Nervous System Defects of Ankyrin_B (−/−) Mice Suggest Functional Overlap between the Cell Adhesion Molecule L1 and 440-kD Ankyrin_B in Premyelinated Axons

A clone containing 17 kb of the ank2 gene was isolated from a 129SVJ genomic lambda fix II library using a probe that covers nucleotides 2443– 2805 of the human ankyrin_B cDNA. A novel NotI site was introduced within an exon in the spectrin-binding domain, and this site was used to introduce a cassette that included the neomycin resistance gene (pMC1neo polyA vector; Stratagene, La Jolla, CA), an in-frame HA epitope, and a stop codon. An XhoI/HindIII fragment containing the thymidine kinase gene (from the vector pIC19R) was added to the construct. ES cells were transfected by electroporation with the linearized targeting construct, neomycin- and gancycolivir-resistant clones were isolated, and homologous recombination was verified by Southern blotting (see Fig. 1 B). ES cells were injected into blastocysts from C57Bl/6 mice and implanted in C57Bl/6 pseudopregnant mothers. Male chimeric progeny were mated with C57Bl/6 females to generate heterozygous mice. Homozygous mice were produced from none-littermate heterozygous mice. ES cell culture, transfection, and blastocyst injection were performed in the laboratory of Dr. Beverly Koller (University of North Carolina, Chapel Hill, NC) (Mansour et al., 1988; Koller et al., 1991).

Magnetic resonance (MR) microscopy was performed on a General Electric (Fremont, CA) 7.1 Tesla unit, 15-cm bore size, equipped with shielded gradient coils. A Helmholtz coil especially designed for MR microscopy of fixed specimens at 300 MHz was used for imaging (Banson et al., 1992). Formalin-fixed brains were suspended in Fomblin (perfluoropolyether; Ausimont, Inc., Morristown, NJ) to minimize dehydration and to limit susceptibility effects at the surface. Data was acquired as anisotropic arrays with an aspect ratio of 1:1:2 with the following acquisition parameters: (a) 18.2-h scan time; (b) 20-ms echo time; (c) 1,000-ms repetition time, diffusion gradient pulse (sine bell-shaped) durations of 5 ms; (d) diffusion gradient separation of 12.2 ms; (e) large diffusion gradients of 35 Gauss/cm, resulting in a diffusion-weighting factor of 965 s/mm², 256 × 256 × 128 digital resolution; and (f) voxel size of 39 × 39 × 78 μm. All images were analyzed using NIH Image 1.68 (public domain from The National Institutes of Health) and VoxelView Ultra (Vital Images, Inc., Fairfield, IA) for three-dimensional rendering.

Immunofluorescence was performed with 5–10-μm frozen brain sections, which were prepared and stained as described (Lambert et al., 1997). Affinity-purified rabbit antibodies were raised against the human COOH-terminal domain (residues 1551–1813) and 440-kD–specific tail domain (Chan et al., 1993) of ankyrin_B using recombinant proteins expressed in bacteria as antigens as described (Lambert et al., 1997). Chick antibody against the 270/480-kD ankyrin_G has been described elsewhere (Lambert et al., 1997). Rabbit polyclonal antibody against neurofascin was described in Davis et al. (1996). Antibody against L1 was purchased from Boehringer Mannheim Corp. (Indianapolis, IN). Electron microscopy was performed with optic nerve fixed with 4% glutaraldehyde and subsequently embedded in Epon. To determine the sizes of axons in the knock-out and the control mouse optic nerves, a sheet of paper containing a hole with dimensions of 8 × 8 cm was mounted on the electron micrographs, and the number of axons that appeared in the hole were counted. Northern and Southern blots (Sambrook et al., 1989) and immunoblots (Lambert et al., 1997) were performed as described. Antibodies used in immunoblotting experiments are anti-L1, a protein A affinity-purified polyclonal antibody (a generous gift from Dr. Patricia Manness of the University of North Carolina at Chapel Hill); polyclonal antispectrin and polyclonal antineurofascin as described in previous reports (Hu et al., 1995; Davis et al., 1996); and polyclonal anti-NCAM, purchased from Chemicon (Temecula, CA).

The ank2 gene encoding ankyrin_B was disrupted by homologous recombination (Fig. 1,A), and ankyrin_B +/+, +/−, and −/− mice (n = 325) were identified by Southern blots (Fig. 1,B) (Materials and Methods). This study is based on the first-generation progeny with a genetic background of 1:1 129SVJ to C57Bl/6. Brains of ankyrin_B (−/−) mice exhibited nearly complete loss of a major 13-kb mRNA transcript encoding the 440-kD ankyrin_B (Kunimoto et al., 1991) as well as loss of less abundant 8- and 5-kb transcripts encoding 220- and 150-kD ankyrin_B polypeptides, respectively (Fig. 1,C). 440- and 220-kD ankyrin_B polypeptides were undetectable in homogenates of brains of ankyrin_B (−/−) mice by immunoblotting (Fig. 1 D). A minor 150-kD polypeptide in the ankyrin_B (−/−) mice still is barely detectable in original gels, although not in the published version. 150-kD ankyrin_B lacks a membrane-binding domain (Otto et al., 1991). Apparently expression of the 150-kD ankyrin_B transcript is less affected by disruption of an exon in the spectrin binding domain (Materials and Methods).

440-kD ankyrin_B is a major isoform in the early postnatal period, as noted previously (Kunimoto et al., 1991). Ankyrin_B is also expressed in heart where the major isoform of 220 kD was undetectable in ankyrin_B (−/−) mice (Fig. 1 D).

Ankyrin_B (−/−) mice were born in nearly Mendelian ratios of 1:3 (81/325), demonstrating normal survival during fetal development. However, over half of the ankyrin_B (−/−) animals were dead or dying (and therefore killed) on the first postnatal day, and 95% by 8 d (n = 81) (Fig. 2,B). A few mutant mice survived until postnatal day 20 (4 out of 81). Mice that did survive to 2 wk were about 25% of the weight of littermates (n = 8) and exhibited abnormal locomotion and balance (Fig. 2 A).

Three-dimensional diffusion-weighted magnetic resonance microscopy (DWM) of formalin-fixed brains from a litter of 1-d-old mice (three mutant and three normal) was performed to obtain general information on gross morphology and accurate measurements of ventricle volumes. We used DWM, which measures diffusion of free proton spins to get optimal anatomical resolution. Brain anatomy is revealed in detail with DWM because proton diffusion within regions composed of oriented axon fiber tracts and those comprised of cell bodies is markedly different (Beaulieu and Allen, 1996).

Total brain and cerebellar volumes in the ankyrin_B (−/−) brains were nearly identical to those of (+/+) and (+/−) littermates (Table I). However, volumetric measurements of the ventricular systems revealed a significant sevenfold enlargement of the lateral ventricles in the ankyrin_B (−/−) brains compared with controls (Table I; Fig. 3). In addition, an enlarged lateral ventricle is easily visualized in Fig. 3. Enlargement of lateral ventricles of ankyrin_B (−/−) mice was not associated with occlusion of the cerebral aqueduct, which is closed in Fig. 3 but was patent in other sections (not shown). The extent of lateral ventricle dilation in neonatal ankyrin_B (−/−) mice is in agreement with that reported for some strains of adult L1 (−/−) mice (Dahme et al., 1997; Fransen et al., 1998b). Fransen et al. (1998b) found a twofold enlargement of both the frontolateral and fourth ventricle in adult L1 (−/−) mice. Ankyrin_B (−/−) mice did not exhibit fourth ventricle enlargement; however, they did not survive beyond day 21, and the full extent of enlargement of brain ventricles could therefore not be determined. L1 (−/−) mice in some genetic backgrounds do not exhibit enlarged ventricles (Dahme et al., 1997). Hydrocephalus is also variable in extent in humans with L1 mutations, even within the same kindred (Fransen et al., 1997). L1 hydrocephalus also is associated with a patent aqueduct of Sylvius in some cases (Fransen et al., 1995; Wong et al., 1995).

L1 mutations in mice and humans are associated with abnormalities in the pyramidal tracts and corpus callosum (Fransen et al., 1997; Cohen et al., 1998). DWM images revealed the presence of the corpus callosum and internal capsule in sagittal sections as well as pyramidal tracts in coronal sections in control mice (Fig. 3). The corpus callosum, internal capsule, and pyramidal tracts could not be resolved in brains of ankyrin_B (−/−) mice (Fig. 3). Histological brain sections also reveal a reduction in the corpus callosum in ankyrin_B (−/−) mice, with nearly complete absence of the anterior portion revealed in a sagittal section (Fig. 5). The corpus callosum also was greatly reduced in coronal sections of ankyrin_B mutant brain (Fig. 5). Hypoplasia of the corpus callosum was observed in 16/26 (postnatal days 1–5) mutant mouse brains examined by coronal sections. Some of the variability in detection of reduction of the corpus callosum may be due to selective loss of anterior relative to posterior regions.

The similarities in phenotype of ankyrin_B (−/−) mice and L1 (−/−) mice suggested the possibility that ankyrin_B and L1 are coexpressed in the nervous system. Evidence in support of segregation of L1 with ankyrin_B and of neurofascin with ankyrin_G is presented in Fig. 4. Neurofascin has previously been identified at axon initial segments and nodes of Ranvier, which are sites that also contain high local concentrations of 270/480-kD isoforms of ankyrin_G (Davis et al., 1996; Lambert et al., 1997). Neurofascin and 270/480 ankyrin_G are colocalized at axon initial segments in the neonatal cerebral cortex (Fig. 4). However, in neonatal mouse brain, neurofascin and 270/480 ankyrin_G were absent from premyelinated axon tracts, including the corpus callosum (Fig. 4, lower panels) as well the optic nerve and other sites in the central nervous system (data not shown). L1 and 440 ankyrin_B, in contrast, codistributed in premyelinated axon fiber tracts in the cerebral cortex (Fig. 4) and throughout the CNS (Figs. 5 and 6).

440-kD ankyrin_B and L1 are colocalized in premyelinated axons located in long fiber tracts in brain (corpus callosum, internal capsule, fornix, striatum, and olfactory tract) (Fig. 5) as well as spinal cord (pyramidal tract and lateral columns) (Fig. 6). Similar patterns of ankyrin_B staining in axon tracts were observed with antibodies against the COOH terminus, which react with all known alternatively spliced forms as with antibodies against 440-kD ankyrin_B (not shown). These patterns of distribution of L1 are consistent with previous studies (Bartsch et al., 1989; Joosten and Gribnau, 1989; Martini, 1994). Brain and spinal cord sections of ankyrin_B (−/−) mice exhibited no staining with antibody against the COOH terminus of ankyrin_B, confirming both the efficacy of gene disruption and specificity of the antibody.

L1 staining is reduced in premyelinated axon tracts in brains and spinal cords of ankyrin_B (−/−) mice (Figs. 5 and 6). The corpus callosum and internal capsule are reduced or absent in mutant animals (Fig. 5) (16/26), as noted above. However, fiber tracts in the brain that still were present and exhibited major reduction in L1 staining include the fimbria and stria terminalis (Fig. 5,A). L1 staining in the pyramidal tract and lateral columns of the spinal cord also is greatly reduced (Fig. 6). Ankyrin_B (−/−) animals exhibited variability with respect to L1 staining, with about 50% having less pronounced differences. For example, L1 staining still is evident in the sagittal section shown in Fig. 5 B. However, littermate control animals always exhibited strong L1 staining.

Loss of L1 in ankyrin_B (−/−) mice is also demonstrated by immunoblotting of brain tissue (Fig. 7). It was not feasible to obtain tissue samples from particular anatomical sites such as corpus callosum or optic nerve (see below), where loss of L1 is most prominent. Therefore, intact brains of a control and a mutant postnatal day 7 mouse were evaluated for levels of L1, neurofascin, NCAM, and spectrin (n = 1) (Fig. 7). Expression levels of neurofascin and NCAM exhibited no detectable change in the brain of the ankyrin_B (−/−) mouse. Moreover, levels of β_G-spectrin also were unchanged (Fig. 7). However, immunoblots of the same gel samples with antibody against L1 revealed a 30–40% percent reduction in amounts of 200-, 135-, and 43-kD cross-reacting polypeptides (Fig. 7). These results demonstrating partial loss of L1 are consistent with reduction of L1 in localized sites detected by immunofluorescence.

Loss of L1 staining in ankyrin_B (−/−) brain and spinal cord raises questions of whether lack of ankyrin_B impairs delivery of L1 to axons, whether L1-expressing axon tracts never form and are simply absent, and/or whether L1-expressing axons are present initially but subsequently degenerate. The optic nerve provides a simplified, easily identifiable anatomical site for addressing issues related to L1 behavior and axonal degeneration in ankyrin_B (−/−) mice. The premyelinated optic nerve has previously been identified as a site where the 440-kD ankyrin_B is the sole ankyrin (Chan et al., 1993), and also expresses high levels of L1 (Bartsch et al., 1989). 440-kD ankyrin_B and L1 both exhibit a dramatic down-regulation coincident with myelination and are not expressed in adult myelinated axons of the optic nerve (Bartsch et al., 1989; Chan et al., 1993).

Transverse sections through optic nerve at postnatal days 1–3 revealed unreduced staining for L1 in ankyrin_B (−/−) mice, even though ankyrin_B staining was absent (n = 7) (Fig. 8). These observations establish that at least early in the postnatal period, optic nerve axons are present, and that L1 is delivered to axons in the absence of ankyrin_B. However, L1 staining disappeared from optic nerves of ankyrin_B (−/−) mice by around postnatal day 7, although the nerve itself was still present (Fig. 8) (n = 6). Optic nerves of normal mice retained L1 and ankyrin_B (Fig. 8). In contrast to L1, levels of NCAM were unchanged in the optic nerve section of ankyrin_B (−/−) mice at postnatal day 7 (Fig. 8). Neurofascin was only found at the initial segments of the premyelinated optic nerve axons and was absent from the optic nerves of both control and ankyrin_B (−/−) mice. Electron micrographs (see below; Fig. 9) reveal that myelination occurred with only a few axons at this stage and was limited to a few loops of myelin. The normal loss of L1 accompanying myelination therefore had not yet begun at the time when L1 disappeared from ankyrin_B (−/−) optic nerves.

Electron microscopy of transverse sections of the postnatal day 1 optic nerve revealed relatively normal axon profiles in ankyrin_B (−/−) mice, although the average axon cross-sectional area was increased about 20% (see Materials and Methods; data not shown). Axons were packed in tightly assembled arrays, indicating that some level of axon fasciculation occurs in the absence of ankyrin_B. Sections of postnatal day 9 optic nerve reveal a major change in axons of ankyrin_B (−/−) mice (Fig. 9). Many axons were increased in diameter up to eightfold compared with the normal axon and contained multivescicular bodies. Longitudinal sections demonstrated increased diameter along the length of ankyrin_B (−/−) axons (not shown). Axon fasciculation also was impaired in sections of ankyrin_B (−/−) optic nerve with many examples of loss of direct axon–axon contacts (Fig. 9).

All ankyrin_B (−/−) mice surviving until postnatal day 21 exhibited nearly complete degeneration of the optic nerve (n = 8) (Fig. 10). The optic nerve of ankyrin_B (−/−) mice thus forms initially with normal axonal delivery of L1 and subsequently degenerates. Ankyrin_B and ankyrin_G are coexpressed before myelination in most other nerves, including the trigeminal nerve (Lambert et al., 1997). The optic nerve, which only has ankyrin_B, apparently is especially vulnerable in ankyrin_B (−/−) mice.

This study describes dramatic consequences of deficiency of ankyrin_B in mice that allow survival until birth but result in hypoplasia of the corpus callosum and pyramidal tracts, dilated lateral ventricles, followed by extensive degeneration of the optic nerve, and death by postnatal day 21. This phenotype is similar to, but more severe, than those observed in L1 (−/−) mice and human patients with L1 mutations. L1 was reduced or absent in ankyrin_B (−/−) mutant mice in premyelinated axons of long fiber tracts in the brain and spinal cord as well as optic nerve at postnatal day 7. These findings provide the first evidence for a role of ankyrin_B in the vertebrate nervous system and support a functional connection between ankyrin_B and L1 that is required for maintenance of premyelinated axons in vivo.

L1 deficiency in mice is compatible with survival to adulthood (Dahme et al., 1997; Cohen et al., 1998; Fransen et al., 1998b) and represents a much less severe phenotype than ankyrin_B deficiency. Other L1 CAM family members (CHL1, NrCAM, NgCAM, and neurofascin) may functionally compensate for L1 in L1 (−/−) mice. In addition, ankyrin_B is likely to have additional roles unrelated to L1, particularly early in development when some L1 CAMs are tyrosine phosphorylated and therefore incapable of binding to ankyrin (Garver et al., 1997; Zisch et al., 1997; Heiland et al., 1998). Conversely, L1 and other members of the family are likely to have ankyrin-independent functions when these proteins are tyrosine phosphorylated. This study demonstrates that early in development, L1 is capable of axonal transport to the optic nerves of ankyrin_B (−/−) mice, implying that its delivery to the plasma membrane and axonal targeting are independent of ankyrin_B. It will be of interest to determine the phosphorylation state of L1 in the early postnatal optic nerve, where this protein is transported to axons in the absence of ankyrin_B (Fig. 6).

The mechanism for loss of L1 in ankyrin_B (−/−) axons is not known, but it could result from reduced stability of L1 in the absence of ankyrin_B. A similar situation in which loss of a cytoskeletal protein results in reduction of integral proteins has been found in mdx mice, where absence of dystrophin results in reduction of dystrophin-associated proteins (Matsumura et al., 1992). A practical consequence of reduced expression of interacting proteins is that a characterization of patterns of gene or protein expression in the ankyrin_B (−/−) mouse brain may yield heretofore unidentified proteins whose functions involve ankyrin_B.

The striking increase in axonal diameter of optic nerve axons suggests the hypothesis that the absence of ankyrin_B leads to loss of mechanical stability of axons. Ankyrin_B could contribute to axonal membrane support by connecting the spectrin–actin network to the plasma membrane, as occurs in erythrocytes (Peters and Lux, 1993). In this case, axonal swelling would be analogous to hereditary spherocytosis of erythrocytes observed in cases of ankyrin_R mutations (Eber et al., 1996). The tail domain of the 440-kD ankyrin_B potentially could extend several hundred nanometers into the axoplasm and also mediate interactions with structural proteins in addition to the spectrin skeleton. L1-mediated connections between axons combined with coupling to the spectrin–actin network via ankyrin could provide an additional level of mechanical support for axon bundles. Such a transcellular network based on L1 and ankyrin is analogous, in a formal sense, to the epidermis, where keratinocytes are connected through desmosomes on their outer surfaces, and desmosomes are interconnected via intermediate filaments. The consequence in both cases is a mechanical buffering system capable of distributing shear stresses throughout the tissue. It is of interest in this regard that the brains of ankyrin_B (−/−) are much more fragile and susceptible to damage during dissection compared with brains of normal mice (data not shown).

Ankyrin_B and L1 are coexpressed and may interact at sites in addition to premyelinated axons. Unmyelinated axons in the molecular layer of the adult cerebellum express 440-kD ankyrin_B as well as L1 (Kunimoto et al., 1991; Martini, 1994; Kunimoto, 1995). 220-kD ankyrin_B is expressed in glial cells, including Schwann cells and astrocytes (Kordeli et al., 1990; Kunimoto et al., 1991; Kunimoto, 1995), and could associate with alternatively spliced forms of L1 present in these cells (Martini, 1994). L1 is expressed outside of the nervous system in cells including lymphocytes (Kowitz et al., 1992). It will be of interest in future experiments to examine L1 distribution and function in ankyrin_B (−/−) mice at sites of axo–glial contact and in the immune system.

Ankyrin and L1 have both been implicated in axon guidance. L1-deficient mice exhibit defects in guidance of corticospinal axons across the pyramidal decussation (Cohen et al., 1998). Ankyrin (Unc-44) mutations in Caenorhabditis elegans result in abnormal axon outgrowth and guidance (Otsuka et al., 1995). It is not clear from the current data if ankyrin_B and L1 collaborate in axonal guidance, or if ankyrin_B has any role in axonal guidance. It will be of interest to trace the paths of axons in the optic nerve and corticospinal tracts of ankyrin_B (−/−) mice to determine possible defects in decussation and in final arrival at neuronal targets. Consequences of ankyrin_B deficiency, including degeneration of the optic nerve, could be secondary to defective targeting of axons to their normal sites accompanied by apoptosis caused by lack of retrograde transport of appropriate factors.

The accumulated L1 mutations characterized in human families provide a valuable database for structure–function studies of the L1 molecule (Fransen et al., 1998a). One missense mutation in the L1 cytoplasmic domain results in conversion of Y1229 to H and clinical disease (Van Camp et al., 1996). The Y1229H mutation of the FIGQY tyrosine is of particular interest with respect to ankyrin. Recent analysis of structural requirements for L1 CAM binding to ankyrin has demonstrated that the FIGQY tyrosine is essential for ankyrin binding (Hortsch et al., 1998; Zhang et al., 1998). Moreover, mutation of the FIGQY tyrosine to histidine abolishes ankyrin binding activity of neurofascin (Zhang et al., 1998). These considerations suggest the possibility that ankyrin binding activity of L1 is also abolished by the Y1229H mutation, and that loss of ankyrin binding results in loss of L1 function in humans.

CAM

cell adhesion molecule

DWM

diffusion-weighted MR microscopy

MR

magnetic resonance