Genetic Analysis of Collagen Q: Roles in Acetylcholinesterase and Butyrylcholinesterase Assembly and in Synaptic Structure and Function

Genomic clones containing the ColQ gene were isolated by screening a 129sv strain mouse genomic library (Stratagene) with cDNAs encoding rat ColQ (Krejci et al., 1997). For targeting vector PRAD1 (Fig. 1 a), a 0.8-kb EcoRI-HindIII fragment containing the exon encoding the PRAD domain as well as adjacent intronic sequences was replaced by a PGK-neo cassette (Tybulewicz et al., 1991). For targeting vector PRAD2 (Fig. 1 a), a 0.6-kb HindIII fragment containing the PRAD exon was replaced by a cassette containing neo plus the Escherichia coli lacZ gene, for monitoring gene expression. The PRAD2 vector also contained a diphtheria toxin-A gene for negative selection (Yagi et al., 1990). Both constructs were transfected into R1 ES cells (Nagy et al., 1993) by electroporation. Homologous recombinants were identified by PCR and confirmed by probing genomic Southern blots with an 0.8-kb fragment of the ColQ gene that was entirely external to the targeting vector (Fig. 1 a). Chimeric mice from one PRAD1 and two PRAD2 ES cell clones gave rise to heterozygous and then homozygous mutant mice. The PRAD2 vector contained a lacZ gene, but no β-galactosidase activity was detected in heterozygotes, either histochemically or immunohistochemically.

For immunohistochemistry, sternomastoid and tibialis anterior muscles were embedded in Tissue-Tek OCT compound (Sakura Finetek USA), frozen in liquid nitrogen–cooled isopentane, and sectioned at 10 μm in a cryostat. Sections were blocked for 1 h with 2% BSA and 5% normal goat serum in PBS, incubated with primary antibodies for 1–2 h, washed, and then reincubated with secondary antibodies for another 1–2 h. Antibodies to the following antigens were used: ColQ (Krejci et al., 1997), AChE (a gift of Terrone Rosenberry, Mayo Clinic), α-sarcoglycan, β-dystroglycan, utrophin (Novocastra Laboratories Ltd.), rapsyn (Phillips et al., 1991), and agrin (a gift of Z. Hall, University of California, San Francisco). FITC-conjugated Vicia villosa agglutinin-B4 (VVA-B4) was obtained from Sigma Chemical Co. and rhodamine-α-bungarotoxin was obtained from Molecular Probes.

Cholinesterase activity was detected by the histochemical method of Karnovsky and Roots (1964) on sections fixed with 2% paraformaldehyde in PBS. To distinguish AChE from BuChE, the reaction mixture was supplemented with 10⁻⁴ M tetraisopropylpyrophosphoramide (iso-OMPA), a selective inhibitor of BuChE, or with 5 × 10⁻⁵ M 1:5-bis (4-allyldimethylammoniumphenyl)-pentan-3-one dibromide (BW284c51), a selective inhibitor of AChE (Austin and Berry, 1953). Where indicated, sections were treated with 1% Triton X-100 for 60 min at room temperature or highly purified bacterial collagenase (800 U/ml, type VII; Sigma Chemical Co.) for 60 min at 37°C before fixation and staining.

For electron microscopy, mice were perfused with lactated Ringer's solution followed by 4% paraformaldehyde and 4% glutaraldehyde in 100 mM cacodylate buffer, pH 7.2. Sternomastoid muscles were dissected and fixed overnight at 4°C. The endplate-rich region of the muscle was refixed in 1% OsO₄ in cacodylate buffer, dehydrated, and embedded in Araldite. Thin sections were stained with uranyl acetate and lead citrate.

Mice were killed and tissues were homogenized in a glass Potter homogenizer in 1 ml of low-salt (LS) buffer, which contained 50 mM Tris, pH 7.0, 40 mM MgCl₂, 1% Triton X-100, 2 mM benzamidine, 40 μg/ml leupeptin, and 25 μg/ml pepstatin. The homogenates were centrifuged for 20 min at 60,000 rpm, the supernatant was removed, and the pellet was homogenized in a second aliquot of LS buffer and centrifuged again. The two resulting supernatants comprised the detergent-soluble (DS) extracts. The pellet was then homogenized with 300 μl of high-salt (HS) buffer, which consisted of LS buffer plus 0.8 M NaCl. After centrifugation, the supernatant (HS extract) and the pellet were both saved.

Samples of the DS and HS extracts were analyzed by sedimentation in 5–20% sucrose gradients containing 50 mM Tris, pH 7.0, 40 mM MgCl₂, 1% Brij96, and 0.4 M NaCl. Centrifugation was in an SW41 rotor for 16 h at 39,000 rpm. About 75 fractions of 150 μl were collected from the bottom of the tube. 60-μl aliquots were transferred to a microtiter plate. Cholinesterase activity was assayed colorimetrically, using acetylcholine as substrate (Ellman et al., 1961). To assay AChE, iso-OMPA (10⁻⁴ M) was included to inhibit BuChE. To assay BuChE, BW284c51 (10⁻⁶ M) was included to inhibit AChE. Alkaline phosphatase (6.1 S) and β-galactosidase (16 S) from E. coli were included as internal sedimentation standards, and their activity profiles were used to establish a linear relationship between the fraction numbers and Svedberg units (Simon et al., 1998). All reagents were from Sigma Chemical Co.

ColQ^−/− mice or control littermates were killed and immediately exsanguinated. Hemi-diaphragms with their associated phrenic nerves were isolated. The two hemi-diaphragms were separated and each was mounted in a 2-ml bath lined with Rhodorsil (Rhône-Poulenc) and superfused with saline of the following composition: 154 mM NaCl, 5 mM KCl, 2 mM CaC1₂, 1 mM MgCl₂, 5 mM Hepes, pH 7.4, and 11 mM glucose. The solution was bubbled with pure O₂.

The membrane potential and miniature endplate potentials (mepps) were recorded from endplate regions at room temperature (22–24°C) with intracellular microelectrodes filled with 3 M KCl (resistance of 16–26 MΩ), using conventional techniques and an Axoclamp-2A system (Axon Instruments). Recordings were made continuously from the same endplate before and throughout application of the cholinesterase inhibitors tested. Data were digitized and analyzed using a computer equipped with an analogue and digital I/O interface board (DT2821; Data Translation) and a program kindly provided by Dr. John Dempster (University of Strathclyde, Scotland). Mepps were analyzed individually for amplitude and time course.

Clones encoding the ColQ gene were isolated from a genomic library and used to produce two targeting vectors, PRAD1 and PRAD2 (Fig. 1 a). Both vectors deleted the PRAD domain of ColQ, which forms the binding site for the catalytic (AChE_T) subunits of AChE (Bon et al., 1997). Both targeting vectors gave rise to homologous recombinant ES cells, germline chimeras, heterozygotes, and homozygotes (Fig. 1, b and c). Phenotypes of PRAD1 and PRAD2 homozygotes were identical, so both are called ColQ^−/− and described together here.

Homozygous ColQ^−/− mice were born in expected numbers, and were externally indistinguishable from heterozygous (ColQ^+/−) and wild-type siblings at birth. Around 5 d after birth (postnatal day [P]5), however, ColQ^−/− mice began to exhibit a tremor when moving. The tremor persisted throughout life. The homozygotes remained active, responsive, and capable of feeding, but failed to thrive, and grew at a slower rate than littermates (Fig. 1 d). About half of the ColQ^−/− mice died at about the time of weaning (P21) and around two-thirds of the survivors died over the next few weeks. 10–20% of ColQ^−/− mice lived into adulthood (>3 mo) but remained much smaller than littermates.

Sections of skeletal muscle from ColQ^−/− mutants and littermates were doubly stained with an antiserum to recombinant ColQ plus rhodamine-α-bungarotoxin, which binds to AChRs and thereby marks synaptic sites. In controls, ColQ was concentrated at synaptic sites and was undetectable extrasynaptically (Fig. 2, a and a′). In ColQ^−/− mutants, AChR-rich synaptic sites were nearly normal in size, shape, and distribution (see below) but, as expected, bore no ColQ (Fig. 2, b and b′).

The ColQ subunit is believed to anchor AChE to the basal lamina at the neuromuscular junction. To obtain definitive evidence on this point, we stained sections of mutant and control muscle with antibodies specific for the AChE catalytic subunit. This antibody recognizes both of the known alternatively spliced products (T and H) of the mammalian AChE gene (Duval et al., 1992). AChE was abundant at all synaptic sites in controls, but undetectable in homozygotes (Fig. 2, c and d).

This result suggested that ColQ is required for accumulation of all synaptic AChE. An alternative explanation, however, was that the association of AChE with synaptic membranes or matrix had merely been weakened in the absence of its collagenous tail, in which case AChE might have been present at synapses in vivo but lost during preparation of sections. Therefore, we performed two additional tests to seek AChE at synaptic sites in ColQ^−/− muscle in situ. First, we injected fasciculin-2, a selective inhibitor of AChE (Rodriguez-Ithurralde et al., 1983) into ColQ^−/− mice and their littermates. This inhibitor was chosen because it penetrates the central nervous system poorly, and is therefore believed to exert its effects predominantly on peripheral AChE. Fasciculin-2 (3–6 μg/g, injected intraperitoneally) caused tremor followed by flaccid paralysis in controls. In mutants, in contrast, effects of fasciculin were minor, and injected animals were reflexive and capable of righting. Second, we recorded mepps from muscle fibers of ColQ^−/− mice and littermate controls (Fig. 3). The decay times of mepps were variable but on average slower in homozygotes than in controls (data not shown), presumably because AChE terminates neurotransmitter action at normal neuromuscular junctions (Whathey et al., 1979), whereas diffusion does so in ColQ^−/− muscle. More importantly, addition of fasciculin greatly slowed the decay of mepps in normal muscle, but had no significant effect on the time course of mepps in ColQ^−/− muscle (from 3.57 ± 0.18 to 3.47 ± 0.07 ms in four experiments; P > 0.2). Together, these toxicological and physiological tests support the idea that ColQ^−/− mutants lack functional synaptic AChE.

We used sucrose density gradient centrifugation to assess the molecular forms of AChE present in ColQ^−/− muscles. Globular forms are extracted into detergent-containing buffers at low ionic strength, whereas asymmetric (collagen-tailed forms) are extracted only at high ionic strength (Bon et al., 1978). Therefore, we first homogenized muscles with detergent-containing LS buffers, then reincubated the pellet with HS buffers to selectively solubilize asymmetric forms.

Control (ColQ^+/+ and ColQ^+/−) muscles contain three asymmetric forms, corresponding to a single helical trimer of ColQ chains attached to one, two, or three tetramers of catalytic subunits (A₄, A₈, and A₁₂, respectively; Massoulié and Bon, 1982). In contrast, no asymmetric forms were detectable in ColQ^−/− muscle (Fig. 4, b and c). Because some asymmetric AChE is difficult to solubilize from tissue (Rossi and Rotundo, 1993), we also assayed the pellet remaining after HS and detergent extraction. AChE activity in insoluble fractions was 14% of that in the HS extract for ColQ^+/+ muscles (n = 4), and 30% of that in the HS extract for ColQ^+/− muscles. However, AChE activity was undetectable in the insoluble fraction from ColQ^−/− muscles. Similar results were obtained with muscles from neonates (P0) and at P20. Thus, all asymmetric AChE in muscle requires ColQ for its assembly and/or accumulation.

Surprisingly, loss of AChE in ColQ^−/− mice was not limited to the asymmetric forms. At birth, three major peaks of AChE activity were distinguishable in detergent extracts of control muscle, corresponding to amphiphilic monomers and dimers (G₁^a and G₂^a) and nonamphiphilic tetramers (G₄^na; Massoulié et al., 1998). G₁^a and G₂^a were retained in neonatal ColQ^−/− muscles, but the G₄^na peak was almost entirely absent (Fig. 4 a). In addition, the pool of globular tetramers extracted by HS was absent from ColQ^−/− neonates (Fig. 4 b). At P20, however, amphiphilic and nonamphiphilic tetramers (G₄^a and G₄^na) as well as G₁^a and G₂^a forms were present in ColQ^−/− muscles (data not shown). Thus, ColQ-independent mechanisms for assembly or retention of G₄ do exist (see Discussion).

BuChE is homologous to AChE, hydrolyzes acetylcholine, and is present at the neuromuscular junction (Chatonnet and Lockridge, 1989; Chapron et al., 1997). Like AChE, BuChE catalytic monomers assemble into both globular and asymmetric multimers, and the asymmetric multimers bear a collagenase-sensitive subunit (Vigny et al., 1978; Toutant et al., 1985). To determine whether ColQ serves as the collagen tail for BuChE, we assayed sucrose density gradients in the presence of a selective inhibitor of AChE, BW284c51. Asymmetric and globular forms of BuChE were detectable in control muscle (Fig. 4, d and e). As reported previously, S values of BuChE differed from those of AChE (Vigny et al., 1978; Krejci et al., 1997), providing assurance that the activity seen in the presence of BW284c51 did not reflect incomplete inhibition of AChE. Moreover, activity was inhibited by the selective inhibitor of BuChE, iso-OMPA (data not shown). Globular isoforms of BuChE were retained in ColQ^−/− mutants, but no asymmetric BuChE was detectable (Fig. 4, d and e). Thus, the ColQ gene appears to encode the collagenous tails of both AChE and BuChE.

Lacking antibodies to BuChE, we used the histochemical stain of Karnovsky and Roots (1964) to localize BuChE in control and mutant muscles. This method relies on hydrolysis of the acetylcholine analogue, acetylthiocholine, which is cleaved by both AChE and BuChE. At P20, reaction product was readily detectable at ColQ^−/− synaptic sites, although levels were substantially reduced compared with controls (Fig. 5, a and b). The AChE inhibitor BW284c51 had no effect on the staining of ColQ^−/− synaptic sites, but decreased staining in controls to approximately the level seen in mutants (Fig. 5, c and d). This result provides independent evidence that AChE is depleted from ColQ^−/− neuromuscular junctions. In contrast, the BuChE inhibitor iso-OMPA decreased staining of control synaptic sites only slightly, but completely abolished staining of ColQ^−/− synaptic sites (Fig. 5, e and f). Thus, the synaptic acetylcholine-hydrolyzing activity in mutants is largely or entirely BuChE. Reaction product was undetectable in controls even after prolonged incubation when a mixture of BW284c51 plus iso-OMPA was added with the substrate, indicating that AChE and BuChE together account for most or all of the ACh-hydrolyzing activity at control synapses (Fig. 5, g and h).

To obtain information on the molecular forms and cellular localization of synaptic BuChE, we treated cross-sections in various ways then stained them in the presence of BW284c51. In both control and ColQ^−/− muscle, reaction product was tightly associated with synaptic sites (Fig. 5, i and j). Pretreatment of sections with either collagenase, which releases matrix-associated enzyme, or with the detergent Triton X-100, which releases membrane bound or intracellular enzyme, reduced staining of synaptic sites in control muscles (Fig. 5, k and m). Combined treatment with collagenase and Triton X-100 removed all detectable BuChE (Fig. 5, o and p). Thus, both matrix- and membrane-associated BuChE are present at control synapses. In ColQ^−/− muscle, collagenase treatment had no effect on the level of BuChE (Fig. 5 l) but detergent treatment abolished all staining (Fig. 5 n), confirming the absence of matrix-associated BuChE and showing that residual activity is membrane-associated. Denervation, which causes loss of nerve terminals and rapid remodeling of terminal Schwann cells, led to complete loss of this membrane-associated BuChE from ColQ^−/− synapses but had little effect on controls (Fig. 5, q and r). Together, these results suggest that both membrane- and matrix-associated forms of BuChE are normally concentrated at the neuromuscular junction. The BuChE of the synaptic cleft is predominantly matrix-associated, asymmetric, and requires ColQ for assembly or localization. Membrane-associated BuChE that persists in the absence of ColQ, in contrast, is likely to be associated with nerve terminals or terminal Schwann cells, although we cannot exclude the possibility that it is postsynaptic.

Although the synapse-associated BuChE in ColQ^−/− muscles did not appear to be localized to the synaptic cleft, it might have been near enough to hydrolyze released acetylcholine. We tested this possibility by determining whether iso-OMPA affected the decay of mepps recorded from ColQ^−/− neuromuscular junctions. No significant effect was seen (τ = 2.70 ± 0.15 ms in normal saline and 2.98 ± 0.49 ms in the presence of 10 μM iso-OMPA; n = 4; P > 0.2). Therefore, although there is synapse-associated BuChE in ColQ^−/− mice, it does not appear to provide a mechanism to compensate for loss of AChE.

To assess the size and shape of neuromuscular junctions in ColQ^−/− muscle, thick longitudinal sections were stained with rhodamine-α-bungarotoxin. In controls, the AChR-rich membrane is a roughly circular plaque at birth, then passes through a perforated-plaque stage (Steinbach, 1981) and eventually matures into a pretzel-like array of distinct AChR-rich branches by P20 (Fig. 6 a). The precise geometry of branches is unique at each synaptic site, but the general pattern is stereotyped. In mutant homozygotes, in contrast, synaptic geometry was variable. Some synaptic sites were smaller than controls but roughly normal in appearance (Fig. 6 b), some retained the immature appearance characteristic of ∼1-wk-old controls (Fig. 6 c), and some appeared fragmented (Fig. 6 d). Approximately 40% of the neuromuscular junctions were normal in geometry, another 40% were fragmented, and the remaining 20% were immature.

Electron microscopy revealed two structural abnormalities at synaptic sites in ColQ^−/− muscles. First, the cytoplasm beneath the postsynaptic membrane was often riddled with holes, consistent with a localized degenerative response (Fig. 7 b). Such local degeneration has been observed after acute inhibition of AChE (Laskowski et al., 1977), and is believed to result from entry of Ca²⁺ through excessively activated AChRs (Leonard and Salpeter, 1979). Consistent with this interpretation, signs of necrosis were seldom observed in nonsynaptic regions of ColQ^−/− muscle, and no signs of muscle fiber degeneration and regeneration (such as central nuclei, which are diagnostic of immature or regenerated fibers) were detectable. Second, nerve terminals were sometimes partially enwrapped by processes of Schwann cells (Fig. 7 d). Such enwrapment has been seen in several pathological situations, and can occur after muscle damage, or as a direct consequence of abnormalities in the synaptic cleft (Duchen et al., 1974; Jirmanová, 1975; Noakes et al., 1995; Patton et al., 1998). Because Schwann cell processes are impermeable to neurotransmitter, their intrusion into the synaptic cleft might represent one adaptive mechanism by which ColQ^−/− synapses limit activation of AChRs.

Interestingly, the incidence of subsynaptic necrosis and Schwann cell enwrapment of nerve terminals varied as a function of age. At P20, nearly two-thirds of synaptic sites encountered (80/125) showed clear signs of necrosis. By 6 mo of age, however, <5% of synaptic profiles (2/71) showed necrosis, and many were indistinguishable from controls (compare Fig. 7, a and c). This difference raises the possibility that damage occurring early is repaired and that compensatory mechanisms, such as Schwann cell enwrapment, prevent damage from recurring. Indeed, the incidence of nerve terminals partially or completely enwrapped by Schwann cell processes increased somewhat, from one-third (42/125) at P20 to over one-half (39/71) at 6 mo of age. This difference suggests either that Schwann cell enwrapment is progressive, or that individuals with more complete enwrapment are more likely to survive. Necrosis and Schwann cell wrapping were seen in only 1 and 4 of 83 control endplates, respectively.

Finally, we stained sections of muscle from ColQ^−/− mice and littermate controls with a panel of antibodies to proteins present at synaptic sites. These included the transmembrane proteins α-sarcoglycan and β-dystroglycan, which are present throughout the muscle fiber but concentrated in the postsynaptic membrane; rapsyn and utrophin, which are selectively associated with the subsynaptic cytoplasm; and agrin, which is concentrated in synaptic basal lamina (Sanes and Lichtman, 1999). The distribution of all of these markers was qualitatively similar in control and homozygous mutant mice (Fig. 8, a–j).

We also stained sections with the lectin VVA-B4. VVA-B4 binds to glycoconjugates that terminate with a β-N-acetylgalactosaminyl residue and selectively stains the synaptic cleft at the neuromuscular junction (Scott et al., 1988). We showed previously that VVA-B4 binds to asymmetric but not globular forms of AChE in muscle, but presented indirect evidence that other VVA-B4–binding proteins were also present in the synaptic cleft (Scott et al., 1988; Martin and Sanes, 1995). VVA-B4 stained synaptic sites nearly as intensely in ColQ^−/− muscle as in control muscle (Fig. 8, k and l), providing direct evidence for the existence of additional synaptic VVA-B4–binding moieties.

Asymmetric forms of AChE are most abundant in skeletal muscle, but are also present in several other tissues (Toutant and Massoulié, 1988). To determine whether asymmetric AChE in nonmuscle tissues requires ColQ for assembly, we analyzed HS extracts of brain and heart. Low levels of asymmetric AChE were detectable in both tissues from control mice but in neither tissue from ColQ^−/− mice (Fig. 9).

We also made a preliminary survey of tissue structure in the major organs of ColQ^−/− mice, but found no abnormalities.

Previous biochemical studies have shown that the ColQ protein is a collagenous subunit of asymmetric AChE, and suggested that it is critical for accumulation of AChE in the synaptic cleft of the neuromuscular junction. Analysis of ColQ^−/− mutant mice has now allowed us to ascertain, in situ, the roles that ColQ plays in the assembly and anchoring of AChE and in the development and function of the neuromuscular junction. Specifically, we have been able to answer the seven questions posed in the Introduction.

First, no asymmetric AChE is present in muscles that lack ColQ, presumably because the ColQ protein forms the collagen tail that endows asymmetric forms with their distinguishing feature. This result was expected but not inevitable. For example, ColQ might have been only one of multiple collagenous subunits, a possibility suggested by immunological studies showing a structural change in the collagen tail during posthatching development in birds (Tsim et al., 1988b). Alternatively, a synaptic collagen that normally plays other roles (Miner and Sanes, 1994, 1996) might have been capable of compensating for lack of ColQ in mutants.

Second, muscles from newborn ColQ^−/− mice lack not only all asymmetric AChE, but also the nonamphiphilic tetrameric form of AChE, G₄^na. This observation was unexpected because the G₄^na molecules have been assumed to be homotetramers of AChE_T catalytic subunits. Possible explanations for this result include the following. G₄^na molecules may be derived from collagen-tailed molecules by proteolysis of the collagen tail, either in vivo or after extraction. Assembly of the tetramer might require transient association of AChE_T with ColQ. Assembly might require an alternatively spliced product of the ColQ gene that contains the amino-terminal PRAD domain, but not the collagenous domain. Such alternatively spliced products do in fact occur (Krejci et al., 1997), but have not yet been characterized. Some of the G₄^na may comprise tetramers associated with a single full-length collagen chain rather than a triple helix; such molecules sediment around 10S (Krejci, E., unpublished results). In any event, muscles of ColQ^−/− mice did contain both amphiphilic and nonamphiphilic tetramers (G₄^a and G₄^na) at P20. These tetramers might have been assembled without any organizer or might contain an alternate organizing subunit such as the still incompletely characterized P subunit (Massoulié et al., 1993). Thus, although the PRAD of ColQ can induce tetramerization of AChE_T (Bon et al., 1997; Simon et al., 1998), and may do so in vivo at P0, there are clearly additional PRAD-independent mechanisms for forming tetramers.

Third, no detectable AChE is present at synaptic sites in ColQ^−/− muscles. Histochemical, immunochemical, toxicological, and electrophysiological assays all gave similar results in this regard. The lack of synaptic AChE could reflect loss of collagen-tailed AChE and/or loss of G₄ forms. We favor the interpretation that all synaptic AChE is collagen-tailed because all of the histochemically detectable AChE can be removed by collagenase treatment of sections (data not shown) or tissue (Hall and Kelly, 1971). In addition, all globular forms (G₁^a, G₂^a, G₄^a, and G₄^na) are present in ColQ^−/− mice at P20, yet synaptic AChE is absent at all ages tested. Globular AChE is associated with isolated basal lamina from frog muscles (Anglister et al., 1994), but this may result from the alterations produced in the experimental system of muscle and nerve degeneration.

Fourth, ColQ is required for assembly or accumulation of asymmetric BuChE as well as AChE. In birds, collagen-tailed molecules containing both AChE and BChE subunits have been detected in muscles of hatchlings (Tsim et al., 1988a), suggesting that both enzymes use a common tail. Likewise, Krejci et al. (1997) showed that antibodies to ColQ interact with both AChE and BuChE as assessed by sucrose density centrifugations. Taken together with these biochemical data, our genetic results strongly support the conclusion that the ColQ gene produces the collagen tails for both AChE and BuChE.

Fifth, the function of the neuromuscular junction is impaired in the absence of ColQ. However, the defects are not as severe as might be expected, considering the dramatic effects of acutely administering extremely selective blockers of peripheral AChE such as fasciculin. In particular, movements of ColQ^−/− mice are relatively normal, while fasciculin-treated normal mice are paralyzed. Moreover, the postsynaptic necrosis observed in ColQ^−/− weanlings, a likely result of excessive calcium influx through AChRs (Leonard and Salpeter, 1979), is largely absent from those mutants that survive to later ages. These observations suggest that neuromuscular junctions compensate for the absence of AChE. Compensation does not appear to involve persistence or upregulation of BuChE: asymmetric BuChE is lost in the mutant and the BuChE inhibitor, iso-OMPA, does not affect the time course of mepps. Instead, ultrastructural studies suggest a simple structural mechanism to limit the amount of neurotransmitter that reaches the postsynaptic membrane: invasion of the synaptic cleft by Schwann cell processes at ColQ^−/− synapses. Enwrapment of nerve terminals by Schwann cell processes is, in fact, actively regulated by another component of synaptic basal lamina, laminin-11 (Patton et al., 1998). ColQ might exert a similar effect. In addition, compensatory changes in quantal size and quantal content may occur; such homeostatic changes have been documented in response to several chemical or genetic perturbations of the efficacy of neuromuscular transmission in both vertebrates and invertebrates (reviewed in Landmesser, 1998). Currently, we are using electrophysiological methods to seek such alterations in ColQ^−/− mice, with the aim of learning how the synapse combines structural and physiological compensations to adapt to the challenge posed by AChE deficiency.

Sixth, initial formation of the neuromuscular junction proceeds in the absence of ColQ and AChE. AChE promotes the extension of neurites from cultured motoneurons by a noncatalytic mechanism (Sternfeld et al., 1998; see also Small et al., 1995), and overexpression of AChE in transgenic frogs and mice perturbs formation of neuromuscular junctions in vivo (Seidman et al., 1994; Andres et al., 1997; Sternfeld et al., 1998). Our results imply that the developmental roles of ColQ and AChE at the synapse are either subtle or masked by compensatory mechanisms. However, it remains possible that expression of AChE at other sites is crucial for synaptogenesis. Moreover, at later stages, there are clear defects in the maturation of ColQ^−/− neuromuscular junctions: some remain immature in geometry and others appear fragmented. These abnormalities could reflect a direct requirement for AChE or ColQ in synaptic maturation or maintenance, but might also be secondary consequences of subsynaptic necrosis or alterations in activity levels. Further experiments will be required to distinguish these alternatives.

Seventh, ColQ is required for assembly or accumulation of asymmetric AChE in all tissues tested, but tissues other than skeletal muscle are not obviously defective in structure or function. However, it is premature to conclude that ColQ plays no roles in these tissues. For example, asymmetric AChE is concentrated in synaptic layers of the retina (Reiss et al., 1996); analyses of these synapses have not yet been undertaken. In addition, it is tempting to speculate that in AChE-poor tissues, ColQ or other PRAD-containing products of the ColQ gene serve to anchor proteins other than AChE to membranes or matrix. Identification of such proteins will guide the search for subtle phenotypes in ColQ^−/− mice.

Finally, it is interesting to note the parallels between ColQ^−/− mice and humans with AChE-deficiency syndrome (CMS-Ic). This syndrome was described by Engel et al. (1977) over 20 years ago, and was reported just this year to result from defects in the human COLQ gene (Donger et al., 1998; Ohno et al., 1998). Some patients appear to be protein nulls, while others have mutations in the carboxy terminus of ColQ that permit formation of asymmetric AChE but prevent its stable association with the basal lamina of the synaptic cleft. All of the COLQ patients are severely myasthenic, as are the ColQ^−/− mice, making the mice a valid model for assessing pathogenic mechanisms and therapeutic interventions.

We thank Mia Nichol, Carol Borgmeyer, Jimmy Gross, and Sherry Weng for assistance.

This work was supported by grants from the National Institutes of Health and Centre National de la Recherche Scientifique, the Direction des Recherches et Etudes Techniques, the Association Française contre les Myopathies, and Human Frontier Science Program Organization. G. Feng was a fellow of the Jane Coffin Childs Memorial Fund for Medical Research.

AChE

acetylcholinesterase

AChR

acetylcholine receptor

BuChE

butyrylcholinesterase

ColQ

collagen Q

DS

detergent-soluble

ES cells

embryonic stem cells

HS

high-salt

LS

low-salt

mepps

miniature endplate potentials

PRAD

proline-rich attachment domain

VVA-B4

Vicia villosa agglutinin-B4