Schwann Cell Glycogen Selectively Supports Myelinated Axon Function

Abstract

Objectives

Interruption of energy supply to peripheral axons is a cause of axon loss. We determined if glycogen was present in mammalian peripheral nerve, and if it supported axon conduction during aglycemia.

Methods

We used biochemical assay and electron microscopy to determine the presence of glycogen, and electrophysiology to monitor axon function.

Results

Glycogen was present in sciatic nerve, its concentration varying directly with ambient [glucose]. Electron microscopy detected glycogen granules primarily in myelinating Schwann cell cytoplasm and these diminished after exposure to aglycemia. During aglycemia, conduction failure in large myelinated axons (A fibers) mirrored the time-course of glycogen loss. Latency to CAP failure was directly related to nerve glycogen content at aglycemia onset. Glycogen did not benefit the function of slow-conducting, small diameter unmyelinated axons (C fibers) during aglycemia. Blocking glycogen breakdown pharmacologically accelerated CAP failure during aglycemia in A fibers, but not in C fibers. Lactate was as effective as glucose in supporting sciatic nerve function, and was continuously released into the extracellular space in the presence of glucose and fell rapidly during aglycemia.

Interpretation

Our findings indicated that glycogen is present in peripheral nerve, primarily in myelinating Schwann cells, and exclusively supports large diameter, myelinated axon conduction during aglycemia. Available evidence suggests that peripheral nerve glycogen breaks down during aglycemia and is passed, probably as lactate, to myelinated axons to support function. Unmyelinated axons are not protected by glycogen and are more vulnerable to dysfunction during periods of hypoglycemia.

Introduction

There is growing appreciation that peripheral nervous system (PNS) pathology can be a consequence of disturbed energy metabolism ¹. Axon loss due to hypoglycemia is an obvious example ^{2, 3} but breakdown in normal, but poorly understood, metabolic interactions between Schwann cells and axons, especially myelinated axons, appears to underlie other forms of neuropathy ⁴. Glycogen, the storage form of glucose, is present in the central nervous system (CNS) and can provide temporary protection against hypoglycemic axon injury ^5–7. Glycogen’s role in the PNS, if any, has not been studied and could have important implications for axon survival during periods of reduced glucose.

Glycogen molecules contain 20 to 50 thousand glucose residues that can be quickly released for use in energy metabolism. The highest concentrations of glycogen are in liver and muscle, but it is also present in the central nervous system (CNS). The role of CNS glycogen, contained exclusively in astrocytes ^{8, 9}, has only recently been determined. In the absence of glucose, or during periods of intense neural activity when fuel demand for ATP production exceeds supply, astrocyte glycogen is broken down to lactate ^10–12, which is released into the extracellular space, taken up by nearby neurons or axons and oxidatively metabolized ^{5, 13}. CNS glycogen, therefore, serves a vital role in ‘energy substrate buffering’.

There have been no systematic studies of glycogen in the PNS; it is not even known if normal mammalian peripheral nerves contain glycogen. The only reports of glycogen in peripheral nerves derive from studies of disease states. Glycogen appeared to be expressed in both axons and Schwann cells in human pathology samples from patients with diabetes and mitochondrial disease ^14–16. Although glycogen has not been directly measured under normal conditions, immunohistochemical studies have reported the presence of glycogen phosphorylase, the key enzyme for glycogen mobilization, in axons of the sciatic and vagus nerves ^{17, 18}. While the presence of this enzyme predicts glycogen content, it does not prove this point. In fact, glycogen in peripheral axons seems unlikely given that neurons and axons in the adult CNS do not contain glycogen ^{8, 9}. Glycogen accumulation appears to be toxic to neurons, and glycogen synthase is inhibited by neuron-specific proteins ¹⁹. If peripheral nerve axons did contain glycogen, this would indicate a fundamental difference in how glycogen supports neural function in the central and peripheral nervous systems.

We studied mouse sciatic nerve, a typical mammalian peripheral nerve, to determine if glycogen was present. It was present, which prompted a search for where it was located and its potential functional role(s). Mouse sciatic nerve is a mixed nerve composed of myelinated and unmyelinated axons of sensory and motor origin. In contradiction to the immunohistochemical data that localized glycogen phosphorylase exclusively to axons ^{17, 18}, we found this enzyme in Schwann cells as well, and our electron microscopy data localized glycogen primarily within the cytoplasm of Schwann cells. As in CNS white matter, glycogen in the PNS supported axon excitability (mainly in myelinated axons) during aglycemia through a mechanism that appeared to involve lactate. These novel findings draw attention to an overlooked set of issues about peripheral nerve energy metabolism that has clinical implications.

Methods

Electrophysiology

Adult male CD-1 mice (28 – 35g) were obtained from Charles Rivers. All experiments were carried out in accordance with UK Home Office guidelines on humane treatment. Mice were killed by cervical dislocation and sciatic nerves dissected free and cut where they emerge from the spinal cord, and just prior to branching into the tibial and peroneal nerves in the thigh region. The sciatic nerves were placed in an interface perfusion chamber (Medical Systems Corp, Greenvale, NY, USA), maintained at 37°C and superfused with artificial cerebrospinal fluid (aCSF) containing (in mM): NaCl 126, KCl 3.0, CaCl₂ 2.0, MgCl₂ 2.0, NaH₂PO₄ 1.2, NaHCO₃ 26 and glucose 10. The chamber was continuously aerated by a humidified gas mixture of 95% O₂ / 5% CO₂. Nerves were allowed to equilibrate in standard aCSF for about 30 minutes before beginning an experiment. Suction electrodes back-filled with the appropriate aCSF were used for stimulation and recording. During an experiment, the CAP was elicited every 30 seconds. The signal was amplified 10× by an Axoprobe 1A amplifier, then amplified a further 100× by Stanford Research Systems Preamplifier (SR560), filtered at 10 kHz and acquired at 20 kHz (Clampex 9.2, Molecular Devices, Wokingham, UK).

Data Analysis

Sciatic nerve axon excitability was monitored quantitatively as the amplitude of the CAP. The sciatic nerve CAP had two discrete components, a large initial peak due to large myelinated axons (A fibers) that occurred within 1 ms of the stimulus artifact, and a much smaller and later component due to small unmyelinated axons (C fibers) that occurred between 5 and 15 ms after the stimulus artifact (see Results). Stimulus pulse strength (30 µs duration, WPI A320 Isostim, WPI, Stevenage, UK) was adjusted to evoke the maximum amplitude of the A peak or the C peak, as appropriate. The amplitude of the CAP represents the number of active axons because currents generated by individual axons within a fiber tract are considered to sum algebraically ^{20, 21}. In experiments where aglycemia caused CAP loss, the latency to onset of CAP failure was monitored as the time from insult onset to the point when the CAP had diminished to 95% of its baseline value, as previously described ⁷.

Lactate biosensor

Lactate and null biosensors ²² were obtained from Sarissa Biomedical Ltd (Coventry, UK). In practice, the lactate signals were so large, relatively, that subtraction of the null signal had no effect on the signal amplitude. The lactate biosensors (25 µm diameter and 500 µm length) were pressed against the sciatic nerve and allowed to equilibrate for 30 to 60 minutes prior to experiments. At the end of a recording the biosensors were calibrated using lactate concentrations of 10, 100 and 1000 µM.

Glycogen assays

Glycogen in sciatic nerves was quantitatively measured using a standard technique ⁵. Briefly, sciatic nerves were transferred to ice-cold 85% ethanol/15% 30 mM HCl solution to prevent glycogen metabolism. Nerves could be stored indefinitely at −30°C prior to assay. A total of 8 sciatic nerves (4 animals) were pooled for each sample (weight of each nerve was ~6 mg). Nerves were washed in 30 mM HCl and then transferred to 400 µl ice-cold 0.1 M NaOH/0.01% SDS. The tissue was sonicated to produce a homogenous slurry and re-acidified by adding 54 µl 1 M HCl. The homogenate was divided into two equal volumes, one for glycogen content the other for protein quantification. The protein assay was performed using the BCA protein assay (BioRad) and a spectrophotometer (SpectaMax, Molecular Devices).

Glycogen was quantified by digesting tissue glycogen to glucose, then assaying for glucose by subtracting the background glucose levels of the tissue. This was performed by first further splitting the homogenate for each sample into two 100 µl aliquots. To the first aliquot, 200 µl of a buffer solution containing amylo-α-1,4-α-1,6-glucosidase (AG: EC 3.2.1.3) at a concentration of 0.30 units/ml was added. 200 µl of the same buffer solution without the enzyme was added to the second aliquot and samples were then gently agitated at 37°C for 1 hour. AG hydrolyses glycogen to glucose, thus the sample left untreated by the enzyme provided the background tissue glucose concentration. All samples were then centrifuged at 10 krpm for 15 minutes at 4°C. A common enzymatic method was used for glucose determination that relies on monitoring the conversion of NADP to NADPH by the enzyme hexokinase ²³. Using a 96-well micro-titer plate, 28 µl of supernatant from the digested tissue sample and 210 µl of a reaction buffer containing 50 µM NADP, 260 µM ATP and 3.9 mM Mg²⁺ were added. The reaction was initiated by addition of an enzyme mixture consisting of glucose-6-phosphate dehydrogenase (EC 1.1.1.49; 0.11 units/ml) and hexokinase (EC 2.7.1.1; 0.3 units/ml). Increases in NADPH concentration were detected by monitoring absorbance on a spectrophotometer at 340 nm. Comparison with a standard curve of glucose (0 – 200 µM) allows calculation of the concentration of glucose in the samples after subtracting the background glucose concentrations. This value was expressed relative to the protein concentration obtained from the assay detailed above for each tissue sample, giving tissue glycogen content as pmol glucosyl units per µg protein.

Tissue processing for microscopy

Adult CD-1 mice were deeply anesthetized with ketamine/xylazine (100/10 mg/kg, i.p.) and perfused through the left ventricle with cold PBS and then a fixative solution containing 4% paraformaldehyde in 0.14 M Sorensen’s phosphate buffer, pH 7.4. Sciatic nerves were removed and immersion fixed for a total fixation time of 20–25 minutes, rinsed with PBS, cryoprotected in 30% sucrose in PBS, and frozen in O.C.T. compound (Tissue-Tek, Torrance, CA).

Immuncytochemistry

Sciatic nerves were processed for immunocytochemistry as previously described ²⁴. Briefly, 10 µm cross-sectional cryosections of the sciatic nerves were cut and mounted on Fisher Superfrost Plus glass slides, and the sections were processed for immunofluorescent detection of glycogen phosphorylase, S-100 and neurofilament. Sections were incubated sequentially in: (1) blocking solution (3% cold fish skin gelatin (Sigma, St. Louis, MO), 3% normal donkey serum (Sigma), and 0.1% Triton X-100 (Sigma) in PBS) for 15 min at room temperature, (2) primary antibodies (goat anti-glycogen phosphorylase (1:100, Santa Cruz Biotechnology, Santa Cruz, CA), mouse anti-S-100 (1:200, Abcam, Cambridge, MA), and chicken polyclonal NF-H (1:5000, Encor, Alachua, FL)) in blocking solution overnight at 4°C, (3) PBS, 6 × 5 min each, (4) secondary antibodies (donkey anti-goat-488 (1:500, Jackson ImmunoResearch, West Grove, PA), donkey anti-mouse-549 (1:500, Jackson ImmunoResearch), and donkey anti-chicken-649 (Jackson ImmunoResearch)) in blocking solution for 6–8 hours at room temperature, (5) PBS 6 × 5 min each. Sections were mounted with Aqua Poly/mount (Polysciences, Warrington, PA) and images were acquired with a Nikon C1 confocal microscope (Nikon, USA, Melville, NY).

Electron microscopy

Following electrophysiological recordings, sciatic nerves were removed from chambers and fixed with 4% glutaraldehyde in 0.14 M Sorensen’s phosphate buffer, post-fixed with OsO₄, and embedded in plastic resin according to standard procedures. Semi-thin and ultrathin cross-sections of sciatic nerves were cut. Ultra-thin sections were stained with uranyl acetate and lead citrate and examined with a JEM-1011 electron microscope (JEOL USA, Peabody, MA) operating at 80 kV. Images were acquired at 10,000× magnification with an AMT-TR-111 digital camera (Advanced Microscopy Techniques, Danvers, MA). For each sample (normal: n = 2, and 0 glucose: n = 2) 12 fields encompassing juxta-nuclear cell bodies of myelinating Schwann cells were acquired. Levels of glycogen granules in these samples were assessed by two observers blinded to experimental conditions. Glycogen granules in Schwann cell cytoplasm were counted and the area of the cytoplasm containing the granules measured using Openlab software (Improvision). Glycogen granule density was expressed as granules per µm² of cytoplasm.

Data analysis

Data are presented as means and standard deviation and statistical tests were carried out using Student’s t-test or one-way ANOVA with Dunnett’s post-hoc test as appropriate.

Results

Glycogen content of sciatic nerve

Glycogen was present in sciatic nerve. Glycogen content of acutely isolated sciatic nerve was 8.61 ± 0.90 pmoles µg protein⁻¹ (n = 3; blood glucose concentration at sacrifice was 9.50 ± 1.80 mM, n = 18, similar to resting glucose levels previously reported in CD-1 mice) ²⁵,, significantly higher than in acutely isolated optic nerve (6.52 ± 0.40 pmoles µg protein⁻¹; p < 0.037), a CNS white matter tract ⁵.

Glycogen content was stable for at least 6 hours in sciatic nerves incubated in 10 mM glucose aCSF (data not shown). Glycogen content in CNS white matter varied with ambient glucose concentration ⁵. We therefore assessed the effect of bath glucose concentration on peripheral nerve glycogen content. We incubated sciatic nerves for two hours in aCSF with variable glucose concentrations ranging from 2 to 30 mM. A pre-incubation period of 60 minutes in 10 mM glucose aCSF preceded each experiment to allow nerves to equilibrate under identical conditions. In nerves bathed in 2, 4, 7, 10 or 30 mM glucose for two hours, the glycogen content was 2.35 ± 0.07 (n = 3), 7.16 ± 0.53 (n = 3), 9.68 ± 1.08 (n = 3), 10.98 ± 1.00 (n = 3) or 15.21 ± 0.98 (n = 3) pmoles µg protein⁻¹, respectively (Fig 1A). These data revealed an asymptotic relationship between ambient glucose concentration and glycogen content, similar to that seen in mouse optic nerve ⁵, but with a steeper slope (Fig 1B).

Figure 1.

Glycogen content of sciatic nerves. A. Nerves bathed for 2 hours in 2, 4, 7, 10 or 30 mM glucose showed increasing concentrations of glycogen (open squares: left axis - pmoles µg protein⁻¹). Sciatic nerves removed from mice, with a mean blood glucose concentration of 9.5 ± 1.8 mM at the time of death, and immediately frozen for glycogen assay had a glycogen content of 8.61 ± 0.90 pmoles µg protein⁻¹ (grey square). B. [Glucose] plotted on a logarithmic scale demonstrating the exponential relation between glycogen content and ambient [glucose]. Left axis as main figure. The open circles and dotted line denote historical data from the mouse optic nerve ⁵, for comparison.

The presence of glycogen in peripheral nerve prompted two questions: (1) where is the glycogen located, and (2) can this glycogen provide a utilizable energy substrate to sustain axonal excitability in the absence of glucose? The cellular location of glycogen was investigated using immunohistochemistry and electron microscopic techniques. To investigate the second question we studied the temporal correlation between glycogen content and the ability of the nerve to conduct action potentials during glucose deprivation.

Ultrastructural identification of glycogen in sciatic nerve

Electron microscopic (EM) studies were carried out on immersion fixed sciatic nerves to identify glycogen ²⁶. Glycogen was visualized as small densely stained spheres of up to 40 nm diameter and was prominently located in Schwann cell cytoplasm (Fig 2A and B). Based on the EM images our strong impression was that the majority of glycogen granules were located in Schwann Cells. We tested this impression by having two observers, naive to our experiments and our hypothesis, score EM images of Schwann cell or axon cytoplasm for the presence or absence of glycogen granules (as defined above). In 60 EM images of Schwann cell cytoplasm, the two observers identified glycogen in 32 and 34 images (53% and 57%), whereas in 28 EM images of axon cytoplasm glycogen granules were identified in 1 and 4 images (4% and 14%).

Figure 2.

Glycogen granules are present in Schwann cell cytoplasm of A fibers. A & C. Low power image reveals glycogen granules present in the cytoplasm. B. Higher power image of the cytoplasm reveals multiple glycogen granules visualized as dark spherical granules located throughout the cytoplasm (g). Ribosomes attached to endoplasmic reticulum (ER) are also present. D. Nerves incubated in aglycemic conditions showed a decreased density of glycogen granules. A & D Scale bar 500 nm. C same magnification as D. B Scale bar 100 nm.

In nerves exposed to 0 [glucose] aCSF for 2 hours, conditions known to deplete glycogen ⁵, the density of Schwann cell glycogen granules decreased (9.66 ± 1.56 granules per µm², Fig 2D) compared to control nerves (21.72 ± 5.08 granules per µm²; p < 0.034, Fig 2C). In a similar manner, the relative density of glycogen granules was determined in myelinating (i.e., associated with A fibers) and non-myelinating (i.e., associated with C fibers) Schwann cells. Glycogen granules were mainly seen in myelinating Schwann cells compared to non-myelinating Schwann cells (20.95 ± 3.65 granules per µm², vs 0.89 ± 0.29 granules per µm², respectively; p < 0.00001).

Expression of glycogen phosphorylase

At present, robust antibodies to immunocytochemically detect glycogen within fixed tissues are unavailable, and therefore, as a surrogate, immunohistochemical labelings were performed to determine the presence of glycogen phosphorylase, a key enzyme in glycogen metabolism. In the CNS, this enzyme appears to be selectively localized to astrocytes ²⁷, the only cell in the adult brain that contains glycogen.

Triple labelling studies were carried out using S-100 or neurofilament, to localize the glycogen phosphorylase to either Schwann cells or axons, respectively. Glycogen phosphorylase immunolabeling was detected in both axons and Schwann cells, as evidenced by the co-localization of the glycogen phosphorylase antibodies with markers for axons (neurofilament) or Schwann cells (S-100) (Fig 3A – D).

Figure 3.

Expression of glycogen phosphorylase in sciatic nerve. A - D. Immunohistochemical studies revealed the presence of glycogen phosphorylase (green), whose cellular location was identified using specific cellular markers for Schwann cells (S-100; red) or axons (neurofilament; blue). D. Merged images demonstrate that glycogen phosphorylase is detected in both Schwann cell cytoplasm (yellow) and axons (cyan). Scale bar 10 µm.

The stimulus-evoked compound action potential (CAP) in sciatic nerve

Electron micrographs of transverse sections of sciatic nerve illustrated the heterogeneous morphology of the constituent axons (Supplementary Fig 1A & B). The axons fell into two main classes; large diameter, myelinated fibers called A fibers (Supplementary Fig 1C) and groups of small diameter, unmyelinated fibers called C fibers. The C fibers were surrounded by Schwann cell cytoplasm (Supplementary Fig 1B & D). The compound action potential (CAP) evoked by a supramaximal stimulus is illustrated in Figures 4A & B. The small stimulus artifact (marked with an asterisk) is followed by a large unipolar peak within 1 ms of the stimulus (Fig 4A) followed by a much smaller bipolar response about 5 to 15 ms after the stimulus artifact (Fig 4B). The large unipolar peak is due to the A fibers and is named the A peak. The smaller bipolar peak is due to the slowly conducting C fibers, and is named the C peak ²⁸. As expected, the A and C fibers had different thresholds for activation (Fig 4C & D). The A fibers (i.e., the A peak) were activated at lower currents than the C fibers (i.e., the C peak; C peak amplitude was measured as the difference between the positive and negative peaks). The A peak reached its maximal amplitude at a current stimulus of about 1 mA, a stimulus intensity that only evoked about a quarter of the maximal C peak response.

Figure 4.

The compound action potential (CAP) recorded from CD-1 mouse sciatic nerve. A. Stimulus evoked CAP illustrates the large rapid A fiber response immediately after the stimulus artifact (*). Scale bars 0.5 ms and 2 mV. B. The A peak is followed by the slower and smaller C fiber response approximately 10 ms after the stimulus artifact (inset). Scale bars 1 mV and 1 ms: inset scale bars 0.05 mV and 1 ms. C. Increasing the stimulus current results in increased amplitude of the A (1) and C (2) fiber response, respectively. Stimulus artifact in 1 removed for clarity. Scale bars 5 mV and 0.25 ms for 1 and 0.25 mV and 1 ms for 2. D. Plotting the stimulus intensity versus the normalized amplitude of the A (□) and C (△) fiber CAPs, demonstrates the lower threshold for activation of the A fibers relative to the C fibers response. E. Aglycemia and glycogen content of sciatic nerve. Aglycemia, indicated by horizontal bar, resulted in onset of delayed failure of the A peak after about 2 hours, and total failure after about 4.5 hours (□), whereas the C peak response failed after about 0.5 hours and disappeared by 3 hours (◇). Glycogen content (open columns; right axis) decreased during aglycemia reaching a nadir after about 2 hours, coincident with the onset of A peak failure.

Effects of aglycemia on glycogen content and nerve excitability

The effect of removing glucose from the perfusing aCSF (i.e., aglycemia) on the A and C fiber CAP responses was investigated. After instituting aglycemia, the A peak remained stable for more than two hours on average indicating that all the A fibers were fully functional. At 2.10 ± 0.34 hours after the onset of aglycemia, the A peak began to fail. It was not lost entirely until 4.5 hours after the onset of aglycemia (n = 4). The C peak, on the other hand, began to fall after only 0.42 ± 0.11 hours, and disappeared completely in about 2.5 hours on average (n = 4). The glycogen content at the onset of aglycemia (preceding bath [glucose] = 10 mM) was 10.98 ± 1.00 pmoles µg protein⁻¹ (n = 3) and fell progressively to a much-reduced new steady state level over about 2 hours (Fig 4E). The time course of glycogen loss was tightly correlated with the onset of A fiber failure measured as a drop in the A fiber CAP to 95% of its control value (see Methods): the A fibers remained fully excitable until glycogen was depleted to near its nadir. Note that some glycogen remains even in the prolonged absence of glucose, a characteristic finding ²⁹. The loss of excitability of C fibers, however, was not well correlated with glycogen loss; these fibers failed well in advance of maximal glycogen loss suggesting that their excitability was not supported, or was only partially supported, by glycogen breakdown.

Effects of glycogen content on the CAP

Persistence of the A fiber CAP during the first two hours of aglycemia appeared to depend upon the presence of glycogen. If true, a corollary would be that latency to the onset of CAP failure should vary as a direct function of glycogen content at the onset of aglycemia; the more glycogen present at the onset of aglycemia, the longer the latency to CAP failure. To precisely define latency we used a curve fitting technique that identified the point when CAP fell by 5% ⁷. This idea was tested by altering glycogen content by pre-treatment with different glucose concentrations (see above; Fig 1) and then measuring the time to onset of CAP failure during aglycemia (Fig 5). As shown in Fig 5A, pre-treatment with high glucose (i.e., 30 mM) significantly prolonged the onset of A fiber CAP failure after initiating aglycemia (2.92 ± 0.42 hours, n = 5), compared to the latency of CAP failure under standard conditions where preceding [glucose] was 10 mM (2.10 ± 0.34 hours, n = 4; p < 0.05). Conversely, pre-treatment with low glucose (i.e., 2 mM) prior to aglycemia, dramatically shortened the onset to A fiber CAP failure (0.28 ± 0.30 hours (n = 4) vs 2.10 ± 0.34 hours (n = 4); p < 0.05; Fig 5A). The powerful relationship between latency of A fiber CAP failure and initial glycogen content (based on average glycogen content associated with the bath [glucose] before onset of aglycemia) is shown in Figure 5C.

Figure 5.

Glycogen content of sciatic nerves and latency to CAP failure during aglycemia. A. A fiber peak in nerves pre-incubated in 2 mM, 10 mM or 30 mM glucose for 2 hours prior to onset of aglycemia at 0 hrs. The peak amplitude was normalized to the value at aglycemia onset. The response was maintained longest in nerves pre-incubated in 30 mM glucose (○), followed by pre-incubation in 10 mM (□) then 2 mM (△) glucose. Horizontal bar indicates aglycemia – also applies to B. B. The latency to C fiber peak failure during aglycemia showed the same pattern of failure as the A fiber response, although the latencies were shorter. Symbols for A also apply. C. Relationship between latency to 95% CAP failure and glycogen content at the onset of aglycemia of the A (□) peak, demonstrating a steep linear relationship. The relationship between latency to 95% CAP failure and glycogen content at the onset of aglycemia of the C (○) peak is not as steep as that of the A peak. D. A comparison of the latency to failure of C fibers pre-incubated in 10 mM glucose (○) and A fibers pre-incubated in 2 (△) or 10 mM (□) glucose prior to aglycemia highlights the beneficial effect of glycogen on A fiber conduction. Horizontal bar indicates aglycemia.

A similar set of experiments was done evaluating the effect of initial glycogen content on latency to C fiber CAP failure during aglycemia (Fig 5B). There was no detectable effect (2 mM glucose vs 30 mM glucose incubation; p = 0.122; Fig 5B and 5C). Therefore, the effect of glycogen was limited to large myelinated axons (i.e, A fibers). In fact, in the absence of significant glycogen (i.e., after pre-incubation with 2 mM glucose), the failure during aglycemia of the A fiber CAP had the same time course as the failure of the C fiber CAP in the presence of glycogen (Fig 5D), highlighting the exclusive benefit of glycogen for myelinated compared to unmyelinated axons.

Inhibiting glycogen breakdown during aglycemia hastens A fiber, but not C fiber, CAP failure

We used the glycogen phosphorylase inhibitor DAB to prevent glycogen breakdown and further test the hypothesis that glycogen supported A fibers during aglycemia ³⁰. Nerves were pre-incubated for 2 hours in aCSF containing 10 mM glucose plus 1 mM DAB. The DAB-treated nerves were then subjected to aglycemia. The latency to A fiber CAP failure was 0.64 ± 0.12 hours, dramatically shorter than the latency to CAP failure in control nerves during aglycemia (2.10 ± 0.34 hours; p = 0.0009, n = 4; Fig 6A and 6D). In contrast to the obvious effect that DAB had on A fiber CAP latency to failure during aglycemia, DAB had no significant effect on C fiber CAP latency to failure (0.50 ± 0.04 hours (n = 4) in DAB vs. 0.58 ± 0.06 hours (n = 4) in control conditions; p = 0.13, Fig 6B and 6D). Similarly, in nerves pre-incubated in 30 mM glucose, without or with DAB, the C fiber CAP latency to failure was 0.75 ± 0.10 hours and 0.71 ± 0.09 hours, respectively, a non-significant decrease (p = 0.32, n = 3, Fig 6C and 6D). It is noteworthy that in the presence of DAB, the latency to onset of A fiber CAP failure becomes similar to the onset of C fiber failure (Fig 6D; cf Fig 5D).

Figure 6.

Effect of inhibiting glycogen phosphorylase on latency to CAP failure and glycogen content. A. Introduction of the glycogen phosphorylase inhibitor DAB resulted in accelerated failure of the A fiber peak (○) relative to control, untreated nerves (□). Horizontal bar indicates aglycemia – applies to B and C. B and C. DAB had no effect on latency to CAP failure of C fibers incubated in 10 mM (B) or 30 mM (C) glucose. D. DAB significantly accelerated CAP failure after onset of aglycemia in A, but not C fibers incubated in 10 mM glucose or 30 mM glucose. E. DAB increased glycogen content in sciatic nerves incubated in 10 mM glucose compared to nerves incubated in 10 mM glucose without DAB (left-hand columns). After 2 hours of aglycemia (filled columns) the glycogen content was elevated in nerves incubated in DAB compared to nerves incubated in the absence of DAB: ns = not significant; * p < 0.05, ** p < 0.005, *** p < 0.001, **** p < 0.0005.

We measured the effects of DAB on glycogen content to validate that glycogen breakdown was blocked. Sciatic nerves were incubated for 2 hours in aCSF containing 10 mM glucose in the presence or absence of DAB, and then exposed to aglycemia for 2 hours. In the absence of DAB, glycogen content fell significantly from 10.98 ± 1.00 to 2.99 ± 0.56 pmoles mg protein⁻¹, p = 0.0014 (n = 4, Fig 6E). In the presence of DAB, glycogen content did not fall significantly (14.17 ± 2.37 vs 12.25 ± 1.50 pmoles mg protein⁻¹, p = 0.30, n = 4, Fig 6E). As expected, the glycogen content after aglycemia was significantly higher in nerves treated with DAB (12.25 ± 1.50 pmoles mg protein⁻¹) compared to control nerves (2.99 ± 0.56 pmoles mg protein⁻¹, n = 4, p = 0.004). Note also that the baseline glycogen content of DAB-treated sciatic nerves (14.17 ± 2.37 pmoles mg protein⁻¹) was significantly higher than in control nerves (10.98 ± 1.00 pmoles mg protein⁻¹, n = 4, p = 0.046), consistent with the fact that glycogen was synthesized but not broken down.

Lactate and sciatic nerve energy metabolism

Extracellular lactate is an efficient fuel for CNS axons ^{5, 7} and neurons ³¹. Moreover, extracellular lactate is produced when astrocyte glycogen is broken down ^32–35 and can be taken up by CNS axons during aglycemia to support oxidative energy metabolism ^{5, 7}. We determined if peripheral nerve axon function could be sustained by lactate in the absence of glucose and if the sciatic nerve had detectable levels of extracellular lactate.

Lactate was equivalent to glucose in supporting the function of sciatic nerve A fibers. The A fiber CAP was stable for hours in the continuous presence of 10 mM glucose (n = 4) or 20 mM lactate (n = 4, Fig. 7A). This concentration of lactate (i.e., 20 mM) was chosen because it is equivalent to the amount that would result from breakdown of 10 mM glucose. Further proof that these metabolic substrates equally supported A fibers was obtained by testing their ability to ‘rescue’ the A fiber CAP following aglycemia. After 2.5 hours of aglycemia the CAP had fallen to about 50% of its baseline value. Glucose (n = 4) or lactate (n = 4) applied at this point led to complete recovery of the CAP (Fig 7B).

Figure 7.

Lactate is equivalent to glucose in supporting sciatic nerve function. A. In the presence of 10 mM glucose (□) or 20 mM lactate (◇) the A fiber CAP is stable for hours. B. In the absence of glucose (from time ‘0’), the A fiber CAP begins to fail after about 2 hours (◇). If glucose (□) or lactate (◇) is introduced after 2.5 hours of aglycemia, when the CAP had fallen to about 50% of baseline value, the CAP recovers fully. C. In the presence of 10 mM glucose (□) the C fiber CAP is stable for hours, but in the presence of 20 mM lactate (◇) the C fiber CAP amplitude gradually decreases. D. In the absence of glucose the C fiber CAP begins to fail after about 1 hour (◇). If glucose (□) or lactate (◇) is introduced after about 2 hours of aglycemia, when the CAP had fallen to about 50% of baseline value, the C fiber CAP (◇) only partially recovers.

Similar results were obtained for the C fiber CAP except that lactate was not quite as effective as glucose in sustaining the C fiber CAP. The C fiber peak was supported by 10 mM glucose (n = 4) for at least 4 hours and was rescued by 10 mM glucose after 2 hours of aglycemia (n = 4, Fig 7C). The CAP drifted slightly lower following rescue by glucose suggesting that C fibers may have sustained a small degree of injury from the preceding aglycemia. Lactate was effective as the sole energy supply for C fibers but after about 1.5 hours the CAP amplitude tended to decline slowly to about 80% of baseline level (n = 4, p = 0.43). The C fiber CAP was largely restored by 20 mM lactate after 2 hours of aglycemia (n = 4, Fig 7D) in a manner not significantly different from glucose rescue (compare Fig 7C and 7D).

We used a highly selective lactate biosensor, or ‘enzyme electrode’, to measure lactate concentration in sciatic nerves under control and aglycemic conditions. The tip of the lactate biosensor was pressed against the outside edge of the nerve (Fig 8A). In nerves perfused with 10 mM glucose there was a constant lactate signal of 118 ± 6 µM (n = 3), indicating that lactate was produced constantly in nerve perfused with glucose and ‘leaked’ into the extracellular space. If the sensor was moved away from the nerve, lactate quickly became undetectable. Removal of bath glucose resulted in a steep fall in the lactate concentration, taking 16.9 ± 4.1 mins to reach 50% of baseline and 33.2 + 4.9 mins to fall to near zero at the edge of the nerve (Fig 8B). For comparison, the A and C fiber CAPs are shown during aglycemia. It is important to keep in mind that the lactate sensor pressing against the nerve was detecting lactate that was diffusing from the complex extracellular space of the nerve. In this position, of course, the probe could not measure the actual extracellular [lactate] because the escaping lactate was immediately subject to massive dilution by bath solution. These results indicated, however, that during aglycemia nerve [lactate] went from a state of excess to being nearly undetectable (see Discussion).

Figure 8.

Lactate is present in the extracellular space. A. An enzyme electrode was used to directly measure extracellular lactate concentration, [lac⁻]_o. The active sensor (grey) of the enzyme electrode was pressed alongside the optic nerve, while simultaneously recording the stimulus evoked CAP via suction electrodes. B. The [lac⁻]_o was zero in the bath (not shown) but about 120 µM at the edge of the sciatic nerve. After about 10 min of aglycemia the [lac⁻]_o begins to fall steeply and reaches nearly zero after 20 min. The response of the A (□) and C (◇) fibers has included for temporal comparison. C. Introduction of DAB (horizontal bar) results in a reversible fall of about 35% in the lactate signal.

To determine if glycogen was a source of the lactate signal recorded from nerves perfused with 10 mM glucose, the glycogen phosphorylase inhibitor, DAB, was applied. Introducing aCSF containing 1 mM DAB resulted in a fall in the lactate signal from 132.1 ± 15.5 µM to 84.3 ± 10.3 µM (n = 3), a 36 % fall. This effect was reversible on washout of DAB (Fig 8C). These results indicated that glycogen-derived lactate contributed significantly to the lactate signal under resting conditions (i.e., in the presence of normal bath glucose).

Discussion

Our results indicated that glycogen is present in mammalian peripheral nerve and is called upon to selectively support myelinated axon excitability under conditions of aglycemia. Glycogen, therefore, serves a similar function in both CNS and PNS axonal pathways. Further parallels were evident: glycogen appears to be located primarily in glial cells in both tissues, astrocytes in the CNS and Schwann cells in the PNS, and is likely broken down to lactate for export to glucose-starved neural neighbors for energy metabolism. A unique feature of PNS glycogen, compared to CNS glycogen, was the long duration of functional support it offered to large myelinated axons in the absence of glucose: 2 hours vs 0.3 hour, a 6-fold longer period of functional support in peripheral nerve compared to CNS white matter. Schwann cell glycogen, therefore, is a major defence against hypoglycemic dysfunction of myelinated axons.

To the best of our knowledge this is the first report to quantitatively identify glycogen in mammalian peripheral nerve, localize the glycogen to myelinating Schwann cells under physiological conditions and show that this molecule can support axon function. Our conclusions are based on the following observations: (1) acutely isolated sciatic nerve contained substantial amounts of glycogen, (2) glycogen granules were primarily present in myelinating Schwann cells, (3) during aglycemia myelinated axon conduction depended on glycogen breakdown, (4) glycogen content could be modulated, with up-regulation prolonging latency to CAP failure during aglycemia, (5) exogenous lactate supported the CAP in the absence of glucose, and (6) lactate was present in nerve extracellular space and fell during protracted aglycemia, and when glycogen breakdown was prevented with the glycogen phosphorylase inhibitor DAB. Taken together, these findings support the hypothesis that Schwann cell glycogen is broken down during aglycemia to lactate that is exported to the extracellular space and taken up by myelinated axons to maintain ATP stores and excitability.

Failure of peripheral axon function (i.e., peripheral neuropathy) is one of the most common forms of neurological disease and can be caused by hypoglycemia (Mohseni, 2001; Ozaki et al., 2010). Our findings indicated that A fibers have a unique metabolic relationship to Schwann cells with regard to energy metabolism. Additionally, the fact that C fibers are much less able to benefit from PNS glycogen would make them especially vulnerable to injury in certain diseases, notably diabetes, where hypoglycemia is common (Mohseni, 2001). The ‘trophic’ functions that Schwann cells provide to axons ¹ can be looked at in a new light based on the fact that these cells are able to support energy metabolism in myelinated axons when glucose is in short supply.

Axon sub-populations behave differently during aglycemia

Professor R. Granit of the Nobel Institute, made this comment when Erlanger and Gasser received the 1944 Nobel Prize: “Erlanger and Gasser showed that the (peripheral) nerve fibers, according to their conduction velocities, could be divided into three main groups of which the first, group A, could be further subdivided. The thickest mammalian fibers, the A-fibers, conduct impulses as fast as from 5 to 100 metre per second, the thinnest, the C-fibers, have conduction velocities below 2 metre per second.” Erlanger and Gasser described the frog sciatic nerve CAP in terms of multiple components identified by peak latency (an index of conduction velocity) and profile ³⁶. These descriptions apply to the rodent sciatic nerve ³⁷, a mixture of myelinated and unmyelinated axons mediating both sensory and motor function. Our CAP recordings were dominated by the A fiber peak arising from large diameter, myelinated axons (Fig 4). Unlike the situation in the CNS, where each oligodendrocyte provides internodal myelin segments to many axons, in the PNS each myelinating Schwann cell makes only one internodal myelin segment. The tiny C peak followed the A peak after a long latency, and was bipolar, as expected for slowly conducting unmyelinated axons ³⁸. We could find no detailed information about fiber composition in mouse sciatic nerve but the rat sciatic nerve is composed of about 27,000 axons of which 6% are myelinated motor axons, 23% and 48% are myelinated and unmyelinated sensory axons, respectively, and 23% are unmyelinated sympathetic axons ³⁹.

Surprisingly, during aglycemia C and A fibers were differentially benefited by glycogen. During aglycemia, the C fiber CAP began deteriorating within about 40 minutes compared to about 125 minutes for the A fiber CAP. Changes in initial glycogen content had dramatic effects on the capacity of A fibers, but not of C fibers, to function during aglycemia (Fig 5C). Doubling baseline glycogen increased the latency to CAP failure during aglycemia by ~100% for A fibers but had no significant effect on C fibers. The C peak also fell before glycogen content reached its nadir, and blockade of glycogen breakdown by DAB had no effect on the time course of C fiber failure during aglycemia. The inability of C fibers to benefit from glycogen during glucose deprivation was initially puzzling. Clearly these fibers could use lactate as a fuel, presuming that glycogen is broken down to this monocarboxylate. It seems likely, however, that the ‘delivery’ of glycogen-derived fuel to C fibers is the problem, as glycogen is predominately contained in myelinating Schwann cells that ensheath A fibers. In addition, C fibers with smaller diameters and lacking compact myelin, have larger metabolic demands than the larger diameter, myelinated A fibers, for an equivalent unit volume ⁴⁰ and diameter ⁴. Thus, glycogen-derived substrate may simply be less able to meet the higher demand in C fibers. Differences in how A and C fibers behave with regard to glucose deprivation and ability to be supported by glycogen might help explain why C fibers are especially susceptible to damage in diseases like diabetes. These questions and related ones are unresolved for the moment.

The saturable relationship between ambient glucose and glycogen content mirrors the relationship in optic nerve ⁵. Although the glycogen content of the sciatic nerve bathed in 10 mM glucose for 2 hours was ~2 fold greater than in the optic nerve, the latency to A peak failure was almost 6 fold greater than that of the optic nerve CAP. This remarkable difference may be because sciatic nerve has an intrinsically lower metabolic rate than the optic nerve, or that the energy substrate stored in Schwann cell glycogen is more efficiently transferred to axons than is the case within the brain. Assessing this hypothesis would not be easy, however.

Peripheral nerve axons are ensheathed by a cellular structure called the perineurium that represents a barrier capable of interfering with the diffusion of molecules, such as glucose or lactate, between axons and bath solution. This structure, however, abundantly expresses glucose transporter 1 (GLUT1) and also monocarboxylate transporter 1 (MCT1) ⁴¹. The presence of GLUT1 in perineurium plasma membranes strongly suggests that glucose can move from the bath into the sciatic nerve without difficulty. Likewise, MCT1 expression in the perineurium would allow facilitated transport of lactate across this barrier.

Glycogen localization

There are no previous studies on the role of glycogen in peripheral nerve and very few determining even its presence. The few studies that do mention glycogen are in pathological situations and tend to show indiscriminate glycogen accumulation in both Schwann cells ^14–16 and in axons ^{42, 43} in human biopsy material. We localized electron-dense particles primarily to the cytoplasm of myelinating Schwann cells. In a few instances, glycogen granules were seen in axons but this was very rare. For this reason, however, we cannot exclude the possibility of a very limited store of glycogen within sciatic nerve axons.

The immunohistochemical studies showed glycogen phosphorylase was located in axons and Schwann cells. The presence of this enzyme in a cell suggests the presence of glycogen, but may not be an infallible marker. For example, neurons express glycogen synthase but do not make glycogen under normal conditions ^{44, 45}. Glycogen is only expressed in neurons under pathological conditions (e.g. Lafora’s disease) ¹⁹. Thus the presence of enzymes associated with glycogen metabolism does not necessarily imply the presence of glycogen. Conversely, however, the absence of the key enzyme glycogen phosphorylase in a particular cell would be strong evidence against its involvement in glycogen metabolism.

How does Schwann cell glycogen support axon function?

In the CNS, astrocyte glycogen breaks down to a monocarboxylate (almost certainly lactate) that is transferred to axons by monocarboxylate transporters or MCTs during aglycemia ^{32, 46, 47}. The most likely, but as yet incompletely tested, scenario is that a similar sequence of events takes place in the sciatic nerve: during aglycemia Schwann cell glycogen would breakdown to lactate, which would be transferred via the extracellular space to myelinated axons to support ATP production and excitability. The presence of lactate in the extracellular space of the sciatic nerve, its ability to substitute for glucose to maintain axon excitability and its fall during aglycemia, are compatible with this hypothesis. That glycogen only supports myelinated axons is consistent with the observation that glycogen is concentrated in myelinating Schwann cells. It also raises the intriguing possibility that the myelinating Schwann cell, intimately contacting the axon it myelinates, conveys glycogen-derived lactate selectively into the internodal space compartment. Pathways for such selective metabolic support have been suggested ⁴. If this were so, however, there would be no advantage for lactate to be transported at high concentration into communal extracellular space. Yet this is exactly what our data indicate. Determination of the spatial distribution of MCT expression in Schwann cells and axons, both myelinated and unmyelinated, may shed light on this question.

It is important to keep in mind that we measured lactate at the outer edge of the nerve, not within the nerve. Thus, we could only detect lactate that was in excess of what was used by the nerve. Moreover, the probe is at the bath-nerve interface where the released lactate would be instantly diluted by flowing bath solution. This undoubtedly limited our ability to detect very low levels of extracellular lactate. Under control conditions in the presence of glucose (and glycogen, of course), lactate was obviously produced in excess of demand and diffused from the nerve into the bath. In the absence of glucose, however, glycogen was the only fuel and the lactate it produced was efficiently extracted from the extracellular space. Under these conditions, there would be little, or no, excess lactate and therefore none could be detected diffusing from the nerve by our probe. In other words, in the absence of glucose, lactate production is diminished (i.e., it is derived solely from glycogen) and lactate usage is maximal because there is no other energy substrate. If our probe were small enough to be placed into the center of the nerve, the picture would be different. Then, we would expect to see a two-phase drop in lactate during aglycemia: an initial drop from the high baseline level to a much lower level when glucose was omitted and a final drop to zero lactate when glycogen was depleted.

The physiological role played by peripheral nerve glycogen, beyond supporting excitability during aglycemia, is not understood. Blood glucose fluctuates significantly and it would be a reasonable safeguard to have an independent energy supply that could take over during relative hypoglycemia. Glycogen might also ‘supplement’ glucose during periods of intense axonal discharge when fuel demand could outstrip supply, if only briefly. The fact that about a third of extracellular [lactate] is due to glycogen breakdown under control conditions, that is in the presence of glucose, suggests that glycogen participates in the maintenance of an easily accessible ‘reserve’ energy supply. Other physiological roles are possible and should be actively explored to better understand the importance of Schwann cell glycogen in the function and long-term integrity of peripheral nerve axons. Certainly, the discovery of glycogen in normal myelinating Schwann cells adds an unexpected and fascinating dimension to how it might partner with its axon neighbors.