Advanced Glycation End Products in Extracellular Matrix Proteins Contribute to the Failure of Sensory Nerve Regeneration in Diabetes

All chemicals were obtained from Sigma-Aldrich (Poole, Dorset, U.K.) unless otherwise stated. High-purity methylglyoxal was prepared by acid hydrolysis of freshly distilled methylglyoxal dimethylacetal and purified by fractional distillation under reduced pressure (18).

All studies and procedures were licensed under the U.K. Animals (Scientific Procedures) Act 1986. Diabetes was induced in adult female Wistar rats (200–250 g, Charles River, U.K.) via intraperitoneal injection of freshly dissolved STZ (55 mg/kg in sterile saline) as previously described (19). Rats were killed at various time points after induction of diabetes (3, 6, 9, 12, and 24 weeks post-STZ, n = 6 at each time point), and the sciatic nerves were rapidly removed and frozen on liquid nitrogen. Age- and weight-matched rats were used as nondiabetic controls (time 0 and 24 weeks, n = 4 at each time point).

A separate group of adult male Wistar rats (250–300 g, n = 12) were anesthetized with isofluorane (2% in oxygen) under sterile conditions, and the left sciatic nerve was exposed at mid-thigh level. The sciatic nerve was crushed with the tips of watchmakers forceps (2 × 15 s), the wound was closed, and animals recovered under observation. Seven days after nerve crush, rats were killed by concussion followed by decapitation. Ipsilateral and contralateral L4 and L5 dorsal root ganglia (DRG) were removed for cell culture experiments.

Sciatic nerves were desheathed, homogenized in 0.25 mol/l sucrose on ice, and centrifuged at 900 g at 4°C. The ECM protein in the pellet was washed with PBS, then extracted with chloroform:methanol (2:1, v:v) and stirred continuously for 24 h at 4°C and centrifuged at 900 g at 4°C. The pellet was washed twice with PBS and frozen at −80°C until analysis.

Laminin and fibronectin modified by fructosyl-lysine and AGEs (AGE-laminin and AGE-fibronectin, respectively) were prepared by incubating the proteins (6.6 mg/ml) with 50 mmol/l glucose in 100 mmol/l sodium phosphate buffer, pH 7.4, at 37°C for 21 days. Laminin and fibronectin were also modified by incubating the proteins (6.6 mg/ml) with 500 μmol/l methylglyoxal in 100 mmol/l sodium phosphate buffer, pH 7.4, at 37°C for 24 h. After the incubation, glycated and control proteins were dialyzed against 30 mmol/l ammonium formate, pH 7.8 and 4°C; lyophilized to dryness; and stored at −20°C.

Protein glycation, oxidation, and nitration adducts were determined by stable isotopic dilution analysis liquid chromatography with tandem mass spectrophotometer detection (7). Glycation adducts determined were fructosyl-lysine and 10 AGEs (CML, CEL, G-H1, MG-H1, 3DG-H, GOLD, MOLD, DOLD, argpyrimidine, and pentosidine). The protein oxidation adduct dityrosine, the protein nitration adduct neurotrophin-3 (3-NT), and amino acids lys, arg, met, tyr, and trp were also determined. Pentosidine was determined by liquid chromatography with fluorimetric detection (20). Authentic standard analyses were prepared as described (7,21). Glycation, oxidation, and nitration adduct residues of protein extracts were determined in exhaustive enzymatic digests (50 μg protein equivalent) (22). Samples were analyzed using a module 2690 separation module with a Quattro Ultima triple-quadrupole mass spectrometric detector and 2475 fluorescence detector (Waters-Micromass, Manchester, U.K.) as previously described (20).

Control or preconditioned (7 days postunilateral sciatic nerve crush injury) adult rat sensory neurons were mechanically and chemically dissociated, purified, and resuspended in modified Bottenstein and Sato's medium in Ham's F12 (containing 10 mmol/l d-glucose) as previously described (6,19,23).

Labtek slides (VWR, U.K.) were either coated directly with 2 μg/ml laminin or derivatized for 30 min with 1 mmol/l m-maleimidobenzoyl-N-hydroxysuccinimide ester (Pierce Chemical, Rockford, IL) before coating for 1 h at room temperature with 10 μg/ml fibronectin (24). Methylglyoxal (500 μmol/l), aminoguanidine (500 μmol/l), methylglyoxal with aminoguanidine (both 500 μmol/l), or PBS was added to appropriate wells and slides incubated for 24 h at 37°C. Slides were washed three times and neurons were seeded onto slides in Bottenstein and Sato's medium and allowed to adhere to the substrate for 1 h before addition of nerve growth factor (NGF) (10 ng/ml) or glial cell-derived neurotrophic factor (GDNF) (50 ng/ml, Promega). Neurons were incubated for 18 h in 5% CO₂ at 37°C and immunostained with anti–β(III)-tubulin antibody (1:1,000, Sigma), a pan-neuronal marker labeling all cell bodies and neurites as previously described (23).

For each experimental condition, images of neurons were acquired from 20 randomly selected fields of view per condition. Assessors were blinded to the experimental condition during analysis. The mean number of neurite-bearing cells was calculated (defined as those with neurites longer than 1.5 times the associated cell body diameter). The length of the longest neurite from each cell was calculated using SigmaScan software (SPSS, U.K.), as was a measure of total neurite outgrowth: a series of concentric circles was overlaid onto an image of each neuron, and the number of times neurites crossed each circle was calculated. This provided both a measure of total neurite outgrowth (total number of cross points) and neurite branching structure (cross points related to distance from the cell body). All these measurements were made from cells in the 20 randomly selected fields in each experiment and repeated in at least four independent cultures (23). Data are expressed as means ± SD. Statistical analysis was conducted using Graphpad Prizm software using ANOVA followed by Bonferroni post hoc test.

To determine whether diabetes altered the extent of adduct residue content on endoneurial ECM proteins, we assessed 14 markers of glycation, oxidation, and nitration by liquid chromatography–mass spectrometry/mass spectrometry. Ten of the 14 markers investigated were present in detectable amounts in endoneurial ECM proteins from control and STZ-induced diabetic rats. These were fructosyl-lysine, CML, CEL, G-H1, MG-H1, 3DG-H, MOLD, pentosidine, MetSO, and dityrosine. The protein damage marker residues below the limit of detection (LOD) were GOLD, DOLD, argpyrimidine, and 3-NT. Estimates of pentosidine residues were <LOD in 33% of cases (STZ-induced diabetic study groups) and were set to the LOD in such cases for statistical analysis. For indexes of diabetes refer to Table 1.

After induction of diabetes, there were significant increases in the amount of fructosyl-lysine, CML, G-H1, and MG-H1 adduct residue contents on endoneurial ECM proteins (Fig. 1,A–D). There were no significant changes in CEL (0.083 ± 0.025 mmol/mol lys), pentosidine (0.0016 ± 0.0009 mmol/mol lys), MOLD (0.004 ± 0.003 mmol/mol lys), MetSO (31.3 ± 8.7 mmol/mol met), and dityrosine (0.29 ± 0.13 mmol/mol tyr) residue contents and a onefold increase of 3DG-H residues (1.26 ± 0.69 vs. 0.67 ± 0.13 mmol/mol arg; P < 0.05) after 6 weeks of diabetes. For MG-H1, there was a relative decline at weeks 9 and 12 of diabetes after reaching a maximum content at 6 weeks of diabetes and rebound to higher contents after 24 weeks. There were no age-related changes in adduct residue contents over the 24-week time course studied (Fig. 1 A–D; no significant differences in AGE content between endoneurial ECM protein samples at time 0 or 24 weeks post-STZ; P > 0.05). These results indicate that the increase in fructosyl-lysine, CML, G-H1, and MG-H1 residue contents of ECM proteins of the sciatic nerve endoneurium correlates with the development of STZ-induced diabetes.

To test the hypothesis that ECM proteins are susceptible to glycation under physiological conditions, we incubated laminin and fibronectin with methylglyoxal or glucose. Note that similar conditions used with human serum albumin (HSA) produced HSA modified minimally with glycation adducts (1–2 molar equivalents) (25). Control fibronectin and laminin had minor but significant contents of glycation adducts fructosyl-lysine, CML, MG-H1, and 3DG-H, ranging from 0.2–2.3 mol/mol protein and some protein oxidation (1–6 equivalents of MetSO residues) but no detectable dityrosine or 3-NT (Tables 2 and 3).

Glycation of fibronectin by methylglyoxal produced ∼28 equivalents of MG-H1. This was the major methylglyoxal-derived AGE formed (97%), the others being CEL (0.06 mol; 0.2%), argpyrimidine (0.79 mol; 3%), and MOLD (0.002 mol; 0.01%); the total methylglyoxal-derived adduct increase was ∼29.2 molar equivalents, representing 88% of the added methylglyoxal (Table 2). Similarly, glycation of laminin by methylglyoxal produced ∼56 equivalents of MG-H1 (88%), together with CEL (0.53 mol: 0.8%), argpyrimidine (6.8 mol; 11%), and MOLD (0.04 mol; 0.07%). The total adduct methylglyoxal-derived adduct increase was ∼63.1 molar equivalents, representing 93% of the added methylglyoxal (Table 3).

Glycation of fibronectin by glucose gave a relatively modest increase in glycation adduct residues: fructosyl-lysine (0.52 mol; 61% detected adduct residue increase), 3DG-H (0.32 mol; 36%), and trace amounts of MG-H1 and pentosidine (Table 2). Glycation of laminin by glucose gave much higher increases of glycation adduct residues: fructosyl-lysine (21 mol; 81%), 3DG-H (3.7 mol; 14%), CML (1.1 mol; 4%), G-H1 (0.2 mol; 0.8%), and trace amounts of CEL, MOLD, and pentosidine. As expected, pentosidine residue content was very low in all proteins analyzed, <0.01 mol% (Table 3).

Because experiments using function-blocking antibodies have highlighted the essential role that laminin and fibronectin play in promoting axonal outgrowth, we hypothesized that glycation of these proteins may contribute directly to failure of axonal regeneration (26,27).

To test this hypothesis, we examined the ability of dissociated adult rat sensory neurons to form neurites in culture. The addition of neurotrophic factors during the first 18 h promotes neurite outgrowth in defined populations of sensory neurons, which enabled us to compare regeneration of different populations of neurons on glycated versus unmodified laminin and fibronectin. Sensory neurons in culture exhibit two distinct forms of neurite outgrowth: an initial, short arborizing form of neurite outgrowth, which occurs in the absence of neurotrophic factors (but is enhanced by addition of NGF, NT-3, or GDNF), followed by a transcription-dependent switch to axon elongation. The arborizing growth is analogous to the collateral sprouting of terminal fields seen in vivo, whereas elongation is analogous to axon regeneration in vivo (28).

Survival of sensory neurons was not altered on methylglyoxal-glycated laminin in comparison to untreated laminin (assessed using trypan blue 18 h after plating, data not shown), nor was the percentage of neurite-bearing cells (Fig. 2 D).

Neurons plated on laminin extended highly arborised neurites in the presence of NGF (Fig. 2,A). In contrast, neurite outgrowth was dramatically lower on methylglyoxal-glycated laminin (Fig. 2,B). Quantification of NGF-stimulated neurite outgrowth showed no significant decrease in the length of longest neurite (Fig. 2,E) but significant reduction in total neurite density and branching, as measured by cross-point analysis (total neurite density: laminin 160 ± 28 cross points vs. methylglyoxal-glycated laminin 90 ± 18 cross points, P < 0.05; at 100 μm from cell body: laminin 32 ± 3 cross points vs. methylglyoxal-glycated laminin 17.4 ± 3 cross points, P < 0.001; Fig. 2,F and G). Typically, the neurons that extended neurites in response to NGF on glycated laminin had fewer and less-branched neurites than those plated on control unmodified laminin. To confirm this decrease was associated with glycation, methylglyoxal-glycation was conducted in the presence of the glycation scavenger aminoguanidine, which prevented the deficits in neurite outgrowth (Fig. 2 C, F, and G).

The GDNF-responsive subpopulation of neurons was similarly disadvantaged when plated on glycated laminin (Fig. 3). There were significant glycation-associated decreases in the length of longest neurite (Fig. 3,B and E), total neurite outgrowth (total neurite density: 64.7 ± 28.6 cross points on laminin vs. 21.2 ± 18.2 cross points on methylglyoxal-glycated laminin, P < 0.05; Fig. 3,B and F) and a reduction in branching structure (Fig. 3 B and G). These deficits were also prevented by inclusion of aminoguanidine.

Diabetes is associated with the initial presence of regenerative axon profiles alongside degenerative structures in peripheral nerve (29). To model this regenerative phenotype, we used a preconditioning nerve-crush injury model that can potentiate the capacity of sensory neurons to mount a regenerative response after a subsequent injury to their axons (30,31). Control sensory neurons extend neurites very poorly on fibronectin compared with growth on laminin; however, neurite outgrowth is enhanced by a preconditioning crush to the sciatic nerve in vivo 7 days before culture (23). Therefore, we used this model to investigate the impact of glycation of laminin and fibronectin on sensory nerve regeneration.

Preconditioned neurons plated on laminin extended some neurites even in the absence of exogenous neurotrophins (Fig. 4,A), but growth was strongly enhanced by addition of NGF (Fig. 4,D) or GDNF (Fig. 4,G). Glycation of laminin significantly reduced preconditioned neurite outgrowth in all treatment conditions (Fig. 4,B, E, and H). This reduction was prevented by inclusion of aminoguanidine (Fig. 4,C, F, and I). Quantification of preconditioned neurite outgrowth showed significant glycation-associated decreases in the length of longest neurite and total neurite density (Fig. 4 J and K).

Preconditioned neurite outgrowth was also observed on fibronectin in the absence of neurotrophic support (Fig. 5,A), which was strongly enhanced by GDNF but not NGF (in agreement with our previous study [23]). In neurons plated on glycated fibronectin, neurite outgrowth was significantly impaired, both in the absence of neurotrophins and in the presence of GDNF (Fig. 5,B, D, J, and K). As before, inclusion of aminoguanidine prevented the glycation-mediated inhibition. Interestingly, the addition of NGF to preconditioned neurons also prevented glycation-mediated inhibition of neurite outgrowth (Fig. 5 J and K).

These results show that ECM glycation has a dramatic inhibitory effect on the ability of sensory neurons to extend neurites in response to neurotrophic factors as well as to regenerate after a preconditioning injury.

The accumulation of proteins damaged by formation of AGE residues in the peripheral nerve in diabetes has been linked to changes in nerve structure and neuronal function and development of diabetic neuropathy (32). We have now conducted experiments that suggest that the glycation of two central ECM proteins associated with neuronal regeneration, laminin and fibronectin, may be linked directly to the failure of axonal regeneration in diabetic neuropathy. We have previously shown that cytosolic protein extracts of peripheral nerve of STZ-induced diabetic rats have increased fructosyl-lysine and AGE residue content (7). Here, we extend those findings by demonstrating that ECM proteins are also glycated in the endoneurium of rat sciatic nerves. Moreover, glycation of ECM proteins with methylglyoxal increases AGE formation and decreases the ability of adult rat sensory neurons to extend neurites.

The epidermis is innervated by the free axon terminals of small-diameter unmyelinated peptidergic and nonpeptidergic sensory neurons (33). These subsets of sensory neurons differ in their responsiveness to specific neurotrophins. The peptidergic population expresses trkA and responds to NGF, whereas the nonpeptidergic population expresses receptors for GDNF: ret and GFR (34,35). Because it is the peripheral processes of these small-diameter neurons that degenerate in clinical (36) and experimental diabetic neuropathy (37,–39), we focused our attention on the response of these populations of small-diameter neurons to ECM glycation.

Experimental diabetes in rats led to a profound increase in early glycation adduct and AGE residues in endoneurial ECM proteins of the sciatic nerve. The most abundant AGE residue was MG-H1, which increased ∼ fivefold after 6 weeks of diabetes. Similarly, fructosyl-lysine residue content increased from sixfold after 3 weeks of diabetes to 13-fold after 12 weeks. This is a higher increase than that seen for plasma glucose concentration but nevertheless is in keeping with the high (14-fold) increase of glucose concentration reported in sciatic nerve in STZ-induced diabetic rats (40). The fivefold increased content of MG-H1 residues is higher than the twofold increase in plasma concentration of methylglyoxal in STZ-induced diabetic rats and may reflect downregulation of glyoxalase 1, which provides the major defense against methylglyoxal glycation in diabetes (41,42). The accumulation of MG-H1 residues may be particularly damaging as it targets functional domains of ECM proteins, particularly arginine residues of RGD and GFOGER motifs, which are the major integrin-binding motifs (43,44).

Further evidence for the reactivity of ECM proteins toward glycation was obtained by glycation of fibronectin and laminin in vitro. We used conditions for glycation by methylglyoxal where similar preparation of HSA produced a low or minimal extent of modification (2.49 equivalents of MG-H1 per mol [45]). Fibronectin and laminin were modified by methylglyoxal to produce 28 and 56 MG-H1 residues per mol protein, respectively. The arginine residue equivalents in these incubations of HSA, fibronectin, and laminin were 2.40 mmol/l, 1.86 mmol/l, and 2.50 mmol/l, and the concentrations of MG-H1 residues formed in methylglyoxal-modified HSA, fibronectin, and laminin were 2.5 mmol/l, 4.2 mmol/l, and 4.1 mmol/l, respectively. Assuming pseudo first-order reaction kinetics for the reaction of methylglyoxal with arginine residues in these proteins, fitting of these data to a first-order integrated rate equation indicates under the conditions studied the mean reactivity of arginine residues in fibronectin and laminin is circa three- to fourfold higher than in HSA. As the RGD residues are preferentially modified, it is likely that the reactivities of arginine residues in RGD motifs of laminin and fibronectin are much higher than this. These deductions provide support for proposing ECM proteins as targets for hot spot dicarbonyl glycation in diabetes.

Similar considerations for the formation of fructosyl-lysine residues in the glycation of HSA (22) fibronectin and laminin by glucose show that the final increased concentrations of fructosyl-lysine residues in the incubation were 114 μmol/l, 8 μmol/l, and 152 μmol/l. As neither lysine residues nor glucose were markedly depleted in these reactions, an appropriate kinetic comparison may be made by considering these data as a measure of initial rates of glycation. The lysine residue concentrations in the incubations of 6.6 mg/ml HSA, fibronectin, and laminin are circa 5.9 mmol/l, 1.2 mmol/l, and 2.3 mmol/l, respectively. The reactivity of lysine residues toward formation of fructosyl-lysine residues in HSA, fibronectin, and laminin is in the ratio 1:0.35:3.5 (not accounting for fructosyl-lysine residue degradation). Glycation of ECM proteins by methylglyoxal in vivo is likely to originate from methylglyoxal formed from cellular metabolism. High-dose thiamine therapy decreased plasma methylglyoxal concentration and glycation of aortal collagen by methylglyoxal by correction of dysfunctional cell metabolism downstream of glucose, without change in fructosyl-lysine residue content (46). Together, our data show that laminin is particularly susceptible to glycation by glucose.

Glycation of laminin with glycoraldehyde (47) or glucose (48) has previously been shown to reduce attachment and neuritogenesis in neonatal sensory neurons (47) and neurite extension from mouse DRG explants (48). In this study, we use two models of regeneration to demonstrate that methylglyoxal-glycation of laminin and fibronectin causes a reduction in neurite outgrowth from specific populations of sensory neurons after stimulation with neurotrophins and a reduction in preconditioned neurite outgrowth. Glycation of fibronectin with methylglyoxal impairs adhesion of smooth muscle cells (49), retinal pericytes (50), and endothelial progenitor cells (51). Similarly, methylglyoxal glycation of the RGD domain of collagen IV inhibits αvβ₃ integrin binding and endothelial cell adhesion (43,44). We previously implicated α5β1 integrin binding to the RGD domain of fibronectin as being of key importance in the neurite outgrowth of preconditioned neurons (23) and suggest that glycation of the RGD domain is also responsible for the impaired neurite outgrowth on glycated fibronectin. ECM integrin binding not only provides a physical anchor point between the intracellular actin cytoskeleton and the outside environment, enabling traction to mechanically drive axonal outgrowth, but also directly regulates important intracellular signaling cascades and subsequent gene transcription at the site of the focal adhesion (52). Indeed, using Affymetrix gene microarray analysis we have identified over 500 genes whose expression is significantly altered in NGF-treated neurons plated on methylglyoxal-glycated laminin compared with unmodified laminin (data not shown). Replacement therapy with exogenous NGF in diabetic rats normalizes key molecular and functional aspects of neuropathy in vivo (53,54). Although the mechanism of the observed NGF-induced rescue of axonal regeneration on glycated fibronectin remains to be elucidated, our initial experiments suggest it is not mediated via RAGE (receptor for AGEs) because function-blocking experiments using anti-RAGE did not affect neurite outgrowth on glycated fibronectin.

Failure of axonal regeneration is a key feature of both clinical and experimental diabetic neuropathy (55). In uninjured peripheral nerve samples from patients with diabetic neuropathy, clusters of regenerating axons can initially be observed alongside degenerating axons; however, as the disease progresses there is a decline in the number of regenerating axons (29). The regenerative response of the peripheral nerve after injury is also blunted in diabetes; little axonal regeneration is observed after sural nerve biopsy (56) or epidermal nerve fiber injury (57). We suggest that AGE accumulation in the endoneurial ECM may contribute to this progressive failure of axonal regeneration in diabetic neuropathy.

In conclusion, we have shown that ECM proteins are glycated in the endoneurium of rat sciatic nerves in STZ-induced diabetes. Furthermore, glycation of laminin and fibronectin with methylglyoxal caused a reduction in both neurotrophin-stimulated and preconditioned neurite outgrowth from sensory neurons. This may provide a potential mechanism for the failure of collateral sprouting and axonal regeneration observed in diabetic neuropathy and may represent an important, if challenging, target for therapeutic intervention.

This work was supported by a project grant from Diabetes U.K. N.J.G. was supported by Research Councils UK (RCUK) and British Pharmacological Society (BPS).

No potential conflicts of interest relevant to this article were reported.