Reduced muscle strength in ether lipid‐deficient mice is accompanied by altered development and function of the neuromuscular junction

Abstract Inherited deficiency in ether lipids, a subgroup of phospholipids whose biosynthesis needs peroxisomes, causes the fatal human disorder rhizomelic chondrodysplasia punctata. The exact roles of ether lipids in the mammalian organism and, therefore, the molecular mechanisms underlying the disease are still largely enigmatic. Here, we used glyceronephosphate O‐acyltransferase knockout (Gnpat KO) mice to study the consequences of complete inactivation of ether lipid biosynthesis and documented substantial deficits in motor performance and muscle strength of these mice. We hypothesized that, probably in addition to previously described cerebellar abnormalities and myelination defects in the peripheral nervous system, an impairment of neuromuscular transmission contributes to the compromised motor abilities. Structurally, a morphologic examination of the neuromuscular junction (NMJ) in diaphragm muscle at different developmental stages revealed aberrant axonal branching and a strongly increased area of nerve innervation in Gnpat KO mice. Post‐synaptically, acetylcholine receptor (AChR) clusters colocalized with nerve terminals within a widened endplate zone. In addition, we detected atypical AChR clustering, as indicated by decreased size and number of clusters following stimulation with agrin, in vitro. The turnover of AChRs was unaffected in ether lipid‐deficient mice. Electrophysiological evaluation of the adult diaphragm indicated that although evoked potentials were unaltered in Gnpat KO mice, ether lipid deficiency leads to fewer spontaneous synaptic vesicle fusion events but, conversely, an increased post‐synaptic response to spontaneous vesicle exocytosis. We conclude from our findings that ether lipids are essential for proper development and function of the NMJ and may, therefore, contribute to motor performance. Read the Editorial Highlight for this article on page 463.

In addition to a variety of proteins, also the lipid environment, particularly the different types of glycerophospholipids, has emerged as an important factor modulating synaptic activity (Rohrbough and Broadie 2005;Davletov and Montecucco 2010). Lipid bilayers and their interplay with the associated proteins provide the framework for endoand exocytotic processes during the synaptic vesicle cycle regulating the release of neurotransmitters (Davletov and Montecucco 2010). The detailed molecular composition of the membranes involved has been shown to be of crucial importance for the regulation of the dynamics of fusion events in the nervous system (Chernomordik et al. 2006).
Ether (phospho)lipids constitute a particular subgroup of glycerophospholipids, which differs from the more abundant diacyl glycerophospholipids by the nature of the chemical bond attaching fatty acids to the sn-1 carbon of the glycerol backbone (diacyl phospholipids: ester bond; ether lipids: O-alkyl, i.e., ether bond). Several subtypes of ether lipids have been identified: plasmalogens, the platelet-activating factor, a potent inflammatory mediator (Prescott et al. 2000) or the lipid part of the glycosyl-phosphatidyl-inositol (GPI) anchor, a post-translational modification that tethers proteins to the outer face of the cell membrane (Kanzawa et al. 2009). Plasmalogens, defined by a double bond adjacent to the O-alkyl bond (vinyl ether bond), are the most abundant representatives, in humans making up about 20% of the total phospholipid mass. Inherited defects in ether lipid biosynthesis cause rhizomelic chondrodysplasia punctata (RCDP), which in its severest form is lethal in early childhood (Steinberg et al. 2006;Wanders and Waterham 2006). Affected patients are confronted with a variety of severe impairments, including growth and mental retardation, shortening of the proximal long bones, epiphyseal stippling, cataracts, facial dysmorphism, joint contractures, and respiratory problems (Braverman and Moser 2012). Although the genetic basis of RCDP is known, many of the underlying molecular processes remain unclear, as the functions of ether lipids have not been fully elucidated. Among the proposed tasks of plasmalogens are: the storage of essential polyunsaturated fatty acids like docosahexaenoic acid or arachidonic acid (Ford and Gross 1989); the stimulation of invariant natural killer T cells (Facciotti et al. 2012); and the generation of second messengers (Braverman and Moser 2012). Several in vitro studies have also proposed a role in antioxidative defense (Zoeller et al. 1999(Zoeller et al. , 1988Broniec et al. 2011), which has not yet been reliably confirmed in vivo (Brodde et al. 2012;Wallner and Schmitz 2011). Most importantly, as constituents of almost all biological membranes, plasmalogens shape membrane properties and structure as well as promote fusion and constriction processes (Thai et al. 2001;Hermetter et al. 1989;Paltauf 1994;Glaser and Gross 1994). Plasmalogens have been described to be enriched in lipid rafts (also termed membrane rafts) (Pike et al. 2002), small heterogeneous membrane domains, which are highly dynamic and compartmentalize cellular processes like signal transduction (Simons and Ikonen 1997;Pike 2006). This still expanding range of functions is reflected by the multifaceted pathogenesis of RCDP and other peroxisomal disorders, but alterations in the levels of ether lipids have also been reported in more common disease conditions (Berger et al. 2016), including Alzheimer's disease (Kou et al. 2011;Goodenowe et al. 2007), Parkinson's disease (Fabelo et al. 2011), Down syndrome (Murphy et al. 2000), or hypertension (Graessler et al. 2009).
The glyceronephosphate O-acyltransferase knockout (Gnpat KO) mouse, in which the gene coding for the first enzyme in the ether lipid biosynthesis pathway (glyceronephosphate (or dihydroxyacetone phosphate) O-acyltransferase, EC 2.3.1.42; Fig. 1) is disrupted, is a wellestablished model of complete, isolated ether lipid deficiency (Rodemer et al. 2003) and, thus, serves as an ideal tool to study the biological relevance of this lipid family for various tissues in the context of a mammalian organism (Gorgas et al. 2006). Gnpat KO mice are characterized by a reduced, but highly variable lifespan, growth deficits, male infertility, and ocular anomalies (Rodemer et al. 2003;Saab et al. 2014;Komljenovic et al. 2009). More detailed studies of the central nervous system revealed abnormalities in cerebellar structures and in evoked neurotransmitter release from presynaptic nerve terminals (Brodde et al. 2012;Teigler et al. 2009). Recently, deficits in myelination and Schwann cell development were demonstrated in the peripheral nervous system of Gnpat KO mice (da Silva et al. 2014).
The neuromuscular junction (NMJ), referring to the synapse between nerve and muscle, because of its large size, accessibility and relatively simple structure, is a widely studied model of peripheral synapses (Sanes and Lichtman 2001). At the NMJ, the neurotransmitter acetylcholine binds reversibly to ionotropic acetylcholine receptors (AChRs) on the surface of muscle fibers. A characteristic feature of the adult NMJ is the organization of the post-synaptic membrane into invaginations (junctional folds) containing extremely dense clusters of AChRs next to a variety of auxiliary proteins, which enable the transmission of the electric signal (Shi et al. 2012). The NMJ differs from central synapses in the elaborate series of maturation steps, which require several weeks. In this process, the heparan sulfate proteoglycan agrin, which is released by the motor axon, plays a central role (Bezakova and Ruegg 2003). Post-synaptically, agrin binds to LRP4, which interacts with the skeletal muscle receptor tyrosine protein kinase MuSK, thereby inducing the activation of MuSK (Kim et al. 2008;Zhang et al. 2008). MuSK kinase activity induces the clustering of AChRs via a complex post-synaptic machinery involving the cytoplasmic linker protein rapsyn (Apel et al. 1997). AChR clusters form early in development but undergo considerable remodeling at later developmental stages. Although they appear as oval plaque-like structures on a flat surface in newborn mammals, they adopt a more complex shape concomitant with formation of the junctional folds during post-natal development (Marques et al. 2000).
Novel determinants for the correct maturation and synaptic function of the NMJ are still being found. Based on their (i) significance for signaling cascades like the AKT (protein kinase B) pathway (da Silva et al. 2014), (ii) involvement in membrane fusion and constriction events (Glaser and Gross 1994), and (iii) abundance in synaptic vesicle and presynaptic membranes (Takamori et al. 2006;Hofteig et al. 1985), ether lipids may well modulate the development and activity of the NMJ. Therefore, in this study, we analyzed the consequences of ether lipid deficiency for formation, maintenance, and function of the NMJ. Based on the observation of an abnormal motor behavior phenotype of the Gnpat KO mouse, potentially involving NMJ deficits, we characterized the morphology and electrophysiological properties of NMJs in these mice.

Mice
Gnpat KO mice (Gnpat tm1Just ; MGI:2670462) were maintained on an outbred C57BL/6 9 CD1 background (Rodemer et al. 2003); homozygous Gnpat À/À (KO) and Gnpat +/+ (wild type, WT) littermates were obtained by mating heterozygous animals. To ensure complete ether lipid deficiency in the target tissue, we confirmed the absence of plasmalogens from dissected muscles (Fig. 2). In line with previous observations (Rodemer et al. 2003), Gnpat KO mice showed significantly decreased survival rates and the longest-living animals were females (Fig. 3a). Slight deviations from the originally reported survival curves probably derive from a drift in the genetic background of the strain and attentive care of KO animals to ensure their survival. The Gnpat genotype was determined at weaning by PCR as described previously (Rodemer et al. 2003) and confirmed at the time of killing. Mice carrying a thermolabile variant of the simian virus 40 large tumor antigen under the mouse major histocompatibility complex H-2K b promoter (Tg(H2-K1-tsA58)11Kio, MGI:3762405, 'Immorto') (Jat et al. 1991) were crossed with Gnpat tm1Just mice for the establishment of immortalized myoblast cell lines. The presence of the transgene was determined at weaning or, in case of myoblast isolation experiments, after killing by PCR, as described previously (Kern and Flucher 2005). All mice were fed standard chow and water ad libitum and were housed in a temperature-and humidity controlled room with 12 : 12 h light-dark cycle and a low level of acoustic background noise at the inhouse animal facility (Medical University of Vienna). All animals received humane care and handling in compliance with institutional and national (Austrian) regulations (BGBl. II Nr. 522/2012) as well as the European Union Directive 2010/63/EU and the use of these genetically modified animals was approved (BMWF-5.011/0003-II/ 10b/2009). For all experiments with adult mice, age-and sex-matched WT littermates were used as controls. For the studies of NMJ development, the sex of the fetuses was not determined. Staging of timed pregnancies was evaluated by vaginal plug detection (E0.5 = noon of the day following overnight mating, vaginal plug at morning inspection).

Behavioral assays
Rotarod (Dumser et al. 2007), balance beam (Carter et al. 2001), weights, and inverted screen (Deacon 2013) tests were performed as described previously with slight modifications. Further information is provided in the Supporting Information and the adapted scoring system used for balance beam testing is specified in Table S1. Investigators were blinded to the genotype of the mice (although the phenotype of Gnpat KO mice is overt in most cases).

Morphologic examination of NMJs
Staining of fetal diaphragms and adult muscles was performed as described previously (Herbst et al. 2002;Kravic et al. 2016). More detailed information is provided in the Supporting Information.

Determination of AChR lifetime
Kinetics of AChR lifetimes in WT and Gnpat KO mice were analyzed by pulse labeling as described recently (Strack et al. 2011). Briefly, on day zero, mice were anesthetized by intraperitoneal administration of xylazin/zoletil and 10 lL 125 I-labeled bungarotoxin (BTX; Perkin Elmer) solution (2.5 lCi) were injected into tibialis anterior muscles of both hind limbs. During the following 24 days, measurements of emitted X-rays from labeled muscles were performed under isoflurane narcosis (air mixture 0.6-1.5% administered by a tube) using a portable Germanium semiconductor counter (GX3018, Canberra Industries). Radiation emerging from the rest of the body was blocked by a lead shield with a circular fenestration positioned at the level of the tibialis anterior muscle. Individual measurements lasted for 300 s. Data were analyzed with the help of an attached multi-channel analyzer (InSpector 2000 DSP Portable Spectroscopy Workstation, Canberra Industries). Obtained data were confirmed by microscopic determination of AChR stability as described (Roder et al. 2010).

Monitoring neural transmission by electrophysiology
Recordings at the NMJ were mainly performed as described earlier (Kravic et al. 2016). Technical details can be retrieved from the Supporting Information. Quantal content was calculated from the ratio between end plate currents (EPCs) and miniature end plate currents (mEPCs). Capacitance was calculated as C = R/s with C as capacitance, R as input resistance, and s as the time constant of the voltage response to the current injection (84% of the rise time curve) Fig. 2 Absence of plasmalogens from muscle tissue of Gnpat KO mice. Plasmalogen levels were determined by detecting dimethylacetals (DMA) after acidic methanolysis and normalizing them to their corresponding fatty acid methyl esters (FAME). Neither C16:0 (a) nor C18:0 (b) DMA were detected in muscle tissue of Gnpat KO mice (WT, n = 4; Gnpat KO, n = 3). n.d., not detected based on previous literature (Gage and Eisenberg 1969). All electrophysiological recordings were performed at 23°C.

Isolation and cultivation of primary myoblasts
Newborn mouse pups (Gnpat KO and WT carrying the H-2Kb-tsA58 transgene) were killed by decapitation and their limbs stored in growth medium (Dulbecco's modified Eagle's medium supplied with 50 units/mL penicillin, 100 lg/mL streptomycin, 10% fetal bovine serum (FBS), 10% horse serum, 0.5% (vol/vol) chick embryo extract). Muscle tissue was isolated and minced in phosphate-buffered saline (PBS) with 1% glucose, taken up in growth medium and dissociated with 0.2% trypsin and 0.01% DNase for 30 min (37°C, 6% CO 2 ). After centrifugation (5 min, 400 g), the pellet was dissolved in growth medium containing 20 U/mL recombinant mouse interferon-c (IFN-c, Peprotech) and a single-cell solution prepared by use of pipet tips and cell strainers (100 lm, BD Falcon). Cells were pre-plated for 35 min at 33°C, 6% CO 2 and non-adherent cells transferred to gelatin (0.2%)-coated dishes. Individual clones of cells were expanded under the permissive conditions (33°C, 6% CO 2 ) and their Note that some KO pups may have died before their birth was recorded. Statistical analysis using log-rank tests revealed highly significant differences between the genotypes (p < 0.001 for males, females or both sexes combined) but not between sexes, although the longestliving animals were females (see inset). (b) In the accelerating rotarod test (maximal time 500 s), the latency to fall was recorded in adult WT (n = 26) and Gnpat KO (n = 27) mice. *p < 0.05 (Mann-Whitney Utest). (c) Scores obtained by the same cohort as in (b) in the balance beam test are depicted as the number of animals per genotype reaching a certain score (defined in Table S1); the better the performance, the lower the score. Statistical analysis using Mann-Whitney U-test revealed a highly significant difference between the genotypes (p < 0.001). (d) For the inverted screen task, the mean latency to fall was recorded for adult WT (n = 30) and Gnpat KO (n = 31) mice. n.s., not significant (Mann-Whitney U-test) (e) Cumulative scores (time 9 weight coefficient) obtained in the weights test are depicted for the same cohorts as in (d). ***p < 0.001 (Mann-Whitney U-test). Box-and-whisker plots in (b), (d), and (e) are drawn according to Tukey's method with the horizontal, bold line indicating the median value. differentiation into myotubes induced by removal of IFN-c, chick embryo extract, and FBS as well as a shift of the growth temperature to 37°C. According to their ability to form myotubes, two Gnpat KO cell lines and one WT cell line were included in clustering experiments.

AChR clustering assay
To induce AChR clustering, myotubes were stimulated with conditioned medium containing neural agrin (A4B8) prepared from HEK 293T cells (Herbst and Burden 2000;Tsim et al. 1992) for 8 h. To visualize surface AChRs, cells were fixed with 4% paraformaldehyde in PBS for 10 min at 23°C, washed twice with PBS for 5 min, and incubated with Alexa 594-conjugated a-BTX (200 ng/mL in 2% FBS/PBS) for 30 min. Cells were washed twice with PBS for 5 min and mounted in Mowiol 4-88. AChR clusters were imaged with a DM-IRB inverted fluorescence microscope (Leica, Wetzlar, Germany) using a 63X oil immersion magnification objective. Metamorph (Molecular Devices) and ImageJ software were used to acquire and quantify images as described previously (Camurdanoglu et al. 2016).

Determination of plasmalogen levels
Plasmalogen levels in muscle homogenates were determined by detecting dimethylacetals as described previously (Dacremont and Vincent 1995). Additional information is provided in the Supporting Information.

Statistical analysis
The groups of WT and Gnpat KO mice were compared with each other using two-tailed Student's t-tests or Mann-Whitney U-tests, depending on the nature of the variable to be analyzed. Statistical details for each experiment can be found in the Figure legends. The number of animals was kept to an absolute minimum; where applicable, required numbers were estimated using Java applets assuming a minimal statistical power of 0.85.

Results
Ether lipid deficiency causes motor impairment Observation of the home cage behavior of Gnpat KO mice revealed several features indicative of motor impairment such as unsteady gait. Therefore, we systematically assessed motor coordination and muscle strength in young adult mice in several behavioral tasks. In the accelerating rotarod test, Gnpat KO mice showed a significantly shorter latency to fall off the turning rod compared with WT mice (Fig. 3b). This difference is probably an underestimate, because some WT mice, after several training sessions, lost interest in the task and stopped running midway during the trial. Differences between WT and Gnpat KO mice were more obvious, when calculating the fraction of animals reaching the maximal time of the task (500 s) in at least one trial [WT: 11/26 animals (42%); Gnpat KO: 4/27 (15%)]. The same cohort of animals was subject to the balance beam task, during which the motor performancein particular, foot placement and slipswhile traversing a horizontal bar was evaluated. Gnpat KO animals experienced severe problems in this task, with their limbs regularly slipping below the horizontal midline of the bar. This resulted in significantly lower (i.e., better) average scores (p < 0.001, Mann-Whitney U-test) ranging from 2 to 4 for WT and from 3 to 5 for Gnpat KO mice (Fig. 3c).
To more specifically address muscle strength, we applied two further simple tests: the inverted screen and the weights tests (Deacon 2013). In the inverted screen task, mice had to cling onto an inverted grid as long as possible or for a maximum of 90 s. In this rather unselective test, both WT and Gnpat KO mice performed similarly and many WT as well as KO animals reached the maximum time (Fig. 3d). The number of animals achieving the maximal time in at least one trial was 27/30 (90%) for WT and 23/31 (74%) for Gnpat KO mice. In contrast, the weights test, which assesses the weightlifting capability of mice, revealed significant differences between the genotypes (p < 0.001, Mann-Whitney U-test), with Gnpat KO mice achieving considerably lower scores than WT controls (Fig. 3e). Because ether lipiddeficient animals are typically smaller than their WT littermates, we considered a normalization to the body weight for each mouse. However, correlation analysis within the genotype groups did not show a significant association between body weight and performance score (Fig. S1). To explore potential gender differences, we also exposed a considerable number of female mice (n = 18 per genotype) to the inverted screen and the weights tests in an independent series of experiments and obtained highly similar results as for males (data not shown). Taken together, our results indicate considerable motor behavior deficits, affecting motor coordination as well as muscle strength, in Gnpat KO mice.
The lack of ether lipids impairs formation of the neuromuscular junction Although various parameters contribute to motor function, motor impairment has been widely associated with disturbances in the formation or the maintenance of the NMJ. Based on the observed behavioral phenotype, we, therefore, conducted a morphological examination of the NMJ in WT and Gnpat KO mice. Because of its thinness and accessibility, the diaphragm muscle is widely used to study motor innervation and NMJ morphology. Thus, we isolated diaphragms from WT and Gnpat KO mouse fetuses at different developmental stages (embryonic day (E) 14.5, 16.5, and 18.5) and visualized pre-and post-synaptic components of the NMJ. At all time points examined, Gnpat KO diaphragms were characterized by abnormally extensive branching of the phrenic nerve and a general widening of the endplate zone (Fig. 4a-c). Main axons appeared partially defasciculated, particularly in the ventral region of the diaphragm, and we found a number of larger axons, each giving rise to several smaller branches (arrows in Fig. 4). Consequently, upon quantification, the area covered by nerves and nerve endings was significantly increased in Gnpat KO samples at all time points examined. Although this defect seemed less pronounced at a later embryonic stage (E18.5, Fig. 4c)most likely because of the larger overall size of the diaphragm as well as the innervated areawe still observed considerably increased phrenic nerve branching also at this stage suggesting that the phenotype is not ameliorated during development. Concomitant with the alterations in nerve appearance, post-synaptic AChR clusters on the muscle fibers were spread across a much wider area in Gnpat KO than in WT diaphragms (Fig. S2a). In spite of these abnormalities, there was a perfect colocalization of preand post-synaptic components of the NMJ indicating that synapses are correctly assembled (Fig. S2b). Furthermore, we did not detect any apparent differences in the structure or the size of individual AChR clusters, which at the embryonic stage appear as oval plaques.
Upon maturation, rodent NMJs are pretzel-like shaped (Sanes and Lichtman 2001). To judge whether mature AChR cluster formation is impaired in Gnpat KO mice in comparison with WT mice, we performed a quantitative analysis of AChR clusters in several hind limb muscles (soleus, gastrocnemius, extensor digitorum longus, and tibialis anterior). The morphometric analysis revealed a smaller volume and surface area of a-BTX-labeled AChR clusters in ether lipid-deficient muscles (Fig. S3a-c). The effect of genotype on these two parameters was statistically highly significant for the soleus muscle. Also the difference in surface area in the gastrocnemius muscle reached statistical significance, whereas in the other muscles, only a trend was observed. There was no difference in the mean gray values of AChR clusters between genotypes indicating identical densities of AChRs (Fig. S3d). However, AChR clusters in Gnpat KO mice were clearly less fragmented than WT clusters in all four muscles analyzed ( Fig. S3a and e).
To test whether an intrinsic clustering defect could account for these findings, we generated immortalized myoblasts and obtained two stable cell lines from Gnpat KO and one from WT mice carrying the Immorto (H-2K b -ts58) transgene. In order to assay AChR clustering, we stimulated differentiated myotubes derived from these myoblasts with neural agrin and analyzed the resulting patches of AChRs by fluorescence microscopy. AChR clustering was clearly impaired in ether lipid-deficient myotubes, as indicated by significantly smaller and fewer AChR clusters per myotube area (Fig. 5a-d). Also the size of Gnpat KO myotubes themselves was decreased in comparison with WT tubes (Fig. 5e).
AChR stability is not affected by ether lipid deficiency Because ether lipids, particularly plasmalogens, shape membrane structure and fluidity, membrane proteins like AChRs are likely to be influenced by alterations in membrane lipid composition evoked by the lack of ether lipids. Particularly lipid rafts, which provide the lipid environment for AChR clustering (Pato et al. 2008;Stetzkowski-Marden et al. 2006), could be negatively affected by ether lipid deficiency. We, therefore, hypothesized that the absence of ether lipids in Gnpat KO mice impairs the stability of AChR clusters. In order to address this question, we labeled AChRs in the tibialis anterior muscles of WT and Gnpat KO mice by injecting 125 I-BTX, which binds irreversibly to muscle-type AChRs, and studied the loss of radioactive emission as a measure of AChR turnover over time. Remarkably, we and area (c), both normalized to myotube area, as well as the Feret diameter of clusters (d) and the size of myotubes (e) are presented as means AE SD (n = 5 experiments/genotype involving a total of > 225 clusters/genotype derived from one WT and two KO cell lines). ***p < 0.001, **p < 0.01, *p < 0.05 (two-tailed Student's t-test). obtained almost perfectly identical decay curves for both genotypes (Fig. 6) demonstrating similar lifetimes of AChRs in WT and Gnpat KO mice. These results were further confirmed by fluorescence labeling of different AChR subsets: surface receptors were stained using Alexa Fluor 647-coupled BTX ('old receptors'). Ten days later, newly generated receptors were stained by Alexa Fluor 555-coupled BTX ('new receptors') and the ratio between new and old receptors was subsequently evaluated as a measure of AChR stability by in vivo confocal microscopy. Similarly, as found in the radioligand assay, also by this method Gnpat KO mice did not differ from WT controls (Fig. S4).

Electrophysiological abnormalities are observed at ether lipid-deficient NMJs
Based on the morphological abnormalities that we observed in NMJs of Gnpat KO mice, we evaluated functional consequences by recording neuromuscular transmission in phrenic nerve-diaphragm explants dissected from adult WT and Gnpat KO mice. We detected, by trend, increased amplitudes of miniature end plate potentials (mEPPs) in Gnpat KO compared with WT muscles (Fig. 7a and b). These were accompanied by significantly increased mEPC amplitudes (means AE SEM: WT (n = 7 mice, 64 fibers total): 2.338 AE 0.109 nA; Gnpat KO (n = 7 mice, 85 fibers total): 3.031 AE 0.196 nA; p = 0.009 [two-tailed Student's ttest]). Rise times of mEPPs did not differ significantly between the genotypes (means AE SEM: WT (n = 5 mice, 47 fibers total): 0.314 AE 0.019 ms; Gnpat KO (n = 5 mice, 63 fibers total): 0.319 AE 0.007 ms; p = 0.829 [two-tailed Student's t-test]). In addition, we detected a strikingly decreased frequency of mEPPs in fibers derived from Gnpat KO mice, demonstrating a reduced number of spontaneous vesicle fusion events at ether lipid-deficient NMJs (Fig. 7c). Interestingly, in all our recordings, we observed a strong and highly significant increase of about 30% in input resistance in Gnpat KO preparations (Fig. 7d), possibly reflecting changes in membrane properties caused by ether lipid deficiency.
To evoke synaptic transmission in response to a presynaptic stimulus, we applied 1 Hz pulses to the phrenic nerve and measured the resulting end plate potentials (EPPs). Remarkably, the amplitude of EPPs did not differ between WT and Gnpat KO diaphragm NMJs (Fig. 7e). From the current flows (EPC and mEPC) accompanying EPPs and mEPPs, we calculated the quantal content, a measure of the number of vesicles released upon a pre-synaptic stimulation event. In line with the observed increase in mEPP amplitude but normal EPPs, the quantal content was significantly reduced in nerve-muscle preparations derived from Gnpat KO mice (Fig. 7f)

Discussion
Because of the poor performance of ether lipid-deficient mice in tests assessing motor coordination and muscle strength, we suspected impairments at the level of the NMJ in these mice. Indeed, morphological analysis of developing and adult NMJs revealed several abnormalities in Gnpat KO mice in comparison with WT littermates. The most striking feature Fig. 6 Normal acetylcholine receptor (AChR) stability in ether lipiddeficient mice as indicated by in vivo radioligand binding. Adult WT (n = 4 limbs derived from two mice) and Gnpat KO (n = 6 limbs derived from three mice) mice were injected intramuscularly (tibialis anterior muscle) with radioactively labeled a-BTX and the loss of radioactivity was monitored over 24 days. Data are presented as means (circles) AE SEM and the curves were fitted as described previously (Strack et al. 2011). Statistical analysis using two-tailed Student's t-tests did not reveal significant differences at any time point. Thus, no correction for the repeated measurements was performed.
was the defasciculated appearance of the main axons innervating the fetal diaphragm as well as a strongly increased area covered by nerve terminals leading to an overall widening of the total endplate zone (Fig. 4). This NMJ phenotype persisted throughout embryonic development indicating a permanent defect of NMJ morphology.
Development of the mammalian NMJ is a highly complex process, which requires the exchange of molecules between pre-and post-synaptic compartments as well as auxiliary cell types like Schwann cells. Given the numerous signal transduction processes involved in NMJ maturation, lipid rafts are probable candidates contributing to the observed alterations in Gnpat-deficient mice, as these small membrane domains serve as organizing platforms for signaling events (Simons and Toomre 2000). Plasmalogens have been identified as constituents of lipid rafts (Pike et al. 2002), but their exact role and how ether phospholipid deficiency affects the stability and function of rafts have not yet been established. In lipid raft fractions isolated from Chinese hamster ovary (CHO) cells with a deficiency in either alkyldihydroxyacetone phosphate synthase, catalyzing the second step in ether lipid biosynthesis (Fig. 1), or peroxisome biogenesis, detergent resistance and lipid composition were normal (Honsho et al. 2008). However, the protein and lipid content of lipid rafts can vary considerably depending on their functional context (Pike 2004;Levental et al. 2011). Thus, disturbances of phospholipid homeostasis may be more devastating in the increased complexity of a mammalian organism than in cultured cells. This is also underlined by the observation that in the brain of Gnpat KO mice, the properties of lipid raft domains and proteins associated with them appear altered (Rodemer et al. 2003). The participation of lipid rafts in NMJ maturation have been reported repeatedly; for example, the characteristic clustering of AChRs has been suggested to require rafts (Zhu et al. 2006;Stetzkowski-Marden et al. 2006;Campagna and Fig. 7 Increased amplitude and strongly reduced frequency of mEPPs but normal evoked potentials at diaphragm muscles of Gnpat KO mice. Miniature end plate potentials (mEPPs), reflecting spontaneous vesicle fusion, were recorded from diaphragms of adult WT (left panel) and Gnpat KO (right panel) mice; representative traces for each genotype are shown in (a). The mEPP amplitude (b), mEPP frequency (c), and input resistance (d) (b-d: WT, n = 7 mice, 88 fibers total; Gnpat KO, n = 7 mice, 118 fibers total) as well as the amplitude of end plate potentials (EPPs) in response to 1 Hz pulses (WT, n = 5 mice, 50 fibers total; Gnpat KO, n = 7 mice, 86 fibers total) (e) and quantal content (the ratio between end plate currents (EPC) and mEPC; WT, n = 5 mice, 43 fibers total; Gnpat KO, n = 7 mice, 86 fibers total) (f) are presented as means AE SEM. ***p < 0.001, **p < 0.01, *p < 0.05, n.s., not significant (two-tailed Student's t-test). Bezakova and Ruegg 2003). Also, agrin, the master organizer of NMJ development, regulates the immunological synapse via modulation of lipid rafts in lymphocytes (Khan et al. 2001). Indeed, our in vitro study of AChR clustering as well as the quantification of clusters in vivo (Fig. 5 and Fig. S3) suggest an impairment in cluster formation upon ether lipid deficiency and it is tempting to speculate that lipid rafts play a central role in the explanation for these findings. A limitation for our clustering experiments is the low number of stable cell lines that we obtained, thus some caution is warranted in the interpretation of these data.

Fallon 2006;
In Drosophila it has been shown that the regulation of NMJ maturation involves the AKT signaling pathway (Natarajan et al. 2013). In the peripheral nervous system of Gnpat KO mice, a deficiency in the phosphorylation of AKT has been observed and was attributed to inefficient recruitment of AKT to the plasma membrane, where the phosphorylation occurs (da Silva et al. 2014). Accordingly, this defect might also have an influence on NMJ development. However, the amplitudes of mEPPs recorded from NMJs in flies with a genetically induced AKT phosphorylation deficiency were lower than control values (Natarajan et al. 2013) arguing against a defective AKT pathway as the sole explanation for NMJ abnormalities in ether lipid-deficient mice, which in our electrophysiological studies showed a trend toward elevated mEPP amplitudes. In line with our observations in Gnpat KO mice, the AKT mutant flies showed a reduced frequency of mEPPs. Ether lipid deficiency could also impair signaling processes at the level of GPI-anchored proteins. Prominent examples of GPIanchored proteins with designated roles in NMJ development include the matrix metalloproteinase regulator reversioninducing-cysteine-rich protein with Kazal motifs (RECK) (Kawashima et al. 2008) or the receptor tyrosine kinase ligand Ephrin A-2 (Lai et al. 2001), for which a role in the stabilization of the post-synaptic apparatus at the NMJ is discussed (Lai and Ip 2003). The GPI anchor serves as a sorting determinant for the integration into lipid rafts (Brown and Rose 1992;Benting et al. 1999); and the lipid moiety of the anchor itself partly depends on ether lipid biosynthesis (Kanzawa et al. 2009). It still remains unclear, whether the absence of ether lipids impairs the function of all or certain GPI-anchored proteins (Wanders and Brites 2010). Apparently, the activity of GPI-anchored proteins is not fully abolished, but compensatory changes in the levels of these proteins have been demonstrated in cells from human ether lipid-deficient patients (Kanzawa et al. 2012).
Comparisons to other genetically modified mouse models with disturbances in NMJ development or neuromuscular transmission could provide insight into the molecular causes of the NMJ phenotype observed in Gnpat KO mice. For example, mice with a complete inability to form muscle-type AChRs  or a deficiency in choline acetyltransferase (Misgeld et al. 2002), as well as agrin KO mice (Gautam et al. 1996) show abnormal nervous innervation of muscles. In particular, the former two mutants display extensive motor nerve branching and a broadened end plate area, reminiscent of our findings in Gnpat KO mice. Elaborate studies in these mice have revealed a complex regulation underlying pre-and post-synaptic specialization at the NMJ, which is not yet fully understood. Ether lipid deficiency might modulate this network, for example, by weakening signaling cascades like MAP kinase pathways, described to be activated by certain ether lipids (Liliom et al. 1998), or the AKT pathway (da Silva et al. 2014). Phenotypically, this could result in the morphological changes detected in the diaphragm of ether lipid-deficient mice.
We also examined the morphological characteristics of AChR clusters in different skeletal muscles of adult mice. There was a consistent trend toward smaller clusters, as judged by volume as well as surface area, in all muscles analyzed of Gnpat KO mice. However, the differences reached statistical significance only in soleus muscle, and, for the surface area, in gastrocnemius muscle. Soleus muscle is a typical example of a muscle consisting of slow-twitch fibers, whereas tibialis anterior and extensor digitorum longus muscles can be classified as fast-twitch muscles. Interestingly, the two fiber types have been suggested to differ in the head group composition of plasmalogens and, moderately, also in their plasmalogen content (Horrocks 1972;Masoro et al. 1966). Remarkably, the gastrocnemius muscle constitutes an intermediate type containing both types of fibers. The fact that in this muscle, a significantly smaller surface area was found in Gnpat KO mice might indicate that indeed slow-twitch fibers are more affected than fast-twitch fibers. A slight impairment in the clustering process evoked by ether lipid deficiency, for example, via destabilization of lipid rafts (as described above) or other interference with the signaling machinery involved in clustering, might explain our findings.
In order to evaluate the functional consequences of ether lipid deficiency for synaptic transmission at the NMJ, we conducted electrophysiological recordings from adult diaphragm muscles in vitro. Intriguingly, we detected a strikingly reduced rate of spontaneous vesicle fusion, as indicated by the reduced number of mEPPs in Gnpat KO mice. The higher mean amplitude of the mEPPs did not reach statistical significance, whereas that of mEPCs was clearly increased. However, the amplitude of evoked potentials (EPPs) was not affected. These changes were accompanied by a considerable elevation in input resistance likely reflecting altered biophysical properties evoked by the depletion of ether lipids and compensatory changes in lipid composition (Dorninger et al. 2015). The proposed involvement of plasmalogens in membrane fusion processes as well as their fast turnover in gray matter has led to suggestions of a role in neurotransmission and related events (Lohner 1996;Han 2005;Farooqui and Horrocks 2001;Rosenberger et al. 2002). In addition, plasmalogens are major constituents of pre-synaptic membranes (Hofteig et al. 1985) and synaptic vesicles (Takamori et al. 2006); and previous studies using synaptosomes derived from Gnpat KO mice supported the hypothesis of impaired vesicular neurotransmitter release under conditions of ether lipid deficiency (Brodde et al. 2012). In our current hypothetical model, an impairment of fusion and constriction events because of the lack of plasmalogens and the resulting compensatory changes (Dorninger et al. 2015) in vesicular and synaptic membranes would account for the majority of our findings. For example, a reduced rate of membrane fusion is a plausible explanation for the decreased frequency of mEPPs as well as for the reduced number of vesicles released at each stimulated event (quantal content) in ether lipid-deficient mice. Concurrently, when a synaptic vesicle fuses with the pre-synaptic membrane, impaired membrane constriction might lead to the release of a large amount of vesicular content instead of a 'kiss-and-run'-like mode, which like in the central nervous system has been proposed to also occur at the NMJ (Verstreken et al. 2002), thus accounting for the elevation in mEPP and mEPC amplitude. Interestingly, altered mEPP properties, particularly reduced frequency of spontaneous vesicle fusion, have been described in many mouse models with a deficiency in proteins involved in the signaling pathways regulating NMJ development and AChR clustering (Shi et al. 2010;Li et al. 2008;Barik et al. 2014). However, in all of these mice, reduced mEPP frequency goes along with decreased mEPP amplitude. In contrast, an increase in mEPP amplitude has been described in relatively few animal models. One remarkable example is mice with a deficiency in SNAP-25, in which exocytosis of synaptic vesicles is impaired and EPPs cannot be evoked; however, mEPPs were recorded at normal frequency but with considerably increased amplitude (Washbourne et al. 2002). The authors proposed reduced activity of acetylcholinesterase as an explanation for the elevated amplitude of mEPPs rather than compromised exoand endocytosis, as hypothesized in our model.
Alternative to these considerations, which focus on a presynaptic deficit upon ether lipid deficiency, our findings could also originate from post-synaptic alterations. Our data demonstrate a considerable increase in input resistance that might be directly (e.g., by changes in membrane properties because of altered lipid composition) or indirectly (e.g., by altered muscle fiber size) related to ether lipid deficiency. Increased input resistance, as also seen in Gnpat KO mice, has been associated with larger mEPP amplitude in some studies (Katz and Thesleff 1957), but such a correlation has been doubted by others (Gage and McBurney 1973). Furthermore, an inverse relationship between quantal content and mEPP amplitude (or quantal size) has been proposed before (Plomp et al. 1992;DiAntonio et al. 1999;Fong et al. 2010) and would imply that the reduced quantal content at ether lipid-deficient NMJs is an adaptive response resulting from the increased mEPP and mEPC amplitude.
A recent paper describes phenotypic aberrations in the peripheral nervous system of mice carrying a homozygous inactivating mutation in PEX10 (Pex10 CY/CY ), a mouse model of a peroxisome biogenesis disorder (Hanson et al. 2014). Next to defects in Schwann cell morphology and axon integrity, these mice display morphological alterations at their NMJs, in particular, a decreased colocalization of preand post-synaptic markersa phenomenon that we did not encounter in Gnpat KO mice, indicating that the NMJ phenotype of Pex10 CY/CY mice is not because of the lack of ether lipids alone. However, similar to our Gnpat KO mice with isolated ether lipid deficiency, abnormal axonal growth was identified in fetal Pex10 CY/CY diaphragms (Hanson et al. 2014). Intriguingly, electrophysiological recordings in these diaphragms indicated normal shape and number of mEPPs but reduced amplitudes of EPPs, which is in sharp contrast to our results in Gnpat KO mice. One probable determinant for these seemingly conflicting findings is the age of experimental animals. Whereas we used adult mice for our electrophysiological studies, the corresponding experiments in Pex10 CY/CY animals involved fetal stages. Some of the phenotypic aberrations in ether lipid-deficient mice manifest with advanced age; thus, it is conceivable that the defects in neuromuscular transmission that we observed are not yet present in developing Gnpat KO mice. Accordingly, the neuromuscular phenotype in fetal peroxisome-deficient mice might be independent of ether lipids and instead be evoked by one of numerous other metabolic defects resulting from generalized peroxisomal dysfunction.
Since poor performance in the balance beam, rotarod, or muscle strength tasks have frequently been associated with neuromuscular dysfunction (Gomez et al. 1997;Pelkonen and Yavich 2011;Shi et al. 2010), it is tempting to ascribe these deficits to the NMJ abnormalities of ether lipid-deficient mice. However, from our electrophysiological data (particularly the normal EPP amplitude and the unchanged decrement after high frequency pulses), we conclude that there is no major impairment of neuromuscular transmission in these animals. More subtle deficits might become detectable in more detailed analyses as described elsewhere (Sandrock et al. 1997). Given the widespread effects of ether lipid deficiency on the mammalian organism, it is generally problematic to assign a phenotypic feature like motor dysfunction to a particular molecular alteration. We consider it likely that motor impairments in Gnpat KO mice result from the combined presence of cerebellar deficits ) and peripheral myelination defects (da Silva et al. 2014), which may be modulated by some of the abnormalities in NMJ development and transmission as reported here.
Taken together, our results indicate that ether lipid deficiency affects both structure and function of the NMJ. Based on the present findings, the lack of ether phospholipids apparently modulates specific NMJ properties rather than abolishing transmission at the NMJ. declare no conflict of interest. All experiments were conducted in compliance with the ARRIVE guidelines.

Supporting information
Additional Supporting Information may be found online in the supporting information tab for this article: Figure S1. No association between the scores achieved in the weights test and body weight. Figure S2. AChR clusters in fetal diaphragms of WT and Gnpat KO mice. Figure S3. AChR clusters in skeletal muscles of adult WT and Gnpat KO mice. Figure S4. In vivo ligand binding for fluorescence-based evaluation of AChR stability in WT and Gnpat KO mice.
Table S1 . Scoring system for the evaluation of balance beam performance.