Disrupted function of lactate transporter MCT1, but not MCT4, in Schwann cells affects the maintenance of motor end‐plate innervation

Recent studies in neuron‐glial metabolic coupling have shown that, in the CNS, astrocytes and oligodendrocytes support neurons with energy‐rich lactate/pyruvate via monocarboxylate transporters (MCTs). The presence of such transporters in the PNS, in both Schwann cells and neurons, has prompted us to question if a similar interaction may be present. Here we describe the generation and characterization of conditional knockout mouse models where MCT1 or MCT4 is specifically deleted in Schwann cells (named MCT1 and MCT4 cKO). We show that MCT1 cKO and MCT4 cKO mice develop normally and that myelin in the PNS is preserved. However, MCT1 expressed by Schwann cells is necessary for long‐term maintenance of motor end‐plate integrity as revealed by disrupted neuromuscular innervation in mutant mice, while MCT4 appears largely dispensable for the support of motor neurons. Concomitant to detected structural alterations, lumbar motor neurons from MCT1 cKO mice show transcriptional changes affecting cytoskeletal components, transcriptional regulators, and mitochondria related transcripts, among others. Together, our data indicate that MCT1 plays a role in Schwann cell‐mediated maintenance of motor end‐plate innervation thus providing further insight into the emerging picture of the biology of the axon‐glia metabolic crosstalk.


| INTRODUCTION
The peripheral nervous system (PNS) is formed by sensory and motor neurons whose cellular bodies are located in dorsal root ganglia and spinal cord gray matter, respectively. Both types of neurons extend long axonal projections into peripheral tissues such as skin, glands, and muscles where they exert their functions of detecting sensory stimuli and executing motor responses such as muscle contraction.
These long peripheral axons are bundled into nerves that contain Schwann cells (SCs), the myelinating glia of the PNS, and are enveloped by connective tissue generated by fibroblasts. The long distances between the neuronal cell body and the distal regions of the axon, and the fact that axons are isolated from the extracellular environment by surrounding myelinating and non-myelinating SC, raise the question of how neurons obtain energy-rich substrates to sustain their metabolism (Hirrlinger & Nave, 2014;Nave, 2010). It stands to reason that transport of such nutrients from the soma to the periphery would be slow and inefficient (for comparison, transport of glycolytic enzymes occurs at a speed of 0.02 μm/s; Brady & Lasek, 1981), and that direct uptake from the extracellular space at the exposed nodes of Ranvier (less than 1% of the total membrane surface area) might be insufficient to meet the energetic needs of active neurons (Nave, 2010). One possibility is that SC are not simply insulators and passive spectators of neuronal function, but instead they act as metabolic intermediaries, taking up glucose from the blood or interstitial space, converting it to small energetic metabolites such as lactate and pyruvate and finally shuttling these substrates to the underlying neuron, where they can be used for mitochondrial oxidative phosphorylation and energy production.
In the CNS, oligodendrocytes and astrocytes express monocarboxylate transporters (MCT) 1 and 4 (1 is expressed in both cell types, 4 in astrocytes exclusively). These glial cells appear to be metabolically coupled to neurons in some CNS regions such as the spinal cord, brain cortex, and hippocampus, providing them with lactate that can be taken up by high-affinity neuronal MCT2 (Lee et al., 2012;Machler et al., 2016;Pierre, Magistretti, & Pellerin, 2002;Suzuki et al., 2011). Other regions, like the corpus callosum, seem to rely primarily on glia-derived glucose (Meyer et al., 2018). MCTs are proton-coupled transporters of monocarboxylates (lactate, pyruvate, and ketone bodies) (Halestrap, 2013). They present different affinities to their substrates (MCT2 has the highest affinity, followed by MCT1 and MCT4 having the lowest affinity) and are expressed in different cell types, which may be related to the cells status as net producers or consumers of monocarboxylates (Halestrap, 2013). In the PNS, SC express MCT1, 2 and 4 whereas DRG neurons express MCT1 and 2 (Domenech-Estevez et al., 2015;Morrison et al., 2015). The presence of different MCTs on both partners in the PNS suggests a metabolic coupling similar to the one operating in the CNS. In order to evaluate the possible role of MCTs in the crosstalk between glia and underlying axons we generated and characterized two conditional knockout models in which MCT1 or MCT4 were eliminated specifically in SC using MPZ-Cre-driven recombination of floxed MCT alleles. We find that MCT1 (but not MCT4) expression in SC is necessary for maintenance of neuromuscular junction innervation. This is associated with transcriptional changes in lumbar motor neurons of MCT1 conditional knockout mice, suggesting that SC's metabolic support to axons is necessary for proper maintenance of motor neurons.

| Western blotting and immunohistochemistry of teased fibers
To further validate MCT1 and MCT4 inactivation, sciatic nerve endoneuria were lysed using a Tissue Lyser II apparatus and stainlesssteel beads to obtain protein extracts (lysis buffer containing 5% SDS, 5 mM EDTA and 80 mM Tris-HCl, supplemented with 1X cOmplete™ Mini EDTA-free Protease Inhibitor Cocktail (Roche), 1 mM NaF and 1 mM NaVO 4 ). Between 20 and 50 μg of proteins (quantified using the Pierce BCA Protein Assay Kit, Thermo Scientific) were separated by SDS-PAGE on 12.5% polyacrylamide gels and transferred onto Immobilon-FL (Millipore) or nitrocellulose (Thermo Scientific) membranes. Membranes were blocked in 5% milk in tris-buffered saline (TBS) with 0.05% Tween-20 for 1 hr at room temperature, washed and incubated with primary antibody (rabbit anti-MCT1 [Stumpf et al., 2019] or rabbit anti-MCT4 [Santa Cruz]) for 48 or 24 hr, respectively, at 4 C with agitation. After washing, membranes were incubated with secondary antibodies for 2 hr at room temperature, scanned using the Odyssey Infrared Imaging System and analyzed with Image Studio Software (Li-Cor, Inc.). Membranes were reblocked, stained for tubulin and imaged the next day. For teased fiber preparation from MCT4 cKO mice, sciatic nerves were freshly dissected and fixed in 4% paraformaldehyde for 15 min at room temperature. After washing in PBS, nerves were cut into smaller pieces, stripped of peri-epineurium and teased on TESPA-coated slides using fine insect pins. Teased fibers were allowed to dry and kept at −80 C until the next day. For immunohistochemistry, fibers were permeabilized with Triton X100 0.1% for 15 min at room temperature, blocked in a solution of 5% BSA, 1% NGS, and Triton X100 0.1% for 1 hr and incubated with rabbit anti-MCT4 primary antibody (Santa Cruz) overnight at room temperature. After washing, fibers were incubated with a secondary antibody, counterstained with DAPI and mounted with Vectashield mounting medium. Antibody references and dilutions can be found in Supplemental Table 2.

| Electron microscopy and analysis of sciatic nerve structure
One-year old mice were anesthetized with isoflurane and sacrificed by decapitation. Sciatic nerves and spinal roots were dissected and immediately fixed in 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer for 2 hr at room temperature, followed by overnight washing in the same buffer. Samples were stained for 4 hr in 1% OsO 4 in cacodylate buffer followed by dehydration in increasing concentrations of ethanol. Then, tissues were infiltrated with propylene oxide and embedded in Epoxy Embedding Medium (Sigma) (modified from Bartesaghi et al., 2015).
For g-ratio quantification, samples were sectioned at 0.5 μm thickness using an Ultracut E (Reichert-Jung) ultramicrotome and stained with toluidine blue in 0.1 M sodium borate buffer. Images were acquired using Zeiss Axioskop 40 light microscope at ×100 magnification and g-ratio determined using Fiji software (Schindelin et al., 2012) and g-ratio calculator plugin as previously described (Arnaud et al., 2009). An average of 859 fibers of one sciatic nerve section per sample were quantified, three samples per genotype were used.
For analysis of Remak bundles, samples were sectioned at 50-60 nm thickness using a Leica EM UC7 ultramicrotome (Leica, Wien, Austria), transferred onto formvar stabilized slot grids and contrasted with uranyl acetate followed by lead citrate. Sections were examined in a Tecnai Spirit BioTWIN transmission electron microscope (FEI Company, Eindhoven, The Netherlands) operated at 100 kV. Digital images were acquired using a 2kx2k Veleta CCD camera (Olympus Soft Imaging Solutions, GmbH, Münster, Germany).
Images of ×1,200 magnification were used to determine the area occupied by Remak bundles (one section per sample was quantified, three samples per genotype). Images of ×11,000 magnification were used to determine number of small unmyelinated fibers per bundle and the diameter of the largest unmyelinated fibers observed. Thirty to fifty randomly selected bundles of one section per sample were quantified, three samples per genotype were used. The experimenter was blinded to the samples' genotype during image analysis. Statistical analysis was performed using an unpaired two-tailed t-test with Welch's correction.

| Motor and sensory nerve conduction studies
Nerve conduction velocity studies were performed using a Nicolet Viking electromyograph system. One-year old animals were anesthetized intraperitoneally with a mixture of ketamine and xylazine. Animals were positioned face down on a styrofoam surface and platinum needle electrodes were inserted in the interdigital muscles to record motor nerve conduction velocity. Stimulating electrodes were placed near the Achilles tendon and sciatic notch and a supramaximal stimulation was applied at 0.5 Hz and 0.1 ms duration. Velocity was calculated by dividing the difference in response latency for each stimulation by the distance between the stimulation sites. Two to three separate measurements were recorded on each leg and the results are expressed as mean velocity per animal. Sensory nerve conduction velocity was recorded subdermally at the base of the tail with stimulation distally at 3 cm, with supramaximal stimulation of 0.7 Hz frequency and 0.1 ms duration. Results are mean of 10 sweeps for a single measurement per animal. All measurements were performed at room temperature. After measurements, animals were sacrificed for tissue collection.

| Motor and sensory behavioral tests
Motor behavior was assessed using the CatWalk system (Noldus Information Technology, The Netherlands) as previously described (Angeby Moller et al., 2018). Briefly, 1-year old mice of both sexes were habituated to the room for 30 min and then allowed to cross the Catwalk system three times. The mouse's home cage was used as bait at the opposite end of the CatWalk. After two training sessions on separate days, baseline measurements were recorded using the Catwalk XT software version 10.6. The parameters analyzed are mentioned in the results section. Mechanical hypersensitivity was assessed using the Marstock OptiHair filaments (Marstock, Germany) and the up-down testing method was used (Chaplan, Bach, Pogrel, Chung, & Yaksh, 1994;Dixon, 1980). Briefly, mice were habituated to a testing enclosure with wire mesh bottom for about an hour prior to testing. Withdrawal thresholds were assessed by a set of filaments with increasing forces, applied perpendicularly to the center of the plantar surface of the hind paw for a maximum of 5 s, or until a brisk response was observed. Results are expressed as 50% threshold values in grams (g), presented as average of the values taken from both hind legs. Mice were habituated to the testing environment on at least two separate occasions. Cold sensitivity was assessed using the acetone drop test. Mice were habituated to the same enclosure with a wire mesh bottom, where one acetone drop was applied gently to the plantar side of one hind paw with the aid of a 1 ml syringe. The time spent with nocifensive reaction (licking, shaking, biting the hind leg) was recorded for 60 s. The test was repeated three times on each leg with 30 min test-free period between sessions. Data are presented as time spent reacting to acetone in seconds (s). Mice were habituated to the testing environment on at least two separate occasions. Heat sensitivity was assessed using a modified Hargreaves box (Dirig, Salami, Rathbun, Ozaki, & Yaksh, 1997). Mice were placed on the glass surface in Plexiglas containers and allowed to habituate to the test environment prior to testing. Paw withdrawal, observed as a brisk reaction, was evoked by an increasing light stimulus positioned to the middle of the footpad. To prevent thermal exposure damage, the stimulus is turned off at a cut-off time of 20 s or 55 C. Data are presented as paw withdrawal latency in seconds. After assessing the withdrawal latency for one paw of all mice, the second paw was measured and this was repeated three times, with 30 min test-free period inbetween sessions. Average of both legs was calculated. Mice were habituated to the testing environment on at least two separate occasions. The experimenter was blinded to the animals' genotypes for the duration of the tests.

| Immunohistochemistry and analysis of distal innervation
One-year old animals were sacrificed and dissected to collect gastrocnemius muscle and hind paw skin for cryosectioning and immunohistochemistry. Muscles of 6-month old animals were also obtained for confirmation analysis of muscular innervation. Briefly, gastrocnemius muscle was fixed in 4% PFA in PBS for 20 min, washed in PBS and cryopreserved in 30% sucrose in PBS overnight, prior to embedding in optimal cutting medium (OCT). Similarly, skin from the hind paws (containing the walking pads) was dissected, fixed in Zamboni's fixative for 2 hr, washed and cryopreserved before embedding in OCT.
Samples were stored at −80 C until further processing. Tissues were sectioned in a Cryostar NX70 cryostat at −20 C and collected on Superfrost Plus glass slides (25 μm longitudinal sections for muscle, 50 μm cross-sections for skin). Muscle sections were permeabilized and blocked in a solution of 5% BSA, 1% NGS and triton 0.1% for 4 hr, followed by overnight incubation with primary antibodies at 4 C. Following washing, samples were incubated with the secondary antibody and fluorescently labeled bungarotoxin for 2 hr at room temperature, counterstained with DAPI and mounted with Vectashield mounting medium. Skin sections were incubated on a slide in blocking solution (1% NGS, 0.15% Triton X-100 in PBS) for 4 hr at room temperature, followed by overnight incubation with the primary antibody at 4 C and the secondary antibody 2 hr at room temperature. Sections were counterstained with DAPI and mounted with Vectashield. Primary and secondary antibodies and stains can be found in Supplemental Table 2.
To characterize the innervation status of neuromuscular junctions (NMJ), ×20 magnification images were acquired using a Zeiss Observer Z1 epifluorescence microscope and an ORCA-Flash4.0 LT camera (Hamamatsu Photonics, Japan). One hundred randomly selected NMJs per sample were classified as fully innervated (>30% overlap between bungarotoxin staining and SV2/NF145), partially denervated (<30% overlap) or fully denervated (0% overlap) (the number of samples is indicated in results). To determine the intraepidermal nerve fiber density (IENFD) in hind paw skin, ×20 magnification images were obtained using a Zeiss LSM800 confocal microscope (pinhole 12 μm). Two to three sections spanning different regions of the hind paw were analyzed per sample. The dermis-epidermis interface (stratum basale) was identified using DAPI staining and all PGP9.5-positive fibers crossing the interface were counted (free standing fiber fragments and branches were not considered as independent fibers). The linear skin length was measured at the upper layer of epidermis and the IENFD calculated by dividing the number of fibers by the linear skin length (expressed as fibers/mm). The experimenter was blinded to the samples' genotype during image analysis.
Statistical analysis of NMJ innervation was performed using a chisquared test for categorical variables and analysis of IENFD was calculated using an unpaired two-tailed t-test with Welch's correction.
p-values are indicated in figures and statistical significance was recognized when p < .05.

| LCM and RNAseq of lumbar spinal cord motor neurons
Lumbar spinal cords of 1-year old mice were freshly dissected and immediately frozen in liquid nitrogen for the laser capture microdissection of motor neuron cell bodies. For LCM-seq and polyA-basedsequencing analyses all steps were carried out as previously described (Nichterwitz et al., 2016) with minor modifications. Briefly, spinal cord tissue was sectioned at 18 μm thickness onto PEN membrane glass slides (Zeiss) and a quick histological Nissl staining based on the Arcturus Histogene Staining Kit protocol was performed prior to capture.
Pools of approximately 70 cells per sample were captured under a Leica DM6000R/CTR6500 microscope using the Leica LMD7000 system.
Collected cells were lysed in 5 μl of 0.2% Triton-X100 (Sigma-Aldrich, in water) with 2 U/μl of RNAse inhibitor (Takara) and 1 μM of DTT (Invitrogen). The library preparation protocol described in Nichterwitz et al. is based on the Smart-seq2 protocol (Picelli et al., 2013) and was used to directly prepare cDNA without prior RNA extraction. Sequencing libraries were prepared using a Nextera XT DNA library preparation The sequencing reads were aligned to the mm10 mouse genome using STAR version 2.7.0e (Dobin et al., 2013) with parameter "-outSAMstrandField" set to "intronMotif." Aligned reads were assigned to genomic features with rpkmforgenes.py (Ramskold, Wang, Burge, & Sandberg, 2009) using Ensembl version GRCm38.95 as annotation. Parameters used in rpkmforgenes.py were "-fulltranscript" and "-rmnameoverlap." Differential expression was analyzed in R (version 3.4.4) with DESeq2 version 1.24.0 (Love, Huber, & Anders, 2014). Default settings were used and an FDRadjusted p-value <.05 was considered significant. No genes were excluded a priori before the DESeq2 procedure.

| RESULTS
3.1 | Generation of MCT1 and MCT4 SC-specific conditional knockout mouse models In the CNS, expression of MCT1 by oligodendrocytes and MCT1/4 by astrocytes is necessary for the metabolic support of neurons. Namely, MCT1 depletion in oligodendrocytes leads to axonal damage and it is reduced in patients and animal models of amyotrophic lateral sclerosis (Lee et al., 2012). Astrocytic MCT4, on the other hand, is essential for long-term memory formation (Suzuki et al., 2011)  well as by immunohistochemistry on MCT4 teased sciatic nerve fibers (Supplemental Figure 1).

| MCT1 and MCT4 cKO mice do not present detectable alteration in their PNS function
Mutant mice developed overall normally, were fertile, and lived for as long as their wild-type littermates. To characterize the function of peripheral nerves, we performed nerve conduction velocity studies on both motor (MNCV) and sensory (SNCV) tracts in 1-year old animals.
MNCV measurements were performed on the sciatic nerve while SNCV was measured on the tail. We did not detect any measurable difference between control and cKO animals in either MNCV or in SNCV (Figure 2a,b). To complement these observations, we per-

| Glial MCT1 is required for long-term maintenance of motor end-plate integrity
To detect potential structural changes in the PNS, we analyzed semithin sections of sciatic nerves from MCT1 cKO or MCT4 cKO mice.
Samples from young animals (P10 and 4 months) showed no observable differences in the structure of peripheral nerve tracts in the absence of MCT1 or MCT4 (data not shown) leading us to conclude that nerves develop normally and, thus, to focus on a more in-depth analysis of 1-year old animals. Also in the aged animals, myelin and axons were preserved in both lines with no detectable differences in g-ratio and distribution of axonal calibers (Figure 3). The number of sciatic nerve myelinated fibers and the g-ratio of spinal roots were also unchanged (Supplemental Figure 3). Electron microscopy analysis of Remak bundles also failed to reveal any differences between control and cKO mice. The area occupied by unmyelinated fibers in We hypothesized that the impact of ablating MCT expression in SC could be affecting neurons in a subclinical manner that cannot be detected in the proximal/central part of the sciatic nerve or behavior of cKO mice. We therefore analyzed the most distal part of sciatic nerve, concentrating on the innervation status of neuromuscular junctions (NMJ) in the gastrocnemius muscle and the density of free nerve endings in the hind paw skin of 1-year old mice. We focused on MCT1 cKO since MCT1 has higher level of expression in SC (Domenech-Estevez et al., 2015;Halestrap, 2013). We found that the cKO as compared to control (Figure 5a). However, the percentage of fully innervated NMJ was reduced and accompanied by a concomitant increase in fully denervated and partially denervated NMJs, revealing that MCT1 expressed by SC has a role in the maintenance of motor end-plate innervation (Figure 5b). This effect had a late onset, as changes in NMJ innervation were not detected at 6-months of age (Supplemental Figure 4).

| DISCUSSION
In this work, we described two conditional knockout models in which MCT1 or MCT4 were eliminated specifically in SCs. We found that absence of MCTs from SC does not affect the development of the PNS, as these mice present normal myelin, nerve conduction velocities and motor and sensory behavior. Interestingly, we discovered that MCT1 (but not MCT4) expression in SC is necessary for long-term maintenance of neuromuscular junction innervation and is associated with transcriptional changes in lumbar motor neurons, suggesting that SC metabolic support to axons is necessary for proper maintenance of motor neuron function. These neurons seem to compensate for the absence of SC-mediated support by upregulating genes involved in mitochondrial function, regulation of transcription and glycosylation, which is important for the function of many transmembrane proteins such as receptors of neurotransmitters and for cell adhesion. Of particular interest, Erk3, Ndufs8, and Fis1 were found to be upregulated, while Tuba1a, Rnd2, G3bp1, and Akip1 were downregulated. Erk3, also known as Mapk6, is an atypical mitogen-activated protein kinase (MAPK) and is involved in dendritic spine formation and development of the nervous system (Brand et al., 2012); Ndufs8 is an essential subunit of mitochondrial NADH:ubiquinone oxidoreductase, or complex I (Procaccio et al., 1997), and its presence is necessary for oxidative phosphorylation, and Fis1 is a positive regulator of mitochondrial fission (Loson, Song, Chen, & Chan, 2013). Tuba1a (α-Tubulin 1A) is a component of the neuronal cytoskeleton (Pubmed gene ID 22142) (Yue et al., 2014) and Rnd2 is a member of the Rho GTPase family involved in regulation of actin cytoskeleton, and is involved in dendrite branching (Negishi & Katoh, 2005). Akip1 promotes the phosphorylation of NFkB subunit p65/RelA by PKA, increasing its retention in the nucleus and enhancing expression of NFkB target genes (Gao, Asamitsu, Hibi, Ueno, & Okamoto, 2008), and conditional knockout of RelA in neurons promotes their regeneration (Haenold et al., 2014), while G3BP1 regulates the formation of stress granules and prevents axonal regeneration (Sahoo et al., 2018). The upregulation of mitochondria-related genes could be a response to reduced metabolic support from SC. The downregulated cytoskeletal genes support the observation of denervation while downregulation of Akip1, a regulator of NFkB signaling via RelA, and G3BP1, an inhibitor of axon regeneration, could be an attempt at preservation of axons (Haenold et al., 2014;Sahoo et al., 2018). The mild effect observed is not surprising, given the fact that non-demyelinating peripheral neuropathies can often start in the later decades of life in humans, progress in a distal-to-proximal fashion and present widely ranging symptoms, from mild foot deformities and difficulty in running, to needing walking aids and an inability to evoke motor potentials in lower limbs (Senderek et al., 2000). It is possible that the time point analyzed in this work, which corresponds to human middle age (approximately 40 years old) (Flurkey, Currer, & Harrison, 2007), reflects the onset of the phenotype which may worsen with aging.
Recent work by Jha and colleagues (Jha et al., 2019) described a SC-specific MCT1 conditional knockout (generated by deletion of exon 2) with a predominantly sensory phenotype. These mice presented progressive thinning of myelin in the sural nerve and increased nodal length, which correlated with reduced SNCV and mechanical sensitivity. This was linked to altered lipid metabolism, reduced expression of MAG and increased expression of c-Jun, suggesting that MCT1 cKO SC may be regressing to less mature states. The authors suggested that ablation of MCT1 from SC was not sufficient to affect glial support to axons. Our observations revealed transcriptional and localized structural changes in the terminals of axons that are in contact with MCT1-deficient SCs which were not characterized in the other model (Jha et al., 2019). It is also possible that the differences observed between these two studies could be due to the different genetic backgrounds of the models. Considering the relatively mild phenotype we observed, we hypothesize, as Jha et al. also defended, that the remaining MCTs expressed by SC can compensate partially for the MCT1 deficiency. This hypothesis should be further evaluated by generation and characterization of a model missing multiple or all MCTs in SC.
The past decade has been rich in discoveries related to metabolic interactions between glia and neurons. In mammalian systems, the astrocyte-to-neuron lactate shuttle (Pellerin et al., 2007) is essential for long-term memory formation (Suzuki et al., 2011) and oligodendrocytes stimulated with NMDA receptor agonists are able to sustain axonal activity ex vivo, linking neuronal activity with glial support (Saab et al., 2016). In the PNS, the knockout of major metabolic regulator LKB1 in SC was shown to lead to axonal degeneration without myelin loss and was accompanied by a neuroprotective increase in lactate release (Beirowski et al., 2014). Our present data show that MCTs expressed by peripheral myelinating glia, which are able to transport lactate, are important for long-term maintenance of motor end-plate innervation, independently of myelin itself. Terminal SC (Griffin & Thompson, 2008) could also play a role in this process; however, the previously published data indicate that terminal SC at neuromuscular junctions do not express MCT1 (Morrison et al., 2015). To

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.