Nidogens-1 and -2 are similar in domain structure but differ in sequence (46% amino acid similarity in human, 43% in mouse 293037). To examine their distribution we sought isoform-specific antibodies. Equal quantities of full-length recombinant nidogen-1 and -2 were separated electrophoretically and immunoblotted with several commercially available antibodies. Antibodies were identified that reacted strongly and selectively with each nidogen isoform (Figure 1A, B). These antibodies were then used to localize nidogens in muscle.

As shown previously 32, nidogen-1 was associated with surfaces of adult muscle fibers whereas little if any nidogen-2 was detectable on most of the muscle fiber surface (Figure 1C, D). Nidogen-2 was, however, present on many small structures associated with muscle fibers. Most of these were capillaries that lie between muscle fibers (see below) but a minority resembled synaptic sites in size, shape and position. We therefore double-labelled muscle sections with nidogen antibodies plus a fluorescent derivative of α-bungarotoxin, which binds specifically to the acetylcholine receptors (AChRs) concentrated in the postsynaptic membrane at the NMJ. This demonstrated that synaptic sites did indeed contain nidogen-2 (Figure 1E).

To localize nidogen-2 within the NMJ, we examined cross- and en face-sectioned NMJs, allowing us to distinguish three distinct domains in BL near synaptic sites: synaptic BL, which occupies the synaptic cleft; extrasynaptic BL, in directly adjoining stretches of the muscle fiber surface; and Schwann cell BL, atop Schwann cell processes that themselves cap nerve terminals (Figure 2A). Nidogen-1 was present in all three of these domains, whereas nidogen-2 was present in both synaptic and Schwann cell BLs but absent from extrasynaptic BL (Figures 2B–D).

On the basis of these findings, we investigated the relationship between nidogen-2 and the antigen recognized by the monoclonal antibody, 9H6, described by Chiu and Ko 38. 9H6 selectively stains NMJs in rat and binds a carbohydrate-dependent epitope on a nidogen-1-like antigen. This antibody was described prior to the discovery of nidogen-2, so it seemed possible that it recognized nidogen-2. Because 9H6 fails to cross-react with mouse tissue, we stained rat muscle with anti-nidogen-2. In rats, as in mice, nidogen-2 was present at NMJs and in capillaries (data not shown), whereas 9H6 stained only NMJs 38. Other differences in staining pattern are discussed below. Moreover, on immunoblots, 9H6 recognizes a protein with a molecular weight of 150 kDa 38, which is the expected size of mammalian nidogen-1 and considerably smaller than nidogen-2 (Figure 1A, B). These differences argue that the 9H6-antigen is not nidogen-2. Thus, synaptic BL may contain both a synapse-specific isoform of nidogen (nidogen-2) and a unique, glycosylated form of nidogen-1.

Synaptic BL can be deposited by muscles fibers, motor nerve terminals or Schwann cells 4539. Both muscle and Schwann cells have been reported to express nidogen-2 4041. To determine whether muscle cells can contribute nidogen-2 to the synaptic cleft, we used a muscle cell line, C2C12. These cells were fused into myotubes on laminin substrata to promote AChR clustering 42. Nidogen-2 was enriched at these AChR-rich sites even though no non-muscle cells were present in the cultures (Figure 2E). Thus, muscle cells are capable of nidogen-2 synthesis, secretion, and accumulation at synaptic sites.

As noted in the introduction, synaptic isoforms have been demonstrated for three of the families of core BL components – laminins, collagens IV and HSPGs 45. Our results add the fourth, nidogen, to this list. The presence of synapse-specific isoforms of common BL components provides a molecular explanation for the observations that synaptic and extrasynaptic BLs are structurally similar and physically continuous but functionally distinct.

Next, we assessed the distribution of nidogen-1 and -2 in the BLs of peripheral nerves. The BL of the perineurium, which surrounds entire axon fascicles, was rich in both nidogens-1 and -2, whereas the BL of the endoneurium, which surrounds individual non-myelinating or myelinating Schwann cells, was rich in nidogen-1 but poor in nidogen-2 (Figure 3A, B). In contrast, antibody 9H6 labels endoneurial but not perineurial BL in rat 38, consistent with the idea that it recognizes a subset of nidogen-1 molecules but not nidogen-2.

A few structures within the nerve fascicle were rich in nidogen-2 (arrowheads in Figure 3B). Although some of these may be capillaries, most were not associated with CD31/PECAM-1, an endothelial cell marker (43 and data not shown). Instead, they are likely to be the BL surrounding non-myelinating Schwann cells and their associated bundle of small-caliber, non-myelinated axons 41. This localization is consistent with the observation that nidogen-2 is present in the BL of terminal Schwann cells (Figure 2C), which are non-myelinating.

Nidogen-1 and -2 were also differentially distributed in BLs of muscle spindles, sensory organs that respond to muscle stretch. Each spindle contains several specialized intrafusal muscle fibers, all of which are surrounded by a capsule BL. Intrafusal fiber BL was rich in nidogen-1, but poor in nidogen-2, whereas spindle capsule BL was rich in nidogen-2 but poor in nidogen-1 (Figure 3C, D). In contrast, BLs of intramuscular capillaries, venules, and arterioles all contained both nidogen proteins (Figure 3E–H and data not shown). To our knowledge, the spindle capsule is the only BL found to date that bears nidogen-2 but not nidogen-1.

The distribution of some laminins, collagens IV, and HSPGs in muscle fiber BL changes as development proceeds 9111244. We asked whether the same was true of nidogens. At birth, both nidogen-1 and nidogen-2 were present throughout both synaptic and extrasynaptic muscle fiber BL (Figure 4A, D). Similar patterns were present through the first postnatal week (data not shown). Moreover, levels of nidogen-1 in synaptic and extrasynaptic BL remained similar into adulthood (Figures 4B, C). In contrast, levels of nidogen-2 decreased in extrasynaptic BL during the second postnatal week, so synaptic and Schwann cell BL contained more nidogen-2 than extrasynaptic BL by postnatal day (P) 14 (Figure 4E). By P21, nidogen-2 was undetectable in extrasynaptic BL (Figure 4F). Thus, nidogen-2 becomes restricted to synaptic sites as the NMJ matures. Because our methods are not quantitative, we do not know whether the decreased abundance of nidogen-2 in extrasynaptic BL is accompanied by an increased abundance in synaptic and Schwann cell BLs.

We asked whether the time course with which nidogen-2 became restricted to the NMJ was similar to that of laminins α4 and α5, which are present extrasynaptically at birth but restricted to synaptic BL in adults 9. In fact, the α4 and α5 laminin chains, like nidogen-2, were present throughout muscle fiber BL in neonates, were present at markedly higher levels synaptically than extrasynaptically by P14, and were largely synapse-specific by P21 (Figure 4G–L). Agrin is lost from extrasynaptic BL with a similar time course 44. Thus, there may be a coordinated alteration in the composition of muscle fiber BL as development proceeds, similar to, but later than, the initial broad distribution and eventual synaptic concentration of AChRs 39. In contrast, laminin β2 and synaptic collagen IV chains are synaptically concentrated from their initial appearance during embryogenesis and during the third postnatal week, respectively 912. Thus, whereas the maturation of extrasynaptic BL may occur in a concerted fashion, the distinctions between synaptic and extrasynaptic BLs arise in a series of multiple steps.

Targeted nidogen-2 null mutant mice are viable and fertile, and no structural or functional defects have been detected in them to date 3336. To seek abnormalities in NMJs of nidogen-2 mutants, we labelled whole mounts of diaphragm muscles from young adults (P56) with markers of pre- and postsynaptic specializations, anti-synaptotagmin 2 (a synaptic vesicle protein) and α-bungarotoxin, respectively, then imaged NMJs by confocal microscopy. Although nerve terminals and postsynaptic membranes were closely apposed to each other, and obviously functional, the topology of the synapse was abnormal in mutants. In controls, NMJs appear 'pretzel-like', with continuous, branched AChR clusters (Figure 5A). In mutants, many NMJs appeared fragmented, with AChRs clustered into small, separate islands (Figure 5B, D, E). Others were plaque-like (Figure 5C), a shape characteristic of neonatal NMJs 45. Thus, although nidogen-2-deficient NMJs are functional, their structure is aberrant.

Neuromuscular abnormalities observed in nidogen-2 mutants could reflect a role of nidogen-2 in the formation, maturation or maintenance of the NMJ. To distinguish these possibilities, we examined mutant muscles at P7, when NMJs are quite immature, and at P21, soon after the early postnatal period of synapse elimination is complete 39. No obvious defects were observed in nidogen-2 mutants at either age (Figure 6A, B). Thus, nidogen-2 appears to be dispensable for formation and remodelling of the NMJ, but required for its full maturation or maintenance. To ask whether the defects are progressive, we examined diaphragms from 1-year-old nidogen-2 mutants. Defects were not appreciably more severe in 12-month-old mutants than in 2-month-old mutants. Thus, nidogen-2 is required for final stages in the maturation of the NMJ.

To test whether defects in NMJ morphology were secondary to muscle damage, we searched for muscle fiber degeneration in nidogen-2-deficient muscle. In healthy muscle, myonuclei are concentrated in the periphery of muscle fibers. In fibers that have undergone degeneration and regeneration, however, myonuclei are centrally located 46. Less than 4% of mutant muscle fibers contained central nuclei (data not shown), whereas up to 80% of NMJs were abnormal (see below). Thus, defects in synaptic structure in nidogen-2-deficient mice were not the result of degenerating muscle fibers.

How might nidogen-2 promote maturation or enhance maintenance of the NMJ? Nidogens are capable of binding various cell surface receptors that are both present in pre- or postsynaptic membranes and necessary for synaptogenesis; these include integrins and leukocyte-common antigen related (LAR) receptor tyrosine phosphatase 3373947. Therefore, one possible explanation for the defects in nidogen-2-deficient NMJs is that nidogen-2 exerts direct effects by binding receptors on either the motor nerve terminal or the postsynaptic apparatus. To test this idea, we used assays previously applied to identify and characterize other synaptic organizing molecules. For presynaptic differentiation, we applied soluble recombinant nidogen-1 or -2 to cultured chick motor neurons, and grew motoneurons on substrates coated with a mixture of recombinant nidogen and laminin-111. In both cases, we assayed the ability of nidogen to promote clustering of synaptic vesicles into aggregates such as those found in nerve terminals 4849. We also asked whether nidogens affected the length or branching of motor neurites. Neither nidogen-1 nor -2 detectably affected motor neurons under these conditions. For postsynaptic differentiation, we assayed the aggregation of AChRs in cultured myotubes 142042. In these assays, nidogen appeared to be detrimental to the health of the myotubes, so it was not possible to gauge their synaptic effects. Thus, these in vitro studies provide no evidence for a direct effect of nidogen on nerve or muscle, although we cannot draw definitive conclusions from them.

Another possible explanation for the defects in nidogen-2-deficient NMJs is that nidogen-2 might be required to recruit or retain other synaptic BL components that are, in turn, required for synaptic integrity. To test this possibility, we immunostained nidogen-2 mutant muscle with a panel of antibodies to synaptic BL components. As expected, no nidogen-2 was present in mutants (Figure 7A). Although nidogen-2 is upregulated in some tissues of nidogen-1 mutants 50, we failed to detect any changes of nidogen-1 level at nidogen-2-deficient synapses (Figure 7B, compare with Figure 2A). The major synapse-specific laminin subunits – laminin α4, α5, and β2 – as well as agrin and the synapse-specific collagens α3–6(IV) were all retained in synaptic BL in the absence of nidogen-2 (Figure 7C–J). Neither levels nor distribution of these proteins were detectably affected by the absence of nidogen-2, although we would not have detected small changes. From these results, we conclude that neuromuscular defects in nidogen-2 mutants are not indirect consequences of loss of synaptic laminins, collagens IV or agrin.

A third possibility is that nidogen-2 selectively binds and presents a matrix-associated synaptic organizing molecule to nerve or muscle. Indeed, several bioactive matrix molecules have been reported to bind with much higher affinity to nidogen-2 than nidogen-1, including tropoelastin, collagen XIII and collagen XVIII/endostatin 37515253. Interestingly, we have found that mice with a targeted mutation of the collagen XIII gene have neuromuscular defects 54, and recent studies have shown that collagen XVIII is critical for motor axon growth and NMJ formation in C. elegans and zebrafish 265556. We therefore used immunohistochemical methods to ask whether synaptic localization of collagens XIII and XVIII are perturbed in nid2^-/- mice. Collagen XIII was concentrated at NMJs in control muscle, as reported previously 5457 and this concentration persisted in the absence of nidogen-2, although we cannot rule out the possibility that its level may be modestly affected (data not shown). In both control and nid2^-/- muscles, collagen XVIII was undetectable in synaptic BL, although it was associated with the BL of terminal Schwann cells (data not shown). Thus, the mechanisms by which nidogen-2 contributes to synaptic maturation and maintenance remain to be determined.

Recently, we and others have observed striking intermuscular differences in neuromuscular phenotypes in mutants lacking agrin or collagen IV chains α3–6 125859. The explanation for these differences remains unknown, but they are of interest because they may provide clues to mechanisms by which BL components act or muscles diversify. We therefore extended our study from diaphragm to three limb muscles – the extensor digitorum longus, soleus and tibialis anterior (Figure 8A–D). The soleus is a predominantly slow muscle; extensor digitorum longus and tibialis anterior are predominantly fast; and diaphragm is mixed. Nidogen-2 was concentrated at synaptic sites in limb muscles as well as in diaphragm (for example, Figure 2 shows limb muscle and Figure 4 shows diaphragm). Nonetheless, the percentage of abnormal junctions varied among muscles from approximately 20% in extensor digitorum longus to 75% in diaphragm (Figure 8E). Likewise, of those NMJs that were malformed, the ratio of those that were fragmented to those that were immature varied from 9:1 in extensor digitorum longus to 2:1 in diaphragm (Figure 8E). Intermuscular differences were less striking in 1-year-old mice, owing to a progressive accumulation of defects in mutant extensor digitorum longus and tibialis anterior muscles (data not shown). Thus, mutant muscles may vary in the rate at which defects accumulate rather than in their absolute susceptibility.

Muscles, and the NMJs within them, differ from each other in many respects 60 and it is not clear whether any documented factors explain the intermuscular differences we have observed. It is intriguing that NMJs more dependent on the presence of nidogen-2 are found in constantly active muscles (that is, diaphragm muscle controls respiration and soleus muscle controls lower limb posture), whereas fewer defects are observed in young, phasically active muscles (tibialis anterior and extensor digitorum longus). Another distinction is between two categories of muscles described by Pun et al. 58. Called 'fast synapsing' or 'fasyn' and 'delayed synapsing' or 'desyn,' they were initially distinguished by the tempo and pattern of NMJ formation within them during embryogenesis. They were subsequently shown to differ in their sensitivity to nerve injury and neurological disease 586162. Interestingly, the two muscles we examined that were most severely affected, diaphragm and soleus, are both 'delayed synapsing' muscles, whereas the two muscles less affected in nid2^-/- mice, tibialis anterior and extensor digitorum longus, are 'fast synapsing' muscles 58.

Biochemical studies have shown that interactions among three main structural components, laminins, collagens IV, and nidogens, are involved in BL assembly 6364. Moreover, all three of these components bind to HSPGs 463. Our results have shown that specific isoforms of all four components are colocalized in synaptic BL. One might therefore presume that interactions among synaptic isoforms would mediate assembly of this specialized domain. Yet, we found that nidogen-2 is dispensable for the synaptic localization of synaptic laminins, collagens IV and agrin (Figure 7). We therefore considered the alternative possibility, that other synaptic BL components might be required for synaptic localization of nidogen-2. To test this idea, we examined the distribution of nidogen-2 in four mutants lacking synaptic laminin chains (β2, α4, α5, or both α4 and α5) and in collagen α5(IV) mutants, which lack all four synaptic collagen IV chains (α3–6) 12. It was not possible to assess nidogen-2 in agrin mutants since these mice die at birth 21, prior to the synaptic restriction of nidogen-2.

Fragmentation and immaturity similar to that observed in nidogen-2 mutants has been reported in collagen α5(IV) and laminin α4 mutants and in laminin α4/α5 double mutants, respectively 121819. However, nidogen-2 remained concentrated in the synaptic BL of NMJs in all of these mutants (Figure 9). Thus, synaptic laminins and collagens IV are dispensable for concentrating nidogen-2 in the synaptic cleft. Moreover, the absence of nidogen-2 is not responsible for fragmentation of synapses lacking synaptic laminins or collagens IV.

Finally, we broadened our inquiry to ask whether any synaptic BL components are necessary for the recruitment and restriction of other BL components to the synaptic cleft. As noted above, nidogen-2 is not required for synaptic localization of laminins, collagens IV or agrin, nor does it require synaptic laminins or collagens IV to become synaptically localized. On the other hand, previous results show that loss of a single laminin subunit or collagen IV chain can lead indirectly to absence of other components of the trimer from the synaptic cleft 91265. Here, we asked whether synaptic laminins are required for localization of synaptic collagens IV or visa versa. We found that synaptic collagens as well as agrin are normally localized to the NMJ in mutants lacking laminins β2, α4, or α5, and that synaptic laminins as well as agrin are normally localized to the NMJ in mutants lacking synaptic collagens α3–6(IV) (Figure 10 and data not shown). Together, these results, summarized in Table 1, indicate that the localization of each family of synaptic BL components occurs independently of other BL components.

Targeted mutant and transgenic mice used in this study have been described previously. They are: nidogen-2 mutants (nid2^-/-) 31, laminin β2 mutants (lamb2^-/-) 16, laminin α4 mutants (lama4^-/-) 18, collagen α5(IV) mutants (col4a5^-/Y) 72, obtained from Jackson Laboratories, Bar Harbor, ME, USA), conditional laminin α5 mutants (lama5^flox/flox) 73, and mice that express Cre selectively in skeletal muscle (HSA-Cre) 74. Mutants lacking laminin α5 selectively in skeletal muscle were generated by crossing lama5^flox/flox and HAS-Cre mice; we refer to the lama5^flox/flox;HSA-Cre mice as lama5^M/M. Mice lacking both laminin α4 and α5 (lama4^-/-; lama5^M/M) were generated by crossing lama4^-/-;lama5^flox/+; HSA-Cre males and females. All mutants and transgenics were maintained on a C57/B6 background. In most cases, littermates of mutants were used as controls. CD1 mice were obtained from Charles River Laboratories, Inc. (Wilmington, MA, USA). All analyses conformed to NIH guidelines and were carried out under an animal protocol approved by the Harvard University Standing Committee on the Use of Animals in Research and Teaching.

Primary antibodies used in this study are listed in Table 2. All fluorescently labeled secondary antibodies were obtained from Invitrogen/Molecular Probes and were used at a 1:1000 dilution. HRP-conjugated secondary antibodies from Vector Laboratories (Burlingame, CA, USA) were used at a dilution of 1:5000.

Full-length human nidogen-1 and -2 fusion proteins were purchased from R&D Systems (Minneapolis, MN, USA). Recombinant proteins (250 ng) were denatured by boiling in Laemmli buffer and separated by SDS-PAGE. Electrophoretically separated proteins were transferred to Immuno-Blot PVDF membrane (Bio-Rad, Hercules, CA, USA) in Tris-Glycine buffer (25 mM Tris, 190 mM glycine, pH 8.4)/20% methanol at 300 mA for 2 hours. Immunoblotted proteins were detected as previously described 75.

Mice were perfused with phosphate-buffered saline (PBS) and tibialis muscles were dissected. Tissue was immediately frozen in OCT on dry ice and 4 μm sections were cut on a cryostat. Sections were collected on gelatin-coated slides and allowed to air-dry for 15 minutes before tissue was fixed by incubating in ice-cold acetone for 10 minutes. For collagen IV antibodies, tissue was treated for 10 minutes in a 1:1 mix of 1 M KCl and 1 M HCl following fixation 12. Sections were washed several times in PBS to remove any residual acid. After fixation, tissue was incubated with blocking buffer (5% non-fat milk in PBS with 0.2% Triton-X100 in PBS) for 30 minutes. Primary antibodies, diluted in blocking buffer, were incubated on the sections for 12 hours at 4°C, then sections were washed several times with PBS. Secondary antibodies, diluted in blocking buffer, were then incubated on the slides for 1 hour at room temperature. For controls, primary antibody incubation was omitted from the immunostaining protocol described above. Sections were washed thoroughly with PBS, cover-slipped with VectaShield, and visualized on an Olympus FV1000 scanning confocal microscope (Olympus America Inc., Melville, NY, USA).

C2C12 cells (ATCC, Manassas, VA, USA) were cultured as previously described by Kummer et al. 42. Myotubes were fixed with ice-cold acetone and stained as described above.