity was evaluated by SDS-PAGE and coomassie staining and biotinylation by a streptavidin

gel-shift assay followed by α-His6 Western blot (Supplementary Fig. 19). C-terminally biotinylated MAG proteins were spotted on a G-STREP SensEye chip (Ssens) with a Continuous Flow

Microspotter (Wasatch Microfluidics) using a 8 × 6 format. The C-terminal coupling of MAG to the surface mimics the native, membrane attached topology. SEC buffer with 0.005% tween was used

as a spotting buffer and the coupling was quenched using 1 mM biotin in SEC buffer solution. GT1b ganglioside liposomes were prepared as described previously76. In brief, the lipids

dipalmitoyl phosphatidylcholine, dipalmitoyl phosphatidylglycerol, cholesterol and GT1b gangliosides were mixed in a molar ratio of 40.3:4.2:40.9:1.3 in a chloroform/methanol mixture (6:1, v

per v). The lipid mixture was dried under vacuum on a rotary evaporator to create a thin film of lipids. Liposomes were formed by addition of 1 ml of SEC buffer per 21.7 μmol of lipid

mixture and 11 freeze-thaw cycles. As a negative control, liposomes with the same lipid composition but lacking GT1b were prepared using the same protocol. Liposomes were extruded through a

100 nm membrane with a mini-extruder (Avanti Polar Lipids) at 70 °C. SPR experiments, with the liposomes in the mobile phase and the MAG constructs attached to the surface, were performed on

an MX96 SPRi instrument (IBIS Technologies) equipped with a CX flowcell, using the CX vesicle run protocol and an association time of 60 min at a temperature of 25 °C. As a running buffer,

SEC buffer without any detergent was used. Preliminary removal of co-purified ligands before the runs and regeneration after runs was performed by washes with 0.5% SDS in phosphate-buffered

saline (PBS) followed by 5M NaCl. Data was zeroed and referenced using SprintX 1.11 (IBIS Technologies). ANIMALS All animal use and care was in accordance with institutional guidelines and

approved by the animal experimentation committee (DEC). Littermate C57BL/6 (Charles River) mice were killed by decapitation at postnatal day 1 (P1) before the brain was removed to prepare

hippocampal neuronal cultures. NEURITE OUTGROWTH ASSAYS MAG1–5-Fc constructs were generated by subcloning into pUPE7.12 vector using BamHI/NotI restriction sites (Fc is C-terminal of MAG),

expressed in HEK293 cells and purified by protein-A affinity purification using standard protocols. Coverslips were all coated overnight at 4 °C with PLL(20 μg ml−1) and 2 h at 37 °C with

different MAG proteins (60 μg ml−1): PBS (PLL control), MAG1–5 wt, MAG1–5 I473E, MAG1–5 N406Q, MAG1–3 wt, MAG1–5-Fc wt or MAG1–5-Fc R118A (at least three independent experiments were

performed). It is expected that proteins immobilize non-specifically to the coverslips, therefore the immobilization efficiency of the different MAG variants was not experimentally verified.

Hippocampal cultures were prepared as described previously77. In brief, the hippocampus was dissected at P1 and collected in L15 dissection medium (Gibco) containing 7 mM HEPES (L15-HEPES).

Cells were dissociated by incubation in 0.25% trypsin in L15-HEPES at 37 °C for 20 min. Following three washes with L15-HEPES, cells were triturated using a fire-polished Pasteur pipette in

growth medium (neurobasal medium (Gibco) with B-27 supplement (Thermo-Fisher), L-glutamine, penicillin/streptomycin and β-mercaptoethanol). Hippocampal neurons were resuspended in fresh

growth medium and plated onto PLL- and MAG-coated glass coverslips. After two days _in vitro_ (DIV2), hippocampal cultures were fixed for 10 min in 4% paraformaldehyde followed by three

washing steps with PBS, blocking with 5% normal donkey serum and 0.1% triton in PBS, and incubated overnight at 4 °C with mouse anti-βIII-tubulin primary antibody (Covance, 1:500 (v per v)).

The next day, cultures were washed three times with PBS and incubated with the secondary antibody (donkey anti-mouse Alexa Fluor 488, 1:750) for 2 h at room temperature. After three PBS

washing steps, coverslips were incubated with 4′,6-diamidino-2-phenylindole (DAPI—Sigma) for 10 min. After several PBS washes, coverslips were mounted with FluorSave. Hippocampal neurons

were visualized using a Zeiss Axioskop A1 using a 20 × objective. Images were analysed using FIJI (version 2.0.0)78 by tracing the longest neurite of a hippocampal neuron (positive for

tubulin and DAPI). Significance was determined using a one-way ANOVA (_P_=0.0001) with a Bonferroni posthoc test for multiple comparisons (Graphpad Prism version 6.07). Every construct was

compared with control PLL or MAG wt. All quantitative assessments in this manuscript were performed while being unaware of the condition to avoid observer bias. During analysis, raw data

were named in a descriptive way, without revealing experimental group information. Fluorescent microscopic imaging was always done with the same settings within experiments, and analysing

techniques were standardized. A single person analysed all data obtained within the experiments. Significant levels were reached for the following proteins: MAG1–5 wt (**, 348 neurons

counted), MAG1–5 N406Q (*, 385 neurons counted), MAG1–5-Fc wt (**, 364 neurons counted) and MAG1–5-Fc R118A (*, 369 neurons counted) compared with PLL (395 neurons counted). Constructs

MAG1–5 I473E (*, 342 neurons counted), MAG1–3 wt (***, 353 neurons counted) and MAG1–5-Fc R118A (****, 348 neurons counted) were significant compared with MAG1–5 wt. *_P_<0.05,

**_P_<0.01, ***_P_<0.001, ****_P_<0.0001. DATA AVAILABILITY Coordinates and structure factors for MAG1–5 deglycosylated (spacegroup _P_322), MAG1–5 methylated (spacegroup _P_6522),

MAG1–3 (spacegroup _P_1) and MAG1–3 ligand bound (spacegroup _P_1) have been deposited in the Protein Data Bank with succession numbers 5LF5, 5LFU, 5LFR and 5LFV, respectively. All SAXS data

is made available at the small angle scattering databank (SASBDB) with the accession codes SASDB55 (glycosylated MAG1–5 wt), SASDB26 (glycosylated MAG1–5 I473E), SASDB36 (glycosylated

MAG1–5 N406Q), SASDB46 (glycosylated MAG1–3), SASDBF6 (deglycosylated MAG1–5 wt) SASDB56 (deglycosylated MAG1–5 I473E), SASDB66 (deglycosylated MAG1–5 N406Q) and SASDB76 (deglycosylated

MAG1–3). ADDITIONAL INFORMATION HOW TO CITE THIS ARTICLE: Pronker, M. F. _et al_. Structural basis of Myelin-associated glycoprotein adhesion and signalling. _Nat. Commun._ 7, 13584 doi:

10.1038/ncomms13584 (2016). PUBLISHER’S NOTE: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. REFERENCES * Simons, M.

& Nave, K. Oligodendrocytes: myelination and axonal support. _Cold Spring Harb. Perspect. Biol._ 8, a020479 (2015). Article  PubMed  Google Scholar  * Quarles, R. H. Myelin-associated

glycoprotein (MAG): past, present and beyond. _J. Neurochem._ 100, 1431–1448 (2007). CAS  PubMed  Google Scholar  * Lopez, P. H. H. Role of myelin-associated glycoprotein (siglec-4a) in the

nervous system. _Adv. Neurobiol._ 9, 245–262 (2014). Article  PubMed  Google Scholar  * Novarino, G. et al. Exome sequencing links corticospinal motor neuron disease to common

neurodegenerative disorders. _Science_ 343, 506–511 (2014). Article  CAS  ADS  PubMed  PubMed Central  Google Scholar  * Lossos, A. et al. Myelin-associated glycoprotein gene mutation causes

Pelizaeus-Merzbacher disease-like disorder. _Brain_ 138, 2521–2536 (2015). Article  PubMed  PubMed Central  Google Scholar  * Braun, P. E., Frail, D. E. & Latov, N. Myelin-associated

glycoprotein is the antigen for a monoclonal IgM in polyneuropathy. _J. Neurochem._ 39, 1261–1265 (1982). Article  CAS  PubMed  Google Scholar  * Poltorak, M., Steck, A. J. & Schachner,

M. Reactivity with neural cell adhesion molecules in sera from patients with demyelinating diseases. _Neurosci. Lett._ 65, 199–203 (1986). Article  CAS  PubMed  Google Scholar  * Wajgt, A.

& Górny, M. CSF antibodies to myelin basic protein and to myelin-associated glycoprotein in multiple sclerosis. Evidence of the intrathecal production of antibodies. _Acta Neurol.

Scand._ 68, 337–343 (1983). Article  CAS  PubMed  Google Scholar  * Jahn, O., Tenzer, S. & Werner, H. B. Myelin proteomics: molecular anatomy of an insulating sheath. _Mol. Neurobiol._

40, 55–72 (2009). Article  CAS  PubMed  PubMed Central  Google Scholar  * Trapp, B. D. & Quarles, R. H. Presence of the myelin-associated glycoprotein correlates with alterations in the

periodicity of peripheral myelin. _J. Cell Biol._ 92, 877–882 (1982). Article  CAS  PubMed  Google Scholar  * Trapp, B. D., Quarles, R. H. & Suzuki, K. Immunocytochemical studies of

quaking mice support a role for the myelin-associated glycoprotein in forming and maintaining the periaxonal space and periaxonal cytoplasmic collar of myelinating Schwann cells. _J. Cell

Biol._ 99, 594–606 (1984). Article  CAS  PubMed  Google Scholar  * Li, C. et al. Myelination in the absence of myelin-associated glycoprotein. _Nature_ 369, 747–750 (1994). Article  CAS  ADS

  PubMed  Google Scholar  * Pan, B. et al. Myelin-associated glycoprotein and complementary axonal ligands, gangliosides, mediate axon stability in the CNS and PNS: neuropathology and

behavioral deficits in single- and double-null mice. _Exp. Neurol._ 195, 208–217 (2005). Article  CAS  PubMed  PubMed Central  Google Scholar  * Spiegel, I. et al. A central role for Necl4

(SynCAM4) in Schwann cell–axon interaction and myelination. _Nat. Neurosci._ 10, 861–869 (2007). Article  CAS  PubMed  PubMed Central  Google Scholar  * Maurel, P. et al. Nectin-like

proteins mediate axon-Schwann cell interactions along the internode and are essential for myelination. _J. Cell Biol._ 178, 861–874 (2007). Article  CAS  PubMed  PubMed Central  Google

Scholar  * Zhu, Y. et al. Necl-4/SynCAM-4 is expressed in myelinating oligodendrocytes but not required for axonal myelination. _PLoS ONE_ 8, 4–11 (2013). Google Scholar  * Lehmann, F.,

Gäthje, H., Kelm, S. & Dietz, F. Evolution of sialic acid-binding proteins: molecular cloning and expression of fish siglec-4. _Glycobiology_ 14, 959–968 (2004). Article  CAS  PubMed 

Google Scholar  * Kelm, S., Pelz, A., Schauer, R., Filbin, M. T. & Tang, S. Sialoadhesin, myelin-associated glycoprotein and CD22 define a new family of sialic acid-dependent adhesion

molecules of the immunoglobulin superfamily. _Curr. Biol._ 4, 965–972 (1994). Article  CAS  PubMed  Google Scholar  * Collins, B. E. et al. Sialic acid specificity of myelin-associated

glycoprotein binding. _J. Biol. Chem._ 272, 1248–1255 (1997). Article  CAS  PubMed  Google Scholar  * Vinson, M. et al. Myelin-associated glycoprotein interacts with ganglioside GT1b. A

mechanism for neurite outgrowth inhibition. _J. Biol. Chem._ 276, 20280–20285 (2001). Article  CAS  PubMed  Google Scholar  * McKerracher, L. et al. Identification of myelin-associated

glycoprotein as a major myelin-derived inhibitor of neurite growth. _Neuron_ 13, 805–811 (1994). Article  CAS  PubMed  Google Scholar  * Tang, S. et al. Myelin-associated glycoprotein

interacts with neurons via a sialic acid binding site at ARG118 and a distinct neurite inhibition site. _J. Cell Biol._ 138, 1355–1366 (1997). Article  CAS  PubMed  PubMed Central  Google

Scholar  * Umemori, H., Sato, S., Yagi, T., Aizawa, S. & Yamamoto, T. Initial events of myelination involve Fyn tyrosine kinase signalling. _Nature_ 367, 572–576 (1994). Article  CAS 

ADS  PubMed  Google Scholar  * Marta, C. B. et al. Myelin associated glycoprotein cross-linking triggers its partitioning into lipid rafts, specific signaling events and cytoskeletal

rearrangements in oligodendrocytes. _Neuron Glia Biol._ 1, 35–46 (2004). Article  CAS  PubMed  PubMed Central  Google Scholar  * Yamauchi, J. et al. Phosphorylation of cytohesin-1 by Fyn is

required for initiation of myelination and the extent of myelination during development. _Sci. Signal._ 5, ra69–ra69 (2012). Article  PubMed  Google Scholar  * Kursula, P., Meriläinen, G.,

Lehto, V. P. & Heape, A. M. The small myelin-associated glycoprotein is a zinc-binding protein. _J. Neurochem._ 73, 2110–2118 (1999). CAS  PubMed  Google Scholar  * Kursula, P., Lehto,

V. P. & Heape, A. M. The small myelin-associated glycoprotein binds to tubulin and microtubules. _Mol. Brain Res._ 87, 22–30 (2001). Article  CAS  PubMed  Google Scholar  * Fahrig, T. et

al. Functional topography of the myelin-associated glycoprotein. I. mapping of domains by electron microscopy. _Eur. J. Neurosci._ 5, 1118–1126 (1993). Article  CAS  PubMed  Google Scholar

  * Attia, J., Hicks, L., Oikawa, K., Kay, C. M. & Dunn, R. J. Structural properties of the myelin-associated glycoprotein ectodomain. _J. Neurochem._ 61, 718–726 (1993). Article  CAS 

PubMed  Google Scholar  * May, A. P., Robinson, R. C., Vinson, M., Crocker, P. R. & Jones, E. Y. Crystal structure of the N-terminal domain of sialoadhesin in complex with 3′

sialyllactose at 1.85A resolution. _Mol. Cell_ 1, 719–728 (1998). Article  CAS  PubMed  Google Scholar  * Alphey, M. S., Attrill, H., Crocker, P. R. & van Aalten, D. M. F. High

resolution crystal structures of Siglec-7. Insights into ligand specificity in the Siglec family. _J. Biol. Chem._ 278, 3372–3377 (2003). Article  CAS  PubMed  Google Scholar  * Zhuravleva,

M. a., Trandem, K. & Sun, P. D. Structural implications of Siglec-5-mediated sialoglycan recognition. _J. Mol. Biol._ 375, 437–447 (2008). Article  CAS  PubMed  Google Scholar  *

Pedraza, L., Owens, G. C., Green, L. a. & Salzer, J. L. The myelin-associated glycoproteins: membrane disposition, evidence of a novel disulfide linkage between immunoglobulin-like

domains, and posttranslational palmitylation. _J. Cell Biol._ 111, 2651–2661 (1990). Article  CAS  PubMed  Google Scholar  * Burger, D., Pidoux, L. & Steck, A. J. Identification of the

glycosylated sequons of human myelin-associated glycoprotein. _Biochem. Biophys. Res_ 197, 457–464 (1993). Article  CAS  Google Scholar  * Furmanek, A. & Hofsteenge, J. Protein

C-mannosylation: facts and questions. _Acta Biochim. Pol._ 47, 781–789 (2000). CAS  PubMed  Google Scholar  * de Beer, T., Vliegenthart, J. F., Löffler, A. & Hofsteenge, J. The

hexopyranosyl residue that is C-glycosidically linked to the side chain of tryptophan-7 in human RNase Us is alpha-mannopyranose. _Biochemistry_ 34, 11785–11789 (1995). Article  CAS  PubMed

  Google Scholar  * Sedzik, J., Jastrzebski, J. P. & Grandis, M. Glycans of myelin proteins. _J. Neurosci. Res._ 93, 1–18 (2015). Article  CAS  PubMed  Google Scholar  * McKerracher, L.

et al. Identification of myelin-associated glycoprotein as a major myelin-derived inhibitor of neurite growth. _Neuron_ 13, 805–811 (1994). Article  CAS  PubMed  Google Scholar  * Wu, Y.,

Vendome, J., Shapiro, L., Ben-Shaul, A. & Honig, B. Transforming binding affinities from three dimensions to two with application to cadherin clustering. _Nature_ 475, 510–513 (2011).

Article  CAS  PubMed  PubMed Central  Google Scholar  * Özkan, E. et al. Extracellular architecture of the SYG-1/SYG-2 adhesion complex instructs synaptogenesis. _Cell_ 156, 482–494 (2014).

Article  PubMed  PubMed Central  Google Scholar  * Harrison, O. J. et al. Nectin ectodomain structures reveal a canonical adhesive interface. _Nat. Struct. Mol. Biol._ 19, 906–915 (2012).

Article  CAS  PubMed  PubMed Central  Google Scholar  * Kinter, J. et al. An essential role of MAG in mediating axon-myelin attachment in Charcot-Marie-Tooth 1A disease. _Neurobiol. Dis._ 0,

221–231 (2013). Article  CAS  Google Scholar  * Snaidero, N. et al. Myelin membrane wrapping of CNS axons by PI(3,4,5)P3-dependent polarized growth at the inner tongue. _Cell_ 156, 277–290

(2014). Article  CAS  PubMed  PubMed Central  Google Scholar  * Quarles, R. H. Effects of pronase and neuraminidase treatment on a myelin-associated glycoprotein in developing brain.

_Biochem. J._ 156, 143–150 (1976). Article  CAS  PubMed  PubMed Central  Google Scholar  * Konat, G., Hogan, E. L., Leskawa, K. C., Gantt, G. & Singh, I. Abnormal glycosylation of

myelin-associated glycoprotein in quaking mouse brain. _Neurochem. Int._ 10, 555–558 (1987). Article  CAS  PubMed  Google Scholar  * Bartoszewicz, Z. P. et al. Abnormal expression and

glycosylation of the large and small isoforms of myelin-associated glycoprotein in dysmyelinating quaking mutants. _J. Neurosci. Res._ 41, 27–38 (1995). Article  CAS  PubMed  Google Scholar

  * Bartoszewicz, Z. P., Lauter, C. J. & Quarles, R. H. The myelin-associated glycoprotein of the peripheral nervous system in trembler mutants contains increased α2-3 sialic acid and

galactose. _J. Neurosci. Res._ 43, 587–593 (1996). Article  CAS  PubMed  Google Scholar  * Schnaar, R. L. & Lopez, P. H. H. Myelin-associated glycoprotein and its axonal receptors. _J.

Neurosci. Res._ 87, 3267–3276 (2009). Article  CAS  PubMed  PubMed Central  Google Scholar  * Cao, Z. et al. The inhibition site on myelin-associated glycoprotein is within Ig-domain 5 and

is distinct from the sialic acid binding site. _J. Neurosci._ 27, 9146–9154 (2007). Article  CAS  PubMed  PubMed Central  Google Scholar  * Wörter, V. et al. Inhibitory activity of

myelin-associated glycoprotein on sensory neurons is largely independent of NgR1 and NgR2 and resides within Ig-Like domains 4 and 5. _PLoS ONE_ 4, e5218 (2009). Article  ADS  PubMed  PubMed

Central  Google Scholar  * Chan, B. et al. SAP couples Fyn to SLAM immune receptors. _Nat. Cell Biol._ 5, 155–160 (2003). Article  CAS  PubMed  Google Scholar  * Vinson, M. et al. Lipid

rafts mediate the interaction between myelin-associated glycoprotein (MAG) on myelin and MAG-receptors on neurons. _Mol. Cell Neurosci._ 22, 344–352 (2003). Article  CAS  PubMed  Google

Scholar  * Walter, T. S. et al. Lysine methylation as a routine rescue strategy for protein crystallization. _Structure_ 14, 1617–1622 (2006). Article  CAS  PubMed  PubMed Central  Google

Scholar  * Battye, T. G. G., Kontogiannis, L., Johnson, O., Powell, H. R. & Leslie, A. G. W. iMOSFLM: a new graphical interface for diffraction-image processing with MOSFLM. _Acta

Crystallogr. Sect. D Biol. Crystallogr._ 67, 271–281 (2011). Article  CAS  Google Scholar  * Kabsch, W. Xds. _Acta Crystallogr. Sect. D Biol. Crystallogr._ 66, 125–132 (2010). Article  CAS 

Google Scholar  * Evans, P. R. & Murshudov, G. N. How good are my data and what is the resolution? _Acta Crystallogr. Sect. D Biol. Crystallogr._ 69, 1204–1214 (2013). Article  CAS 

Google Scholar  * McCoy, A. J. et al. Phaser crystallographic software. _J. Appl. Crystallogr._ 40, 658–674 (2007). Article  CAS  PubMed  PubMed Central  Google Scholar  * Bukrinsky, J. T.,

St. Hilaire, P. M., Meldal, M., Crocker, P. R. & Henriksen, A. Complex of sialoadhesin with a glycopeptide ligand. _Biochim. Biophys. Acta_ 1702, 173–179 (2004). Article  CAS  PubMed 

Google Scholar  * Freigang, J. et al. The crystal structure of the ligand binding module of axonin-1/TAG-1 suggests a zipper mechanism for neural cell adhesion. _Cell_ 101, 425–433 (2000).

Article  CAS  PubMed  Google Scholar  * Liu, H., Focia, P. J. & He, X. Homophilic adhesion mechanism of neurofascin, a member of the L1 family of neural cell adhesion molecules. _J.

Biol. Chem._ 286, 797–805 (2011). Article  CAS  PubMed  Google Scholar  * Coles, C. H. et al. Proteoglycan-specific molecular switch for RPTP σ clustering and neuronal extension. _Science_

332, 484–488 (2011). Article  CAS  ADS  PubMed  PubMed Central  Google Scholar  * Stein, N. CHAINSAW: a program for mutating pdb files used as templates in molecular replacement. _J. Appl.

Crystallogr._ 41, 641–643 (2008). Article  CAS  Google Scholar  * Bunkóczi, G. & Read, R. J. Improvement of molecular-replacement models with _Sculptor_. _Acta Crystallogr. Sect. D Biol.

Crystallogr._ 67, 303–312 (2011). Article  Google Scholar  * Cowtan, K. DM: an automated procedure for phase improvement by density modification. Joint CCP4 ESF-EACBM Newsletter Protein

Crystallography 31, 34–38 (1994). * Emsley, P. & Cowtan, K. Coot: model-building tools for molecular graphics. _Acta Crystallogr. Sect. D Biol. Crystallogr._ 60, 2126–2132 (2004).

Article  Google Scholar  * Murshudov, G. N. et al. REFMAC5 for the refinement of macromolecular crystal structures. _Acta Crystallogr. Sect. D Biol. Crystallogr._ 67, 355–367 (2011). Article

  CAS  Google Scholar  * Afonine, P. V. et al. Towards automated crystallographic structure refinement with phenix.refine. _Acta Crystallogr. Sect. D Biol. Crystallogr._ 68, 352–367 (2012).

Article  CAS  Google Scholar  * Chen, V. B. et al. MolProbity: all-atom structure validation for macromolecular crystallography. _Acta Crystallogr. Sect. D Biol. Crystallogr._ 66, 12–21

(2010). Article  CAS  Google Scholar  * Krissinel, E. & Henrick, K. Inference of macromolecular assemblies from crystalline state. _J. Mol. Biol._ 372, 774–797 (2007). Article  CAS 

PubMed  Google Scholar  * Konarev, P. V., Volkov, V. V., Sokolova, A. V., Koch, M. H. J. & Svergun, D. I. PRIMUS: a Windows PC-based system for small-angle scattering data analysis. _J.

Appl. Crystallogr._ 36, 1277–1282 (2003). Article  CAS  Google Scholar  * Svergun, D. I. Determination of the regularization parameter in indirect-transform methods using perceptual

criteria. _J. Appl. Crystallogr._ 25, 495–503 (1992). Article  CAS  Google Scholar  * Franke, D. & Svergun, D. I. DAMMIF, a program for rapid ab-initio shape determination in small-angle

scattering. _J. Appl. Crystallogr._ 42, 342–346 (2009). Article  CAS  PubMed  PubMed Central  Google Scholar  * Svergun, D., Barberato, C. & Koch, M. H. J. CRYSOL - a program to

evaluate X-ray solution scattering of biological macromolecules from atomic coordinates. _J. Appl. Crystallogr._ 28, 768–773 (1995). Article  CAS  Google Scholar  * Petoukhov, M. V. et al.

New developments in the ATSAS program package for small-angle scattering data analysis. _J. Appl. Crystallogr._ 45, 342–350 (2012). Article  CAS  PubMed  PubMed Central  Google Scholar  *

Rose, R. J., Damoc, E., Denisov, E., Makarov, A. & Heck, A. J. R. High-sensitivity Orbitrap mass analysis of intact macromolecular assemblies. _Nat. Methods_ 9, 1084–1086 (2012). Article

  CAS  PubMed  Google Scholar  * Ahn-Yoon, S., Decory, T. R. & Durst, R. a. Ganglioside-liposome immunoassay for the detection of botulinum toxin. _Anal. Bioanal. Chem._ 378, 68–75

(2004). Article  CAS  PubMed  Google Scholar  * Van Battum, E. Y. et al. The intracellular redox protein MICAL-1 regulates the development of hippocampal mossy fibre connections. _Nat.

Commun._ 5, 4317 (2014). Article  CAS  PubMed  Google Scholar  * Schindelin, J. et al. Fiji: an open-source platform for biological-image analysis. _Nat. Methods_ 9, 676–682 (2012). Article

  CAS  PubMed  Google Scholar  Download references ACKNOWLEDGEMENTS We thank the staff of the ESRF beamline ID23-1, Swiss Light Source beamline PX and Deutsches Elektronen-Synchotron

beamline PETRA III P14 for assistance with diffraction data collection and the staff of ESRF beamline BM29 for SAXS measurements. We thank Mike F. Renne and Jonas M. Dörr for assistance in

preparing the GT1b liposomes used in the SPR experiment. This work was funded by a Vidi grant (723.012.002) from the Netherlands Organization for Scientific Research (NWO) to B.J.C.J. and an

Investment Grant NWO Medium (721.012.004). J.S. and A.J.R.H. are supported by Proteins@Work (project number 184.032.201), and by the Gravity Program Institute for Chemical Immunology, both

funded by NWO. S.L. and R.J.P. are funded by an ALW-Vici Grant from NWO, Stichting ParkingsonFonds and Dynamics of Youth Seed Money grant from Utrecht University. AUTHOR INFORMATION AUTHORS

AND AFFILIATIONS * Department of Chemistry, Crystal and Structural Chemistry, Bijvoet Center for Biomolecular Research, Faculty of Science, Utrecht University, Padualaan 8, Utrecht, 3584 CH,

The Netherlands Matti F. Pronker & Bert J. C. Janssen * Department for Translational Neuroscience, Brain Center Rudolf Magnus, University Medical Center Utrecht, Universiteitsweg 100,

Utrecht, 3584 CG, The Netherlands Suzanne Lemstra & R. Jeroen Pasterkamp * Department of Chemistry and Department of Pharmaceutical Sciences, Biomolecular Mass Spectrometry and

Proteomics, Bijvoet Center for Biomolecular Research and Utrecht Institute for Pharmaceutical Sciences, Faculty of Science, Utrecht University, Padualaan 8, Utrecht, 3584 CH, The Netherlands

Joost Snijder & Albert J. R. Heck * Department of Chemistry, Van’t Hoff Laboratory for Physical and Colloid Chemistry, Debye Institute of Nanomaterials Science, Faculty of Science,

Utrecht University, Padualaan 8, Utrecht, 3584 CH, The Netherlands Dominique M. E. Thies-Weesie Authors * Matti F. Pronker View author publications You can also search for this author

inPubMed Google Scholar * Suzanne Lemstra View author publications You can also search for this author inPubMed Google Scholar * Joost Snijder View author publications You can also search

for this author inPubMed Google Scholar * Albert J. R. Heck View author publications You can also search for this author inPubMed Google Scholar * Dominique M. E. Thies-Weesie View author

publications You can also search for this author inPubMed Google Scholar * R. Jeroen Pasterkamp View author publications You can also search for this author inPubMed Google Scholar * Bert J.

C. Janssen View author publications You can also search for this author inPubMed Google Scholar CONTRIBUTIONS M.F.P. and B.J.C.J. designed the experiments. M.F.P. generated constructs,

purified proteins and did all the structural biology (SAXS and X-ray diffraction). M.F.P. and D.M.E.T.-W. performed SE-AUC experiments and analysed the data. J.S. performed mass spectrometry

experiments and analysed the data together with A.J.R.H.; S.L. performed neurite outgrowth experiments and analysed the data together with R.J.P.; B.J.C.J. supervised the project. M.F.P.

and B.J.C.J. wrote the manuscript with input from all authors. CORRESPONDING AUTHOR Correspondence to Bert J. C. Janssen. ETHICS DECLARATIONS COMPETING INTERESTS The authors declare no

competing financial interests. SUPPLEMENTARY INFORMATION SUPPLEMENTARY INFORMATION Supplementary Figures 1-19, Supplementary Note 1 and Supplementary References (PDF 12745 kb) PEER REVIEW

FILE (PDF 235 KB) RIGHTS AND PERMISSIONS This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are

included in the article’s Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to

obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/ Reprints and permissions ABOUT THIS

ARTICLE CITE THIS ARTICLE Pronker, M., Lemstra, S., Snijder, J. _et al._ Structural basis of myelin-associated glycoprotein adhesion and signalling. _Nat Commun_ 7, 13584 (2016).

https://doi.org/10.1038/ncomms13584 Download citation * Received: 27 July 2016 * Accepted: 17 October 2016 * Published: 06 December 2016 * DOI: https://doi.org/10.1038/ncomms13584 SHARE THIS

ARTICLE Anyone you share the following link with will be able to read this content: Get shareable link Sorry, a shareable link is not currently available for this article. Copy to clipboard

Provided by the Springer Nature SharedIt content-sharing initiative