Abnormal Glycogen Storage by Retinal Neurons in Diabetes

Abstract

Purpose: It is widely held that neurons of the central nervous system do not store glycogen and that accumulation of the polysaccharide may cause neurodegeneration. Since primary neural injury occurs in diabetic retinopathy, we examined neuronal glycogen status in the retina of streptozotocin-induced diabetic and control rats.

Methods: Glycogen was localized in eyes of streptozotocin-induced diabetic and control rats using light microscopic histochemistry and electron microscopy, and correlated with immunohistochemical staining for glycogen phosphorylase and phosphorylated glycogen synthase (pGS).

Results: Electron microscopy of 2-month-old diabetic rats (n = 6) showed massive accumulations of glycogen in the perinuclear cytoplasm of many amacrine neurons. In 4-month-old diabetic rats (n = 11), quantification of glycogen-engorged amacrine cells showed a mean of 26 cells/mm of central retina (SD ± 5), compared to 0.5 (SD ± 0.2) in controls (n = 8). Immunohistochemical staining for glycogen phosphorylase revealed strong expression in amacrine and ganglion cells of control retina, and increased staining in cell processes of the inner plexiform layer in diabetic retina. In control retina, the inactive pGS was consistently sequestered within the cell nuclei of all retinal neurons and the retinal pigment epithelium (RPE), but in diabetics nuclear pGS was reduced or lost in all classes of retinal cell except the ganglion cells and cone photoreceptors.

Conclusions: The present study identifies a large population of retinal neurons that normally utilize glycogen metabolism but show pathologic storage of the polysaccharide during uncontrolled diabetes.

Glycogen represents the only carbohydrate reserve in mammalian cells and is synthesized from UDP-glucose to yield a polymer of glucose units joined primarily by α-1,4-glycosidic linkages, through the action of glycogen synthase (GS), with branch points at α-1,6-glycosidic linkages, introduced by the glycogen branching enzyme.¹ The storage unit for glycogen is the glycogen particle, a highly organized macromolecular complex that may have up to 12 shell-like tiers with a diameter of 42 nm and accommodation for 55,000 glucosyl residues.^1,2 In most instances the particle does not achieve its maximal capacity, and in skeletal muscle the average particle diameter is 25 nm, corresponding to seven tiers.^3,4 The glycogen particle is seeded by a self-glucosylating protein primer called glycogenin,⁵ and the current model corresponds to the beta particle⁴ in the ultrastructural classification described by Wanson and Drochmans⁶ and Drochmans.⁷ When required this energy reserve is mobilized by glycogen phosphorylase (GyP) and the debranching enzyme, with liberation of glucose-1-phosphate and free glucose in what represents the rate-limiting step of glycogenolysis.^1,8 The principal enzymes of glycogenesis and glycogenolysis, GS and GyP, respectively, are subject to multiple levels of regulation, both allosterically by major metabolites, and reciprocally by phosphorylation. Phosphorylation inhibits GS but increases the activity of GyP, and both effects can be accomplished simultaneously through the action of protein phosphatase-1 (PP1). Importantly, GS and GyP remain closely associated with the glycogen particle through “protein targeting to glycogen,” the glycogen-targeting subunit of PP1.^2,9 

Although liver and muscle represent the principal repositories of glycogen in mammals, glycogen can be synthesized by many different types of cells, including heart, adipose tissue, kidney, and brain¹; however, there is a long-standing conviction that glycogen is not stored by neurons in the adult central nervous system (CNS).^10,11 It has been proposed that glycogen storage in the brain is the sole preserve of the astroglia, while neuronal glycogen synthesis is actively suppressed and may actually induce apoptosis.¹² A very recent study has challenged this paradigm and proposed that neurons in the brain do store glycogen, albeit at a low level, and also possess the necessary GyP for its utilization as an energy reserve during hypoxia.¹³ Although primary neurodegeneration and functional hypoxia represent features of early diabetic retinopathy,^14–19 glycogen metabolism in retinal neurons during diabetes has not yet been associated with either of these pathologies. Several studies have documented increased glycogen synthesis in the diabetic retina, although the cellular distribution of the polysaccharide has not been addressed.^20–22 It is therefore important to document the distribution of glycogen in retinal neurons and to consider the possible roles of this important glucose storage molecule in the context of diabetic retinopathy. 

In the present study, we examined the distribution of glycogen in neurons of control retina and in those of diabetic rats after 2- and 4-month durations of diabetes. We also sought to identify which, if any, retinal neurons express GyP, the initiator of glycogenolysis. Additionally, we wanted to assess the activation state of GS in diabetic retina by comparing the distribution of phosphorylated glycogen synthase (pGS) in retinal tissue sections from diabetic and control rats. Glycogen synthase is inactivated by phosphorylation, and in muscle and brain, also undergoes nuclear sequestration.^12,23 Although the known phosphorylation sites are not involved in its nuclear translocation,²⁴ we considered that the intranuclear enzyme was likely to be inactive and phosphorylated. As previous studies have shown that cytoplasmic GS is widely distributed in the retina, chiefly in the Müller glia,²⁰ we sought to examine the distribution of the inactive phosphorylated form of the enzyme, to assess its nuclear sequestration in retinal neurons, and how such compartmentalization may be affected by diabetes. 

Materials And Methods

Experimental Animals

Since a large number of suitable archival eyes from diabetic and age–sex matched control rats, obtained and processed according to the protocol described below, were available in our laboratory, we considered it both efficient and congruent with the 3R principles of ethical research practice²⁵ not to induce diabetes in any further animal groups. 

Male Sprague-Dawley rats weighing between 240 and 280 g were randomly assigned to diabetic or control groups, and diabetes was induced by a single intraperitoneal injection of streptozotocin (65 mg/kg body weight) in 0.1 M citrate buffer pH 4.6 (Sigma Poole, Dorset, UK). The diabetic state was characterized by cessation of growth, polyuria/polydipsia > 200 mL/d and persistent blood glucose levels > 15 mM/L. After 2- and 4-month durations of diabetes, diabetic animals and control rats were killed by an overdose of sodium pentobarbital. The mean blood sugar in the diabetic groups at the time of culling was 25.5 mM/L (SD ± 5.4). The experiments were carried out under the United Kingdom 1986 Animals Scientific Procedures Act with supervision by the British Home Office, and conformed to the ARVO statement for the use of animals in ophthalmic and visual research. 

The singular disadvantage incurred in the use of archival eyes for the current investigation was that only batch blood sugar data were available for the diabetic animals from which the study eyes were obtained, so that cell glycogen content could not be correlated precisely with individual blood sugar level. Retinal tissue blocks from one eye of 2-month-old diabetic and control animals (n = 6/group) were available for electron microscopic analysis and at least one eye from 4-month-old diabetic (n = 11) and control rats (n = 8) were available for wax histology. 

Tissue Processing and Staining

Eyes from diabetic animals and sex–age matched controls for electron microscopy had been fixed in glutaraldehyde and processed according to standard protocols, while those for use in wax histology had been fixed in 10% formalin (see Supplementary Methods). 

To further assess the presence of glycogen in retinal neurons of nondiabetic rats, eyes of additional 4-month-old control male Sprague Dawley rats (n = 6) were preserved in Davidson's fixative (DF), which provided superior retinal morphology, glycogen preservation, and immunohistochemical staining with several antigens, particularly GyP. Histochemical staining for glycogen was performed with the periodic acid Schiff (PAS) reaction with negative controls digested with alpha amylase. Following heat-induced antigen retrieval, immunohistochemical staining was performed on wax sections with rabbit polyclonal anti-Glycogen Phosphorylase H-300 (sc-66913; Santa Cruz Biotechnology, Dallas, TX, USA) and anti–phospho-Glycogen-Synthase (07-817; Merck Millipore, Temecula, CA, USA). Secondary detection used either the DAKO Envision system for brightfield microscopy or Alexa-568–labeled goat anti-rabbit (Life Technologies, Paisley, Scotland, UK) for fluorescence and confocal microscopy (see Supplementary Methods). Immunohistochemical controls were performed by substitution of the primary antibodies with an irrelevant rabbit primary antibody or nonimmune serum (Supplementary Fig. S1). Combined PAS histochemical and immunohistochemical staining was used to co-localize glycogen deposits and pGS for brightfield microscopy. As a glycogen antibody from a different host species to that used for pGS staining was unavailable, an alternative method was devised in which alkoxyamine-biotin coupling with FITC-streptavidin was used in place of Schiff's reagent for fluorescent staining of periodate-oxidized glycogen and co-localization with pGS in confocal microscopy (see Supplementary Methods and Supplementary Fig. S5D). 

Morphometric Data Collection

As the glycogen-filled amacrine cells in diabetic rats were confined to a single cell layer at the inner aspect of the inner nuclear layer (INL), the number of PAS-positive cell bodies was counted in a linear fashion in a continuous strip of central and equatorial retina from each eye. Sections were selected from central axial planes of each eye, including or immediately proximal to the optic disc, and eight contiguous high-magnification images were collected from each side of the disc, providing a sample of 4.8 mm of retina/eye. Randomized sets of images from 11 diabetic and 8 control eyes were analyzed in a single-blind fashion by two independent investigators. The changes in pGS staining in diabetic eyes were also quantified in six random high magnification images from the central retina in immuno-peroxidase stained preparations. A mean of 960 INL cells were counted from each retina, and staining was classified as maximal or “reduced” in relation to the maximally stained cells in each section (term “reduced” includes cells in which nuclear pGS staining was not detected). Suitably stained sections for nuclear pGS analysis were available from 10 diabetic and 8 control rats. 

Statistical Analysis

The numbers of PAS-stained amacrine cells/mm of central-equatorial retina of diabetic and control rats were compared using an unpaired t-test (2-tailed) with a Welch correction for unequal variance. The same analysis was used to compare the percentage of nuclei in the INL showing reduced or absent pGS immunostaining in diabetic versus control retina. A P value < 0.05 was considered statistically significant. Statistical analysis was performed using GraphPad Prism (GraphPad, San Diego, CA, USA). 

Results

Neuronal Glycogen in Diabetic and Control Retina

After 2 months of diabetes, electron microscopy revealed marked deposition of glycogen (Figs. 1A–E) in cells identified as amacrine cells according to their ultrastructure and location at the internal border (Figs. 1A, 1D, 1E) of the INL. Notably these cells had smooth nuclear profiles (Figs. 1A, 1B, 1E), and glycogen was not observed in those amacrine cells that have a characteristically crenelated nuclear envelope and show numerous dense bodies in their soma (Supplementary Fig. S2). The glycogen deposits ranged from small aggregates distributed throughout the perinuclear cytoplasm (Fig. 1C), to gross confluent accumulations that often showed artefactual vacuolation and fracturing (Figs. 1A, 1B), presumably because there was insufficient cytoplasmic protein to bind the glycogen particles during tissue processing. Glycogen was present on all sides of the cell nucleus, but tended to accumulate on the aspect facing the INL (Figs. 1B, 1E). The glycogen deposits occurred as discrete particles (∼30 nm diameter) of beta glycogen (Fig. 1C inset). Occasionally, glycogen aggregates were observed in small cell processes (Fig. 1D) within the inner plexiform layer (IPL), but not in the major vertical process that characterized the affected amacrine cells (orientation indicated by arrows in Fig. 1E). Such vertical processes were also observed in amacrine cells expressing GyP in immunohistochemical preparations (Fig. 3D) and by confocal microscopy (Supplementary Fig. S5A). 

Figure 1

 

Electron micrographs of 2-month-old diabetic (A–C) and control (F) rat retina. (A) Dense accumulations of glycogen (arrows) in the perinuclear cytoplasm of an amacrine cell (Ac) at the interface of the INL and IPL; scale bar: 2 μm. (B) Adjacent amacrine cell shows a glycogen-filled (arrows) saccular distension of the cell soma at the pole of the cell facing the INL; scale bar: 1 μm. The large confluent glycogen masses show artifactual fracturing and vacuolation (A, B). (C) Less confluent glycogen deposition (G) demonstrates the ultrastructural features of discrete beta-particles (inset); scale bar (main): 200 nm, scale bar (inset): 100 nm. (D) Dense mass of glycogen fills a presumptive amacrine process (arrow) in IPL proximal to an Ac body with obvious masses of glycogen adjacent to the Golgi apparatus (Go); scale bar: 1 μm. (E) Amacrine cell with perinuclear glycogen masses (G) shows a major vertically orientated cell process within IPL (arrows); scale bar: 1 μm. (F) Small aggregates of glycogen particles (thick arrows and inset) are present in an amacrine cell of control retina. The cell containing the glycogen has a smooth nuclear profile, similar to those showing glycogen storage in the diabetic retinas; scale bar: 1 μm.

In control rats, electron microscopy showed discrete glycogen granules throughout the retinal neuropile, in both neurons and glia; however, the only conspicuous aggregates of the polysaccharide were observed in a subclass of the amacrine cells (Fig. 1F). Such aggregates occurred in cells with the same location and nuclear profile as those showing glycogen storage in diabetic rats; albeit at a much lower level than in the diabetics. As in the diabetic amacrine cells, the glycogen aggregates were typically irregular and located in the perinuclear cytoplasm where they appeared to lie free within the cytosol and not enclosed by any endomembrane structures (Figs. 1C, 1F insets). 

After 4 months of diabetes, light microscopy revealed that glycogen-engorged amacrine cells were distributed throughout the whole retina, but predominantly in the central and equatorial regions (Figs. 2A, 2B). The most significant glycogen accumulations appeared as PAS-positive, amylase-labile deposits clustered around the cell nucleus (Fig. 2C), particularly at the aspect of the nucleus within the INL and remote from the IPL; this was most obvious when the glycogen had been digested to reveal clear cytoplasmic distensions as in Figure 2D. Highly localized dense deposits of glycogen were consistently found throughout the inner segments of the photoreceptor cells of both diabetic and control retina (Figs. 2C, 2D). In addition to amacrine and photoreceptor cells, dense accumulations of glycogen could occasionally be found in cell bodies within the ganglion cell layer (GCL; Figs. 2E, 2F), although this was uncommon and only noted in diabetic specimens that also showed amacrine cells heavily congested with glycogen. 

Figure 2

 

Periodic acid Schiff–stained sections from wax-embedded eyes of 4-month-old diabetic rats, including consecutive sections predigested in alpha-amylase to differentiate glycogen deposits from vascular basement membranes and other PAS-positive retinal proteins (B, D, F); cell nuclei were stained with methyl green (A–D) or hematoxylin (E, F). Images (C–F) of eye injected with fixative at enucleation shows superior glycogen localization but with artifactual retinal detachment. (A) Glycogen-filled amacrine cell bodies (arrows) in diabetic retina are unstained in the amylase-digested control (B). (C) In addition to glycogen-filled amacrine cells, dense spherical deposits of glycogen are present in the photoreceptor inner segments (arrows and inset). In amylase-digested control (D) the stores of glycogen in the amacrine cells appear as empty vacuoles (arrows), while deposits in the photoreceptors have been removed. (E) Periodic acid Schiff–positive cell body in a cell body within the GCL (arrow) that is unstained in the amylase-digested control ([F], large arrow). Scale bars: 50 μm.

Figure 3

 

Immunohistochemical stains for GyP from eyes of formalin-fixed 4-month-old diabetic and control rats (A–C) and additional Davidson-fixed control eyes from 4-month-old controls (D–F). (A) GyP immunoreactivity in control retina is present in amacrine and ganglion cell bodies, and cell processes in the IPL, especially in a presumptive synaptic layer within the IPL. (B) GyP staining in diabetic retina shows the same pattern as in controls but with increased intensity in the cell processes of the IPL. GyP-positive amacrine cells in diabetic eyes show staining within saccular distensions corresponding to glycogen deposits at the aspect of the cells facing the INL (arrows and inset). (C) Higher magnification of GyP-positive amacrine cells in diabetic retina showing the saccular distensions described in B (arrows). (D) Contiguous segments from central retina of Davidson-fixed control eye shows enhanced GyP immunostaining of amacrine cells, ganglion cell bodies (Gc), and axons in the NFL. The GyP-positive synaptic layer within the IPL here appears trilaminar (white arrows) and several amacrine cells show the vertical processes described in Figure 1E; retinal vessels (V). Higher magnification shows GyP immunoreactivity in individual ganglion cell axons within the nerve fiber bundles (arrows). (F) Higher magnification of ganglion cells shows GyP immunostaining in their cell processes within the IPL (arrows). Scale bars (A, B, D): 100 μm; (C, E, F): 50 μm.

In spite of the fact that glycogen deposits were relatively easy to find in amacrine cells of control retina by electron microscopy, they were insufficient to be detectable in light microscopy of formalin-fixed retina stained by PAS. However, the rapid fixation achieved with DF in the additional control eyes did permit detection of glycogen in amacrine cells at the light microscopic level (Supplementary Fig. S3), although the PAS staining was weak in comparison to the experimental diabetic retina. 

Quantitative analysis of the glycogen-filled amacrine cell bodies in diabetic retinas (n = 11) produced a mean of 26 cells/mm central retina (SD ± 5) with a range of 18 to 33, compared to 0.5/mm (SD ± 0.2) in controls (n = 8). Only a small number of PAS-positive profiles were present in control retina and probably represented ambiguous oblique cuts through capillary basement membranes, as a layer of capillaries is distributed in the same plane of the retina as the amacrine cells. The difference between glycogen-filled amacrine cell bodies in diabetic and control retina was highly statistically significant (P < 0.0001; Supplementary Fig. S4A). 

GyP in Retinal Neurons

In control retina, immunohistochemistry for GyP revealed strong staining in the GCL, a scattered population of amacrine cells at the inner aspect of the INL and in a sharply defined lamina within the IPL, assumed to be a synaptic layer (Figs. 3A–F). Weaker staining was present throughout the IPL and the nerve fiber layer (NFL). In diabetic retina, the general pattern of GyP staining was similar to controls; however, more amacrine cell bodies showed strong staining for GyP and the overall intensity of staining in the IPL was greater than in control retina (Figs. 3B, 3C). Notably the GyP-positive amacrine cells in the diabetic retina showed strong staining within cytoplasmic distensions at the side of the nucleus facing the INL (Figs. 3B, 3C); these structures demonstrated identical morphology to the perinuclear glycogen stores observed in PAS-stained preparations. The additional control eyes preserved with DF showed superior morphology than the formalin-fixed diabetic and control eyes, as well as improved immunostaining for GyP (Figs. 3D–F). Following fixation in DF, a much larger population of amacrine cells were seen to express GyP in the control retina (Fig. 3D), and GyP immunoreactivity in ganglion cell bodies and dendrites, and axons in the NFL was more evident (Figs. 3D–F, Supplementary Fig. S5D). 

pGS in Retinal Neurons

Immunostaining for pGS in control retina revealed weak cytoplasmic staining throughout the retinal neuropile but intense nuclear localization that obscured the delicate methyl green chromatin stain in all retinal cells, with the exception of the vascular cells and rod photoreceptors (Fig. 4A). A ring of immunostaining surrounded the central heterochromatin in the rod cells (Fig. 4A), and confocal microscopy confirmed that the pGS was confined to the thin layer of euchromatin that surrounds the dense core of heterochromatin in rat rod cells (Fig. 5E). 

Figure 4

 

Immunohistochemical stains for pGS from formalin-fixed 4-month-old diabetic (B–F) and control Sprague-Dawley rats ([A], top), and additional Davidson-fixed control eyes from 4-month-old Sprague-Dawley rats ([A], bottom). ([A], top) Nuclear staining for pGS is present in all classes of retinal cells except those of the retinal vasculature. Only rings of immunoreactivity are present in the rod cells of the ONL as staining is excluded from the central mass of heterochromatin in these cells; the small number of more homogenously stained nuclei in the outer region of the ONL is likely to reside in cone cells. ([A], bottom) Davidson-fixed retina shows attached RPE-choroid and identical but more intense staining for pGS as in formalin-fixed controls. Note intense pGS staining in RPE cell nuclei (arrows). (B) Sections of retina from two different 4-month-old diabetic eyes; both show loss nuclear staining for pGS throughout the INL, ONL, and RPE (arrows, bottom panel). Nuclear pGS is retained by the majority of the nuclei in the GCL. Strong cytoplasmic pGS staining is only seen in the saccular distensions of glycogen-filled amacrine cell bodies (arrows and inset) and the Müller cell endfeet at the ILM (continuous layer at top edge of section) in the retina shown in the bottom panel. (C) Higher magnification of 4-month-old diabetic retina shown in the bottom panel of (B) above, showing cytoplasmic but lack of nuclear pGS staining in glycogen-filled amacrine cells (arrows). Nuclear pGS is also absent from the RPE (thick arrow). (D) Higher magnification of 4-month-old diabetic retina shown in the bottom panel of (B) above showing pGS staining in vertical processes of Müller cells spanning the ONL. (E, F) Four-month-old diabetic retina double stained with PAS for glycogen and immunoperoxidase for pGS. Several amacrine cells show glycogen-filled saccular distensions stained positively with PAS (arrows) but no pGS immunostaining in their cell nuclei. Scale bars (A, B): 100 μm; (C–F): 50 μm.

Figure 5

 

Confocal microscopy of formalin-fixed control and 4-month-old diabetic rats stained for pGS. Images (A) and (C) include the DAPI stain for nuclear chromatin, while (B) and (D) show the respective red channel only images of pGS localization. (A, B) Confocal microscopy confirmed that pGS was limited to the nucleus in all the neural cells of the retina and was located within the nucleoplasm in euchromatic regions of the nucleus. (C, D) In 4-month-old diabetic retina, loss of pGS staining was evident in the rod but not the cone cell nuclei (arrows) of the ONL; pGS was also lost from many, but not all, nuclei at all levels within the INL. Loss of pGS was not evident in the nuclei of the ganglion cells. (E) Four-month-old control retina shows pGS staining in the nuclei of the RPE and rod cell nuclei within the ONL. The tight apposition of the pGS to the central heterochromatin of the rod cells suggests that the enzyme lies within the peripheral euchromatin and not the cell cytoplasm. (F) Compared to control in image (E), 4-month-old diabetic retina shows depletion of pGS from the RPE and rod cell nuclei within the ONL. (G) Amacrine cells of 4-month-old diabetic retina double-stained with alkoxyamine-biotin-streptavidin-FITC (green) for glycogen and pGS (red) show loss of nuclear pGS and yellow-green glycogen deposits due to colocalization with cytoplasmic pGS. (H) High magnification image of similar area to that depicted in (H) shows discrete deposits of pGS in the nucleus of a glycogen-filled amacrine cell. Scale bars (A–D): 50 μm; (E–H): 20 μm.

In contrast to the almost universal nuclear staining for pGS in control retina, 4-month-old diabetic rats showed loss of nuclear pGS in all layers of the retina except the GCL. The INL showed a distinctly patchy distribution with loss of nuclear pGS at all levels (Figs. 4B–F), confirming that the change was not confined to one class of cells, as particular cell types are confined to distinct levels within the INL. Interestingly, some diabetic animals demonstrated strong cytoplasmic staining for pGS in the end-feet of the Müller cells (lower panel of Figs. 4B, 4C) at the internal limiting membrane (ILM), while others showed little pGS at the ILM but positive staining in vertical processes that spanned the outer nuclear layer (ONL; upper panel of Figs. 4B, 4D), corresponding to the outer processes of the Müller cells. Importantly, the only significant cytoplasmic staining for pGS in the INL was present in saccular distensions of amacrine cells, many of which lacked nuclear staining for pGS (Figs. 4B inset, 4C). These cells corresponded in both morphology and location to those showing the stores of glycogen observed by electron microscopy, PAS staining, and immunostaining for GyP. Combined staining with PAS and immunohistochemistry for pGS confirmed that the glycogen-filled amacrine cells in diabetic retina often showed loss of nuclear pGS (Figs. 4E, 4F); however, colocalization of glycogen and pGS in confocal microscopy revealed that although this was generally the case (Fig. 5G), loss of nuclear pGS was often incomplete (Fig. 5H). 

Nuclear sequestration of pGS was also lost or significantly reduced in the majority of the retinal pigment epithelium (RPE) cells in diabetic rats (Figs. 4B–F; Figs. 5E, 5F). This was also the case in the majority of the rod photoreceptors, although it was notable that a small number of cells scattered close to the outer margin of the ONL retained pGS in the diabetic retinas (Fig. 4C). Confocal microscopy confirmed that these were cone cells as the enzyme was clearly located within the central euchromatin that is typical of cones²⁶ (Figs. 5C, 5D, Supplementary Fig. S5C). 

Quantitative analysis of nuclear pGS in the INL of diabetic retina (n = 10) showed that a remarkably consistent percentage of the cells had reduced pGS nuclear staining, with a mean of 72% (SD ± 2.8) and a range of 68% to 76%, compared to a mean 0.52% (SD ± 1.27) and range of 0.32% to 0.68% in controls (n = 8). The difference between nuclear pGS in the INL of diabetic and control retinas was highly statistically significant (P < 0.0001; Supplementary Fig. S4B). 

Discussion

This study has shown that in contrast to the situation in the brain,^10,11 a significant population of the inner retinal neurons is equipped to utilize glycogen metabolism, particularly a large subclass of the amacrine cells and the majority of ganglion cells. Importantly, the amacrine cells that appear to utilize glycogen under physiological conditions are clearly unable to regulate its storage during uncontrolled diabetes, and their cytoplasm becomes congested with the polysaccharide to a degree that may be considered as pathologic. These observations have implications for neuronal dysfunction, and possibly also for their survival in diabetic retina. 

Glycogen Metabolism in CNS Neurons

The importance of glycogen metabolism in neuronal survival was highlighted by a 2007 study of the mechanism underlying the pathologic glycogen storage in Lafora disease, a genetic neurodegenerative condition characterized by myoclonus epilepsy²⁷: Evidence was presented that, under physiological conditions, glycogen synthesis in CNS neurons is actively suppressed and that glycogen storage in these cells may trigger apoptosis.¹² The authors concluded that although GS is expressed in neurons, its enzymatic activity is inhibited through phosphorylation, and that GyP, the key enzyme in glycogen catabolism, was undetectable at the protein level.¹² 

Glycogen Storage in Diabetic Retina

Early studies suggested disturbance of retinal glycogen metabolism during diabetes,^28–30 and a small electron microscopic study by Sosula et al.³¹ showed a disproportionate increase in retinal neurons. However, these observations have been largely overlooked in recent discussion of neuronal dysfunction in diabetic retinopathy, possibly because glycogen is not well retained in current fixation protocols and requires special staining. In the present study, we have shown that a significant proportion of amacrine cells exhibit excessive glycogen accumulation in the diabetic retina. These are the same cells that under physiological conditions store modest amounts of the polysaccharide and express GyP for its utilization. This offers good evidence that these cells normally employ glycogen as a local energy reserve, perhaps for utilization in times of increased neural activity or starvation. Importantly, such amacrine cells appear unable to regulate their glycogen content during diabetes and become congested with the polysaccharide. Previous studies have shown that although GS activity increases in several tissues in diabetic animals, including brain, only retina showed an increase in glycogen content.²² Glycogen metabolism in the retina may differ fundamentally to the rest of the CNS, although gross glycogen storage in the perikarya of hypothalamic neurons has been demonstrated in long-term diabetic rats, and was associated with cell-specific neurodegeneration.³² 

In the current study, the pattern of nuclear sequestration of pGS was seen to change significantly in the diabetic retina. The nuclear pGS was considered inactive as in addition to its phosphorylation state, the nuclear localization sequence of the enzyme includes the arginine-rich cluster of amino acids that are involved in its allosteric activation by glucose-6-phosphate (G6P).²⁴ Cells throughout the INL, ONL, and RPE of diabetic retina showed loss of nuclear localization of pGS. This change was thought to represent dephosphorylation and cytoplasmic translocation of the enzyme. Although it is not possible to completely exclude the possibility of an overall downregulation of GS, our view is supported by a recent study showing increased retinal GS activity, but with no quantitative alteration in GS expression during diabetes.²¹ Historically, much attention has been paid to the phosphorylation state of GS, and recently in relation to its role in the “silencing” of GS in neurons¹²; however, the bulk of available evidence suggests that phosphorylation plays only a “fine-tuning” role in GS regulation, and that allosteric activation by G6P represents the dominant control mechanism.³³ Indeed Roach et al.¹ emphasize that the presence of G6P may overcome phosphorylation-mediated inactivation of GS and restore full enzymatic activity. This may explain experiments in diabetic rats that showed increased GS activity and a 3-fold increase in retinal glycogen compared to nondiabetic controls, but without a measurable change in the level of pGS.²¹ Nuclear-cytoplasmic shuttling of GS has been extensively characterized in muscle^23,24 and has been demonstrated in neurons in vitro¹²; however, this study offers the first evidence of nuclear sequestration of pGS in CNS neurons in vivo and alteration in response to diabetes. Significantly, the amacrine cells that contained excessive glycogen stores in the diabetic retina often showed loss or reduction of nuclear pGS, but positive staining for the enzyme in their glycogen-filled cytoplasm. Such cytoplasmic staining is not unusual as both phosphorylated and nonphosphorylated forms of the enzyme may remain associated with the glycogen particle.¹ This suggests that in addition to phosphorylation, GS may also require nuclear sequestration to ensure inactivation. 

The observation that excessive glycogen storage was rarely encountered in retinal ganglion cells, and that their nuclei continued to sequester pGS during diabetes, suggests that the ganglion cells maintain some measure of control over GS activity, even during chronic hyperglycemia. 

The retinal pigment epithelium was not a focus in the present study as normal glycogen storage and utilization by RPE cells has been well characterized³⁴; however, it was of interest that there was a loss of nuclear pGS in the RPE of diabetic rats, which is suggestive of increased activity, and supports evidence of increased glycogen storage in the RPE of diabetic rats³⁵ and human patients.³⁶ 

The significance of the discrete glycogen aggregates observed in photoreceptor inner segments of both diabetic and control retina is unknown; however, a previous ultracytochemical investigation has shown that these deposits are located in cone cells.³⁷ Indeed the persistent nuclear sequestration of pGS in cone cells compared to rods in the diabetic retina suggests continued control of glycogen metabolism in cones during diabetes. Previous studies of primate and human retina indicate that in these species cones express GyP but rods do not, suggesting an active glycogen metabolism in cones compared to rods.³⁸ This correlates with a recent study showing that rod-derived cone viability factor promotes cone survival by stimulating glucose uptake and aerobic glycolysis,³⁹ a function for which stored glycogen would confer significant advantage. 

Glycogen Storage in Retinal Neurons: Pathologic Implications

The presence of glycogen in retinal neurons, at least in modest amounts has been previously observed by electron microscopy in the amacrine cells of the cat⁴⁰ and rat,³¹ suggesting that neuronal glycogen deposition per se is not pathologic, and supports recent studies from other parts of the CNS.¹³ However, it is likely that the level of glycogen occupancy described above in the perinuclear cytoplasm of the amacrine cells in diabetic rats would eventually disrupt the normal synthetic functions of the endoplasmic reticulum–Golgi complex and downstream transport mechanisms. 

Glycogen storage in the diabetic retina has not been previously associated with primary neuronal dysfunction or cell loss in diabetic retinopathy, and although the retina represents a target in classical glycogen storage diseases, notably Lafora disease^41,42 and Pompe disease,⁴³ the pathologic features of these conditions differ greatly from diabetes. The glycogen stores observed in diabetic amacrine cells in the present study occurred as aggregates of normal beta-particles lying free within the cytoplasm, while in Lafora disease, discrete membrane-bound bodies contain filamentous aggregates of abnormal polyglucosans. Again the glycogen aggregates in Pompe disease, although consisting of apparently normal beta-glycogen are sequestered within abnormal lysosomal bodies.⁴⁴ Nevertheless, amacrine cells show degenerative changes in models of diabetic retinopathy^15,45 and have been implicated in diabetic neurophysiological dysfunction as measured by alterations in the oscillatory potentials of the electroretinogram.^46,47 Whether glycogen accumulation contributes to functional disturbance or cell death in the affected cells will require mechanistic studies with agents that can modulate the glycogen load. Importantly, pathologic glycogen storage may also be relevant to neuronal dysfunction in diabetic patients as primate retina contains a subpopulation of amacrine cells that express GyP, confirming a need for glycogen metabolism in these cells.⁴⁸ 

In conclusion, it appears that the underlying mechanisms and sequelae of glycogen storage in CNS neurons are complex and that this is particularly true for the retina. In a neural tissue dependent on glucose metabolism, glycogen as the only readily-accessible short-term store is likely to play an important role. This has added significance when considering the metabolic profile of the retina and how it reacts to metabolic stress, and especially diabetes. Future studies will need to determine if neuronal glycogen storage is influenced by local or systemic insulin, correlates with stage of the disease, and whether regulation of blood glucose can reverse the change. At a more basic level, it will be important to address why ganglion and amacrine cells are dependent on an endogenous glycogen metabolism and why the amacrine cells are less able to regulate it during diabetes. 

Acknowledgments

Supported by Fight for Sight (FFS‐17358-1) and The Juvenile Diabetes Research Foundation (JDRF-5-CDA-2014-225-A-N). 

Disclosure: T.A. Gardiner, None; P. Canning, None; N. Tipping, None; D.B. Archer, None; A.W. Stitt, None