Early diabetic neuropathy: Triggers and mechanisms

Abstract

Peripheral neuropathy, and specifically distal peripheral neuropathy (DPN), is one of the most frequent and troublesome complications of diabetes mellitus. It is the major reason for morbidity and mortality among diabetic patients. It is also frequently associated with debilitating pain. Unfortunately, our knowledge of the natural history and pathogenesis of this disease remains limited. For a long time hyperglycemia was viewed as a major, if not the sole factor, responsible for all symptomatic presentations of DPN. Multiple clinical observations and animal studies supported this view. The control of blood glucose as an obligatory step of therapy to delay or reverse DPN is no longer an arguable issue. However, while supporting evidence for the glycemic hypothesis has accumulated, multiple controversies accumulated as well. It is obvious now that DPN cannot be fully understood without considering factors besides hyperglycemia. Some symptoms of DPN may develop with little, if any, correlation with the glycemic status of a patient. It is also clear that identification of these putative non-glycemic mechanisms of DPN is of utmost importance for our understanding of failures with existing treatments and for the development of new approaches for diagnosis and therapy of DPN. In this work we will review the strengths and weaknesses of the glycemic hypothesis, focusing on clinical and animal data and on the pathogenesis of early stages and triggers of DPN other than hyperglycemia.

Key Words: Diabetes; Pre-diabetes; Neuropathy; Impaired glucose tolerance; Hyperglycemia; Insulinopenia; Insulin-resistance

INTRODUCTION

Diabetes mellitus is a complex of metabolic disorders associated with insufficiency of insulin secretion, insulin action or both, and is manifested by hyperglycemia[1-3]. Diabetes is diagnosed when fasting blood glucose exceeds 6.9 mmol/L, or casual or 2-h glucose in a glucose tolerance test exceeds 11 mmol/L[4]. Control of blood glucose in vertebrate organisms is accomplished essentially by the action of two pancreatic hormones, i.e., insulin and glucagon, with the participation of epinephrine, ACTH, growth hormone and glucocorticoids occurring under special circumstances, such as stress. Insulin is released by islet beta cells in response to an increase of blood glucose (usually after a meal). It suppresses glucose production and stimulates the uptake and storage of glucose in skeletal muscle and liver (Figure 1). It also suppresses lipogenesis in the fat tissue and stimulates amino-acid synthesis in skeletal muscle. During a fasting state, when blood glucose is low or during stress requiring mobilization of energy, insulin secretion is suppressed and glucagon is released into the circulation by pancreatic α cells, opposing the action of insulin to increase the release of stored energy resources for use by the organism (Figure 1)[5,6].

Figure 1 Hormonal regulation of systemic blood glucose.

The metabolic effects of insulin are mediated by activation of its cognate receptors that are expressed in target tissues (skeletal muscle, liver, fat) in large excess compared to the amount needed for normal regulation of glucose metabolism (spare receptors[7-10]). This lays a background for and signifies the paramount importance of the glucose control mechanisms. Indeed, in type 1 diabetes, which is usually associated with idiopathic autoimmune attack and destruction of islet beta cells, more than 90% of islet cells need to be destroyed or less than 10% of insulin production should remain for overt hyperglycemia to manifest[11]. Similarly, in type 2 diabetes, associated in its early stages with decreased sensitivity of insulin-responsive tissues to the action of the hormone (insulin resistance) and compensatory hyperinsulinemia, no clinical diabetes develops before muscle and fat tissue sensitivity to insulin is decreased below 50%-35% of normal[9]. In rats, and similarly in humans, plasma insulin may vary over the range of 1 to 4 ng/mL with fasting blood glucose exceeding the diabetic threshold of 6.9 mmol/L (solid line in Figure 2)[12-17] only in rare cases. In most cases, insulin must decrease to less than 0.5 ng/mL level (15% of control, 3.5 ng/mL level) in order to manifest overt hyperglycemia and diabetes.

Figure 2 Relationships between rat plasma insulin and fasting blood glucose concentrations. Data are from normal and streptozotocin-injected adult Sprague-Dawley rats (the authors’ unpublished observations). STZ-rats having moderate pancreatic impairment and moderately decreased plasma insulin (vertical dashed lines) do not develop overt hyperglycemia.

Such a large safety factor for glucose control is of critical biological importance; however, clinically the impairment of insulin signaling in this disease process starts long before it manifests in overt hyperglycemia. With the discovery of insulin and improvement in techniques for blood glucose measurement, diabetes is not a life threatening disease by itself[18]. Therefore the long pre-clinical progression of diabetes could be a relatively minor issue; however, diabetes is associated with a variety of life threatening complications, among which distal peripheral neuropathy (DPN), cardio-vascular disease (CVD), retinopathy and renal disease are most frequent[2]. The problem is signs of these complications are frequently present prior to overt hyperglycemia and diabetes (Figure 2). Realization of this fact led to the definition of pre-diabetes as a state with moderate impairment of blood glucose control and high risk of development of overt diabetes, retinopathy and CVD[3,19]. Pre-diabetes is diagnosed when fasting blood glucose exceeds 5.6 mmol/L and/or 2-h glucose in a glucose tolerance test (GTT) exceeds 6.9 mM (impaired fasting glucose (IFG) and impaired glucose tolerance (IGT) dashed and solid horizontal lines in Figure 2, respectively)[4].

Establishing the correct lower limit for diagnosis of pre-diabetes is important because it determines whether clinical tests for complications should be performed and recommendations in life style and diet modifications should be presented to patients. In a recent study of young adult men, it was shown that fasting glucose exceeding 4.82 mmol/L constitutes an independent risk factor for developing type 2 diabetes in otherwise healthy subjects[20]. In another study of adult healthy men without diabetes or pre-diabetes, progressive loss of β-cell function and decrease in the first-phase of insulin secretion were detected at fasting glucose between 5.0 and 5.4 mmol/L[21]. Another clinically important issue is to understand the pathogenesis of diabetic complications, starting with pre-diabetic patients. This knowledge is required for early detection, prognosis and treatment of diabetic complications. Here we will discuss data and hypotheses for the triggers and progression of one such complication, which is distal peripheral neuropathy.

DISTAL PERIPHERAL NEUROPATHY

The prevalence of peripheral neuropathy in diabetic subjects approaches 70% and about 50% of these are cases of DPN[22-24]. The disease usually progresses to involve cardiac autonomic nerves, and as a result it is a major factor in mortality of diabetic subjects. DPN is also the major reason for loss of protective limb mechanical sensations, traumatic ulceration injures and therefore amputations[23,24]. Finally, about 11% of DPN cases are associated with chronic pain symptoms that severely diminish the quality of life and are frequently associated with depression[22,23]. The etiology of DPN is unknown and prediction of progression and treatments of the symptoms of DPN are limited[22,25-27].

Perhaps the largest problem associated with DPN, complicating its classification and treatment, is the variety of clinical presentations of this disease (Figure 3)[28-30]. Aside from a generally bilateral manifestation, distal to proximal advance and prevalence of sensory over motor impairment signs and symptoms, any two randomly selected cases of DPN may have nothing in common at the time of diagnosis and, to the extent it is known, their history and the pattern of their future progression may be very different. There are two broad categories of sensory symptoms of DPN, positive and negative[26,29]. Positive symptoms include pain, paresthesias and aberrant, exaggerated sensitivity to normally painless or moderately painful stimuli (allodynia and hyperalgesia). Negative symptoms consist of loss of sensory perception in one or several modalities. Motor symptoms, if present, manifest as muscle weakness, and thus they are also negative. Finally, a classical electrophysiological, and also negative, sign of DPN is a decrease in nerve conduction velocity (NCV) and amplitude of compound action potentials (APs) in peripheral sensory and motor nerves[23,24,26,27,31]. Furthermore, three categories of DPN, acute painful remitting neuropathy, chronic painful neuropathy and painless neuropathy with ulcer can be outlined as separate clinical entities[24,32,33]. The relationships among these entities, however, if any exist, are not known. Within each of these groups, signs of demyelination of large peripheral fibers (decrease in NCV) may or may not co-exist with signs and symptoms of large fiber axonopathy (decrease in amplitude of compound APs, loss of vibration sensation and/or loss of stretch reflexes). Loss of warm and cold perception, impairment of unmyelinated and small myelinated fibers, may or may not co-exist with signs of large fiber abnormalities[24,31,32,34-36]. Furthermore, the pain normally conducted by small unmyelinated peripheral axons may or may not be present at the same time with any of the above mentioned symptoms[22-24,26,36,37]. Finally, the modalities of pain and the degree of involvement of autonomic nervous system impairment constitute another large set of variables[22,24,33,38-40].

Figure 3 Signs and symptoms of distal peripheral neuropathy. Categories of symptoms most frequently manifested in humans with diabetes are given in bold.

From experiments in animals, and by analogy with other neuropathies, it can be suggested that the pathogenesis of negative symptoms and signs of DPN is likely to be associated with demyelination and axonal atrophy and degeneration[41,42]. Failure of re-innervation will make these symptoms essentially irreversible[42-44]. Mechanisms of neuropathic pain, paresthesias and hyperalgesia are less understood[27,40]. However, it is generally accepted that abnormally intense spontaneous input from primary afferent fibers to the spinal cord is a primary trigger of these symptoms[24,27,45,46]. At least three usually overlapping conditions, resulting in such abnormal activity of peripheral axons, are well established. First is impairment of endoneurial circulation and following it ischemia[47,48]. Second is impairment of axon-glia relationships and segmental or paranodal demyelination[49,50]. The third condition is an axonal injury and following it Wallerian degeneration and neuroma[51,52]. In addition, increased excitability of regenerating, and therefore not yet properly myelinated, nerve fibers may add to the generation of aberrant peripheral discharge and pain[53,54]. At least at advanced stages, evidence for axonal, glial and vascular injuries are detectable in most cases of DPN[41,55].

Further insight into the pathogenesis of DPN is provided by its diffuse, bilateral presentation and distal to proximal progression. The former suggests that systemic rather then local conditions underlie the clinical pathology, while the latter could indicate two possibilities (Figure 4). First, DPN may be a manifestation of dying back degeneration, with the primary insult consisting of the impairment of synthesis or efferent axonal transport of proteins, therefore affecting the function of the longest axons in the body that are most dependent on these mechanisms[42]. Failure of protein synthesis, including synthesis of some important neurotrophic molecules, in diabetes could result in impaired nerve regeneration and dying back axonopathy[43,44,55,56]. Alternatively, accumulation of the effects of multiple injuries randomly located along the axon, for example demyelinating injures, may result in a clinical picture that is practically indistinguishable from dying back neuropathy[41,42,57]. The longest axons in the body will most likely be hit by a critical number of such local insults and their function will fail first. Micro-vascular disease followed by local impairment of blood supply to the nerve fascicules may be a basis for random demyelinating insults progressing later to axonal degeneration[58-60]. It is also conceivable that both mechanisms are operating at the same time. Both demyelination and axonal degeneration were detected in human DPN by electrophysiological tests and biopsy studies[24,41,61]. Peripheral axons are the longest in the human body, and this might be an important factor explaining PNS involvement superseding the CNS complications. Another potentially important reason for the high vulnerability of PNS to diabetic injury is the relatively weak anti-oxidative defenses of peripheral neurons (see Section 3).

Figure 4 Hypotheses on distal-to-proximal progression of DPN. In normal PNS (top), neurons synthesize proteins in the cell body and transport them down the axon at the rate determined by axonal structural and functional needs. Impairment of synthesis or axonal transport of proteins will result in dying-back neuropathy, in which neurons with longest processes are affected first (middle). Alternatively, the neuropathy may result from effect of random local insults to the axon, with probability of accumulation of a critical number of such insults being higher for neurons having longer axons (bottom). Short arrows indicate non-specified axonal or neuronal injuries. Long solid or dashed arrows indicate normal or compromised axonal transport (respectively).

While the discussion above appears to encompass all the major features of DPN, multiple questions related to the pathogenesis of DPN remain unresolved. Thus whether or not dying back axonopathy or multiple local injuries or both mechanisms lead to the disease, it is not clear why, in most cases of neuropathy, sensory symptoms prevail over signs of impairment to motor axons innervating the same distal areas of the human body. Furthermore, neither of these mechanisms explains a variety of clinical presentations of DPN nor answers the question of why some diabetic patients live without any symptoms of DPN for years[62]. Finally, major questions that remain to be answered include identifying pathogenic triggers of DPN, pathogenesis of individual symptoms, and relationships among different symptoms and signs of the disease[22,24].

To illustrate the importance of the latter question, two hypothetical scenarios of DPN are shown in Figure 5. In the first scenario (Figure 5, top), there is a single pathogenic process triggering and maintaining the disease. The course of the disease and its clinical manifestations at the time of diagnosis and neurological evaluation (ovals in the Figure 5) will be determined by the duration of DPN and individual differences in the genetic backgrounds of patients. From the point of view of the symptoms revealed by a neurological exam, the second scenario (Figure 5, bottom) is identical to the first one; the critically important difference, however, is that different sets of symptoms in this scenario are triggered and driven by entirely independent pathogenic mechanisms (symptoms “a”, “c”, and “d”vs symptom “b”). Some of the branches of the pathogenic process (“b” in the first scenario) may enter an irreversible stage. Therefore, the early detection of DPN is an obligatory condition for the successful treatment of this disease. The cartoons demonstrate also that identification of all participating triggering mechanisms and symptoms associated with them is another critical step for the efficient treatment of DPN.

Figure 5 Hypothetical branching (top) and multi-trigger (bottom) pathogenesis of DPN. In the first scenario (top) all manifestations of the disease result from the unique branching pathogenic process, and symptom “b” discovered at the time of neurological exam B is not corrected by treatment because it has already progressed to an irreversible stage (dashed lines). Earlier diagnosis and institution of treatment (at time A) may critically change the outcome of therapy in this scenario. Alternatively (lower scenario), several independent factors may trigger and maintain the progression of DPN. In this case the therapy may fail to treat symptoms not because they are irreversible, but because the correct cause of the pathology is not identified and is not treated (dashed lines).

CHRONIC HYPERGLYCEMIA AND PATHOGENESIS OF DPN

DPN follows both type 1 and type 2 diabetes, and systemic hyperglycemia is the most obvious symptom that these types of the disease have in common[24,63,64], suggesting hyperglycemia as a universal trigger for DPN. Indeed, insulin treatment or treatment with insulin-sensitizing drugs to control hyperglycemia reverses some symptoms of DPN and delays its progression in general[65,66]. The Diabetes Control and Complications Trial (DCCT) data show that strict control of hyperglycemia in type 1 diabetes patients without clinical neuropathy decreased development of DPN in 60% of cases over 5 years of follow up study[67,68]. Well within the framework of the glycemic hypothesis, the failure to prevent all cases of DPN could be explained by the fact that glucose control can never be perfect. Type 1 and 2 patients are, on average, euglycemic for only about 62% of the day, while during the remaining 30% and 8% of the day they have various degrees of hyperglycemia and hypoglycemia, respectively[69]. Therefore, to avoid hypoglycemic crisis, the acceptable target value for blood glucose in controlled subjects is usually set to values above normal (6.7 to 10 mmol/L in DCCT and 6 mmol/L in U.K. Prospective Diabetes Study; of type 2 diabetic patients[65,68]). Another explanation for incomplete efficacy of glucose control is that, after long-standing diabetes, some neuropathic mechanisms may enter either an irreversible stage or a stage of progression that is already independent of the original trigger. With the limitations imposed by generally late diagnosis of diabetes and DPN[24,70,71], slowing of NCV, paraesthesia and painful symptoms appear to be the earliest (closest to the initiating pathogenic insult) manifestations of DPN. Therefore, it appears to be in good agreement with the glycemic hypothesis that in patients with newly diagnosed diabetes, NCV slowing and paresthesias (hyperglycemic neuropathy) can be frequently completely recovered with the establishment of euglycemia[31,42].

Further support for the glycemic hypothesis comes from research in animals, specifically rat models of type 1 diabetes (STZ-induced or spontaneous in BB-rats) and spontaneous type 2 diabetes in Zucker fatty rats[66,72] that appear to be the best studied animal models with regard to neuropathy. The short life span of rodents severely limits evaluation of chronic human diseases in these animals. Another limitation is that in behavioral tests, evoked pain manifested by limb withdrawal can be tested, but neither spontaneous pain nor changes in non-nociceptive sensory thresholds can be reliably measured in animals. Nonetheless, in general agreement with both clinical data on the earliest signs and symptoms of human DPN and the glycemic hypothesis, slowing of sensory[73-76] and motor[74,75,77-83] NCV and manifestations of evoked pain (hyperalgesia[75,84-87] and allodynia[85]) were shown to develop within the first month of onset of hyperglycemia in diabetic rats. With a longer time allowed (six to twelve months of diabetes) signs of axonopathy, demyelination and nerve degeneration can also be detected in diabetic animals[56,74,80,82,88-90]. Finally, early in the course of diabetes in rat models impairment of endoneurial blood flow and micro- and macrovascular reactivity are reported by many investigators[75,91-94]. Skin and arterial blood flow is abnormal early in diabetic patients[95-98], but no reduction in sural nerve blood flow was detected in humans with diabetes and mild DPN[99]. Thus, it is not clear whether impaired endoneurial blood flow represents a rat-specific component of DPN or if it is missed in humans because of its transient character and usually late detection of DPN in diabetic patients. In agreement with the glycemic hypothesis, all abnormalities found in rat models are reversible with normalization of blood glucose in insulin replacement experiments, and some of them (decreased pain pressure threshold) can be induced in normal rats by chronic in vivo perfusion of a DRG, or a segment of sciatic nerve with hyperglycemic solution[100,101].

Finally, support for the glycemic hypothesis is provided also by studies of cellular pathology associated with experimental diabetes. These studies show that practically all signs and symptoms of DPN observed in animal models may be linked to hyperglycemia-induced metabolic impairment of nerve, glial and endothelial cells in PNS. A detailed description of these studies is beyond the scope of the present work and can be found in a number of recent reviews[94,102-109]. The purpose of Figure 6 is only to provide a brief outline of cell metabolic abnormalities associated with diabetes and emphasize the findings directly relevant to the following discussion of triggers of early DPN.

Figure 6 Hyperglycemia, derangement of cell metabolism and oxidative/nitrosative stress. Hyperglycemia associates with accumulation of fructose-6-phosphate and hexosamine pathway (1) to N-acetylglucosamine, accumulation of dihydroxyacetone phosphate and associated activation of PKC (2), activation of polyol sugar pathway (3), and glucose autoxidation and non-enzymatic protein glycation (4). These metabolic events are either regular physiologically important components cell metabolism (1, 2 and 3) or are normally under strict control of intrinsic intracellular defense mechanisms (4). However, under conditions of chronic hyperglycemia activation of these pathways leads to a global derangement of the cell and tissue homeostasis, which culminates in an uncontrolled cascade of abnormal protein modifications, oxidative/nitrosative stress and pro-inflammatory conditions.

Although the scheme in Figure 6 is simplified and omits many essential steps for the cascade of events [the poly (ADP-ribose) polymerase (PARP)[108]; activation and consequences of lipid peroxidation[110] are not shown], it clearly demonstrates the complexity of this cascade. Some events appear to be more specific to one type of cell than to another, and there is as yet no agreement on which events (for example activation of polyol pathway[111] or abnormally intensified oxidative phosphorylation[106]) plays a leading role in cellular impairment. Nonetheless, most of the data available indicate that all the various pathways activated by hyperglycemia converge in generation of excess reactive oxygen species (ROS). This process eventually overwhelms the intrinsic anti-oxidant mechanisms of the cell and ends in oxidative/nitrosative stress and pro-inflammatory conditions in the tissues[112-115]. The efficacy of treatment with anti-oxidants in correction of DPN in animals and humans supports this view[116-119].

In diabetic animals, oxidative stress injury develops in parallel in all major cellular elements of PNS. Injury to glial cells is responsible for the demyelination component of DPN, which may explain the decrease in NCV and painful manifestations of the disease. Oxidative stress in neurons might be responsible for axonopathy, impaired regenerative capacity of axons and negative symptoms of DPN. Glial cell injury will affect the nerve neurotrophism adding to the progression of neuronal defects. Finally, oxidative stress and impairment of nitric oxide (NO) production in the endothelium of epi- and endoneurial blood vessels results in impairment of endoneurial circulation and endoneurial hypoxia, exaggerating and speeding up the direct effect of hyperglycemic conditions on glial cells and neurons. Oxidative stress is pro-inflammatory, which affects the production of cytokines by glial cells, and provokes the recruitment of immune response cells into the affected tissue. This might be another important component of the pathogenesis and progression of DPN. Thus, combining these data and observations lays the foundation for a view of the natural history of DPN similar to that depicted in Figure 7 according to an expanded view of the glycemic hypothesis.

Figure 7 Pathogenesis of DPN with hyperglycemia as a major trigger of PNS injury.

Perhaps the most attractive aspect of such a view of the pathogenesis of DPN is that while it suggests a unifying trigger and mechanism (hyperglycemia and oxidative stress) for all symptoms of DPN, it nevertheless remains flexible enough to leave room for individual variability in the rates of progression and spectrum of manifestations of the disease. Indeed, the actual effects of uniform pro-oxidative and pro-inflammatory conditions may differ sharply depending on differences of individual cells and tissues in their intrinsic anti-oxidant defenses and individual organisms in their immune defenses. The injuries to a single myelinating or non-myelinating Schwann cell, endothelial cell or neuron will unlikely have even subclinical significance. The death of several myelinating cells will result in a decrease of NCV in a given axon and injury to several endothelial cells may cause the closure of a given capillary. Yet there are many axons and there is regeneration of damaged axons, and there are many capillaries and regeneration of damaged and collateral capillaries. It is only after cellular defenses against the oxidative stress are overwhelmed in many cells that the sub-clinical signs of the disease may be expected to appear, and it is anticipated to take an even longer time before clinical manifestations will themselves appear. In agreement with this, en mass CNS cells appear to have a much higher capacity for anti-oxidative mechanisms than do PNS cells and this might be one of the reasons why diabetic neuropathy affects the CNS much later than it affects peripheral nervous functions[90,113,120].

DIFFICULTIES OF THE GLYCEMIC HYPOTHESIS

While apparently logical and consistent with many clinical observations and the results of animal studies, the glycemic hypothesis cannot completely explain all the data. Thus, experiments in rodents consistently demonstrate that the polyol pathway (conversion of glucose to sorbitol by aldose reductase and then to fructose by sorbitol dehydrogenase; pathway 3 in Figure 6) is an important source of reactive oxygen species, and inhibition of aldose reductase prevents or reverses many signs of DPN seen in diabetic animals[121]. The same treatment in humans, however, has shown questionable efficacy so far[27,53,122,123]. Other problems include the failure of a pre-clinical slowing of NCV[24,124] or increase in glycosylated hemoglobin[22] (HA1c, integral measure of persistent hyperglycemia; pathway 4 in Figure 6) to predict development or severity of symptoms of DPN. Also, the finding of an inverse correlation between hyperglycemia and pain severity in diabetic patients with remitting painful neuropathy at presentation[33] is inconsistent with the glycemic hypothesis. These and other similar inconsistencies certainly could be attributed to the complexity of the disease, differences in studied patient populations, or inadequate design of drug trials in terms of timing, duration or dose[121]. Similarly, the lack of absolute efficacy in glucose control in preventing DPN in humans with diabetes[68] could also be explained within the framework of the glycemic hypothesis considering the difficulty of maintaining blood glucose concentrations within the relatively narrow range of normal values, which suggests that the actual threshold for neuropathic effects of hyperglycemia is lower than was previously thought. Therefore, recent findings of increased incidence of DPN in patients with pre-diabetes, many of whom have an impaired glucose tolerance (IGT) but not fasting hyperglycemia[22,25,70,71,125-127], appear to present the most serious challenge for the glycemic hypothesis.

The possibility that PNS injury may be triggered by exaggerated and prolonged postprandial hyperglycemic episodes, without necessarily requiring chronic hyperglycemia, should be considered to reconcile the glycemic hypothesis with observations of DPN in glucose intolerant patients[70,128-130]. Indeed, indices of large fiber function (ankle and knee reflexes and vibratory perception thresholds) were shown to decay with impairment of glucose tolerance in humans without diabetes and neuropathy[131]. Furthermore, pain is a frequent symptom of pre-diabetic DPN. An acute glucose infusion, which could be considered as an analog to a postprandial glucose surge, decreased thresholds to electrical stimulation in healthy adult volunteers[132] and decreased pain pressure thresholds in type 1 diabetic patients without clinical neuropathy[133]. In the latter study, however, no association between acute hyperglycemia and heat pain, warmth/cooling or vibration perception thresholds, was found[133]. Inconsistent with the idea that postprandial glucose changes have a neuropathic effect, no correlation was detected between short-term fluctuations in blood glucose and pain scores or heat pain thresholds in the study of type 1 and type 2 diabetic subjects with painful neuropathy[134]. Furthermore, no correlation between the number of glycemic excursions and the number of painful episodes was found in the study of type 1 patients with painless neuropathy[135].

Difficulties with the glycemic hypothesis are not unique to the human clinic. Thus in the STZ-rat model of diabetes, NCV could be corrected by a low level of insulin therapy below that required to correct hyperglycemia[89,136]. Pain pressure and von Frey filament thresholds studied in the same model demonstrate no correlation with the degree of hyperglycemia[84,101,137]. Furthermore, aldose reductase inhibitors (blockers of the polyol sugar pathway; Figure 6), given at doses sufficient to correct nerve sorbitol and fructose and heat pain thresholds, do not correct von Frey filament threshold in STZ-hyperglycemic rats[138,139]. As another example, pain pressure thresholds in type 2 diabetic Zucker rats could be corrected with insulin-like-growth factor II (IGF-II) that has no effect on blood glucose[86]. All these examples are taken from experiments in overtly diabetic and hyperglycemic animals. Therefore, formally the possibility remains that hyperglycemia was the triggering event for the observed abnormalities, but it is not required for the progression and maintenance of these pathologies. Whether DPN develops in pre-diabetic rats as it does in humans has not yet been studied. Since previous work has focused on diabetes, little attention has been devoted to the development of pre-diabetic animal models and studies of DPN in these models. Nonetheless, neuropathic decreases of mechanical and thermal nociceptive thresholds[140] and slowing of motor NCV[141] were observed in studies in Zucker-fatty rats. Since these are insulin-resistant but normoglycemic animals (type 2 pre-diabetes) the impaired glucose tolerance could be responsible for DPN in these animals. Recently, we described decreased pain pressure threshold in rats that were injected with STZ but remained normoglycemic[84]. These rats also had normal glucose tolerance, maintained normal levels of HbA1c, and normal concentrations of sorbitol in the nerve, suggesting that not only fasting but casual glucose also was maintained within physiological limits (Figure 8).

Figure 8 Pain pressure thresholds are not correlated with hyperglycemia. Weight (A), blood glucose (B), nerve sorbitol (C) and pain pressure thresholds (D) in control, STZ-normoglycemic and hyperglycemic rats (white, grey and black columns respectively; 2 wk after injection of 65 mg/kg STZ).

Thus there is solid evidence that hyperglycemia is an important factor of DPN. However, there are also both clinical and animal studies indicating that in addition to chronic and/or postprandial hyperglycemia, other pathogenic mechanisms must exist that trigger and maintain at least some of the symptoms of DPN. Identification of these factors is of critical importance for our understanding of both the natural history of pre-diabetic DPN and the pathogenesis of diabetic DPN in general[27,88].

INSULIN SIGNALING AND DPN

In the search for triggers of DPN other than hyperglycemia, it is important to note that successful reversion or postponing of DPN in clinical glucose control trials does not necessarily prove the glucose hypothesis[142]. Glucose metabolism is regulated by insulin via type B receptors (IR-B) abundantly expressed by liver, skeletal muscle and fat cells. Insulin receptors, however, are also expressed in central and peripheral nervous systems[143,144]. Furthermore, in PNS the highest densities of IRs are located on endothelial cells, paranodal loops of Schwann cells and medium and small size primary sensory neurons[143,144]. All these locations are strategically critical points of PNS function considering what is known about the pathogenic mechanisms of DPN. While the nervous system has mostly type A receptors (IR-A), the insulin affinities of these and IR-B receptors are similar (K_0.5 is 3 to 6 ng/mL[145]) and well within the range of circulating concentrations of insulin (1 to 6 ng/mL[12-17]). Thus, it is very possible that correction of insulin levels in treatments for Type 1 diabetes or insulin-sensitizing therapy in cases of type 2 diabetes not only corrects glucose metabolism, but also has independent effects on the function of PNS. Serious consideration of this hypothesis is warranted, first because it provides an explanation for at least some failures of glycemic controls to reverse DPN, and second because it may explain the development of DPN in pre-diabetic patients.

Insulinopenia (Type 1 diabetes and pre-diabetes)

As described in the Introduction, overt type 1 diabetes is preceded by a state of partial pancreatic damage and moderate insulinopenia in which insulin production is still satisfactory for blood glucose control[11]. The type 1 pre-diabetic state is probably short and little is known about neuropathy in these patients. The results of a recently completed follow-up to DCCT study of type 1 diabetic patients, however, has shown that regardless of the level of glycemic control, neuropathy is less prevalent in the group of patients that maintained intensive vs. conventional insulin therapy[146]. Further support for the possibility of a direct role of insulinopenia in DPN comes from experiments in the STZ-rat model of type 1 diabetes. It was demonstrated that local insulin application to the nerve prevents motor NCV slowing in STZ-hyperglycemic rats[147]. Similarly, systemic[136] or intrathecal[89] application of insulin can correct sensory and motor NCV in STZ-treated rats without having an effect on hyperglycemia. Finally, in our experiments in rats that were injected with STZ and developed moderate insulinopenia but not fasting hyperglycemia (Figure 8), pressure pain thresholds were decreased in proportion to the degree of insulinopenia, and low dose insulin-replacement therapy corrected this defect without changes in the systemic blood glucose level (Figure 9). Taken together these data suggest that at least some signs of neuropathy (slowing of NCV, pressure-evoked pain in rats) may indeed be triggered by insulinopenia with no relevance to the blood glucose level. Another notable aspect of these experiments is that correction of nerve conduction[89,136,147] and pain pressure thresholds (Figure 9) with insulin treatment could be achieved without changes in systemic blood glucose levels, leading to an important implication that the thresholds of “metabolic” and “neuropathic” effects of insulinopenia may differ.

Figure 9 Insulin-dependence of pain pressure threshold in STZ-normoglycemic rats. Two weeks after injection of STZ pain pressure threshold of STZ-NG rats is decreased in proportion to the plasma insulin level (R = 0.97; A). Insulin replacement initiated one week after STZ injection (horizontal bar in panel B) does not change PPT of control or hyperglycemic rats, but corrects it in STZ-normoglycemic animals (empty and filled circles, and filled triangles, respectively). Insulin replacement, does not affect blood glucose (C), but normalizes plasma insulin level (D) in STZ-NG rats. In C and D white, grey and black columns represent control, STZ-NG and STZ-HG rats, respectively.

To date, detailed information on the relation between insulin and nerve conduction is not available. However, comparison of better studied “dose-response” relationships between insulin, blood glucose and pain pressure thresholds (Figure 10A) allows speculation that in the rat, control of glucose metabolism may tolerate at least five times lower insulin levels than does nerve function. Given that this difference was confirmed in both animals and humans, the outcome of these studies is of a great importance. This finding may explain the development of neuropathy in pre-diabetes and also suggests that neuropathy may start at stages preceding pre-diabetes, and some therapeutic interventions to correct insulin levels or insulin resistance (see next section) are warranted in pre-diabetic patients.

Figure 10 Putative “insulin-glucose metabolism” and “insulin-nerve function” relationships in normal and insulin-resistant rats. In A: Normal rats: Empty circles represent fasting glucose and filled circles pain pressure thresholds measured in control, STZ-NG and STZ-HG rats, pooled together and normalized to show relative changes of these parameter between control and diabetic animals. The data were fitted by Hill curves calculated based on the assumption that insulin binds to the receptor with an affinity of 1 ng/mL (dotted curve) and 10% and 65% occupancy of these receptors is required for maximum metabolic (intersection with dashed curve) and nerve (intersection with solid curve) effect of the hormone, respectively. In B: Insulin-Resistant Rats: Insulin resistant state was simulated by increasing K_0.5 of insulin binding to the receptor to 20 ng/mL. Points of intersections of the dotted curve with the dashed and solid curves recalculated with the new K_0.5 parameter show (arrows and labels) that maintaining glucose metabolism now requires about 6 ng/mL of plasma insulin, and nerve function requires at least 28 ng/mL of the hormone.

Differences in threshold concentrations of a ligand are usually determined by the differences in receptor properties. However, insulin affinities of IR-B and IR-A isoforms of the insulin receptor expressed in cells of organs responsible for glucose metabolism and in nerve tissue are too close to account for apparent differences in the concentrations of hormone required for maintaining normal blood glucose concentrations and normal pain pressure thresholds[145]. On the other hand, strong correlative relationships between insulin and pain pressure thresholds (Figure 9A) suggest a nearly direct link between insulin regulation and pressure pain mechanisms. The hypothesis of spare receptors[7,8,10] is the easiest way to explain this discrepancy. Thus as shown in Figure 10A, both metabolic and neuropathic effects of insulin may be described adequately within the concept of insulin binding with the same affinity (K_0.5 = 1 ng/mL) in nerve and in glucose controlling organs, if 65% of the receptor occupancy is needed to control nerve function, and only 10% of occupancy is required for glucose metabolism. Hill equations were used in this simulation; however, since the purpose was merely to illustrate the potential possibility of the given scenario, parameters of the equations were adjusted by a trial and error approach and no attempt was made to optimize the parameters. Some alternatives to this scenario will be discussed in section 6.1 of this review.

Insulin resistance

Prevalence of type 2 to type 1 diabetes is about 9 to 1 and most of the cases of human pre-diabetes are type 2 pre-diabetes or metabolic syndrome cases[24,125]. Over the long-term, insulin production is impaired in type 2 diabetes further increasing the incidence of DPN in this population by mechanisms described above. In a ten-year study of the natural history of type 2 diabetic patients, it was found that decreased serum insulin and increased blood glucose concentrations are independent predictors of DPN[148]. However, in early type 2 diabetes and pre-diabetes there is a compensatory hyperinsulinemia. Because of this hyperinsulinemia type 2 pre-diabetes usually spans a much longer period of time than does type 1 pre-diabetes. Thus, after 5 years only 20% to 35% of patients with impaired fasting glucose or impaired glucose tolerance develop overt hyperglycemia (see[70]). Despite this compensatory hyperinsulinemia, however, many type 2 pre-diabetic patients do develop DPN[125,126]. This latter observation suggests that in terms of neuropathic outcome, insulin resistance and insulinopenia may be equivalent states. It also suggests that increased production of insulin may fail to compensate for decreased sensitivity of PNS to regulation by insulin.

Not much experimental data exists to verify the validity of either of these suggestions. In studies in normal human volunteers, warmth detection threshold correlated with insulin but not fasting or 2-h GTT glucose, leading the authors to suggest that insulin resistance may determine some sensory functions of PNS[149]. This is also supported by observations of decreased NCV[141] and pressure pain thresholds[86] and the authors’ unpublished observations in the Zucker fatty rat model of type 2 pre-diabetes. However, whether it is possible that hyperinsulinemia is effective in regulating glucose metabolism but fails to compensate for nerve insulin resistance is at this time absolutely not known. Furthermore, it is unknown if this mechanism plays a role in DPN associated with type 2 diabetes. In theory such a possibility does exist, and Figure 10B illustrates the scenario that we believe provides a useful working hypothesis for future experiments. Using “dose-response” relations as a starting point, shown in Figure 10A, the insulin resistant state that leads to type 2 diabetes may be modeled as a decrease in affinity of IR from 1 to 20 ng/mL, which leads to proportional rightward shifts of insulin-glucose metabolism and insulin-nerve function relationships. If the requirements of 10% and 65% occupancy of IR remain unchanged, maintenance of normal glucose metabolism under these new conditions will need about 6 ng/mL of circulating insulin and nearly 30 ng/mL insulin concentration will be the minimum needed to maintain normal PNS function. The calculated 6 ng/mL insulin concentration is in the range of insulin concentrations measured in Zucker fatty rats (5.1 to 11.7 ng/mL; model of compensated insulin resistance[150-152]). These latter numbers are, however, significantly lower than predicted by the model insulin concentration needed to maintain nerve function. Therefore, it may be speculated that the “set-point” or natural goal of compensatory hyperinsulinemia is merely to correct glucose metabolism, which is vitally important for the organism, with no concern about the less significant problem of nerve function.

Cellular mechanisms

Thus, while the connection remains speculative, the data above suggest that impairment of insulin signaling in PNS (because of decreased insulin production, insulin resistance or both) may be an important factor in the pathogenesis of DPN. Further studies are needed to confirm this hypothesis and further studies are also needed to understand the cellular mechanisms of insulin action in PNS.

Glucose is a major fuel for neurons of peripheral and central nervous systems. However, unlike that in major target tissues of insulin regulation, uptake of glucose in nervous tissue is an insulin-independent process. Therefore, the simple explanation of the neural effects of insulin to regulate the energy supply does not appear to be applicable. There should be some other role of IR in the nervous system. In the CNS, these receptors are involved in the insulin control of feeding behavior, reproductive and cognitive functions and neuromodulation[27,153]. Insulin also clearly has neurotrophic functions. It stimulates neurite outgrowth, is involved in peripheral nerve regeneration and is required for survival of sympathetic neurons (see[27]). These effects are likely very important for regeneration, which is suppressed in long-term DPN. For short term diabetes, the possibility of insulin regulation of axon-glia relationships, vascular permeability, and function of nociceptive primary afferent neurons[143,144] may be of importance. The possibility of a selective acute effect of insulin on endoneurial blood flow (decrease) was also demonstrated by experiments in normal rats[154].

Note also that the negative cellular effects of hyperglycemia and insufficiency of insulin signaling seem to converge at some point. Pain pressure thresholds are decreased in rat models of local hyperglycemia with a time course and to a degree that is very similar to those in STZ-normoglycemic rats[101]. Since many studies support oxidative/nitrosative stress as a central event of hyperglycemic impairment, it is reasonable to suggest that oxidative/nitrosative stress can interfere with insulin regulation in some or all of the hyperglycemia-induced steps of the pathogenic cascade depicted in Figure 6. Insulin directly regulates inner mitochondrial membrane potentials and may affect oxidative phosphorylation[136]. Insulin also suppresses expression of NADPH oxidase[155] and controls expression of Nf-κB and associated inflammatory reactions[155,156]. In fact, the anti-inflammatory effects of insulin have been known since the discovery of the benefits of insulin therapy in systemic inflammatory responses to trauma or bacterial infection[157]. Insulin signaling also was shown to be linked to the regulation of Na, K-ATPase[158,159] and endothelial NO production[158,160]. These data suggest that insulinopenia does have the potential to produce the same or very similar neuropathic effects as were attributed previously solely to hyperglycemia.

OTHER FACTORS OF SIGNIFICANCE

While the above outlined insulin-signaling hypothesis of DPN appears compelling, it will certainly be corrected and modified in many of its segments to conform to the results of future studies. It can also be stated here, without reservation, that no complete picture of the pathogenesis of DPN will be created unless the roles of insulin-like growth factors and C-peptide are considered in addition to hyperglycemia and insulin signaling in PNS[142].

Insulin-like growth factors (IGFs)

IGFs are produced in the kidney, spinal cord, skeletal muscle and peripheral glia. IGFs possess multiple neurotrophic functions, including control of neuronal survival, neurite outgrowth and regeneration, and expression of genes encoding axonal cytoskeletal proteins (tubulins and neurofilaments)[142,161-166]. Interestingly, IGF-1 appears to be involved in the regulation of resistance to oxidative stress[167]. IGFs primarily act via specific receptors, but since IGFs are present in the circulation in a 100-fold excess compared to insulin, they may also bind to and activate IR, mimicking some but not all affects of insulin[168,145,165]. It might be important in this regard that, in peripheral nerve, IGF-I receptors are co-localized with IR in sensory neurons[169,170] and Schwann cells[171] and a large fraction of them are likely hybrids composed of protein subunits of both IR-A and IGF-I receptors[143,169]. These hybrid IR/IGF receptors have substantially higher affinity to IGF-I than to insulin[172,173]. However, even when present in physiological concentrations, insulin may still bind to and activate some portion of hybrid as well as homomeric IGF-I receptors[161,174]. In addition, IGF-I may act by suppressing growth hormone and improving insulin sensitivity and insulin production in type 1 and type 2 diabetic subjects[174-178]. Insulin on the other hand, may modify kidney production of IGFs, and via regulation of IGF binding proteins, it may control the activity of circulating IGFs[161,175,176]. Whether, any of these multiple mechanisms participate in the apparent dissociation of the metabolic and neuropathic effects of insulinopenia remains to be determined.

Abnormal expression and levels of circulating IGFs and/or changes in expression of receptors for IGF were measured in diabetic human subjects[174], in STZ-rats (see[179-181]), and in the type 2 diabetes Zucker diabetic fatty (ZDF) rat model[86]. Furthermore, in obese Zucker rats, both insulin- and IGF-I resistances were shown to develop and mediate impaired glucose tolerance in this model of pre-diabetes[182]. Thus potentially, via impairment of protein synthesis, insufficiency of IGFs may add to the pathogenesis of regenerative capacity, neurodegeneration and irreversible stages of DPN[142]. This suggestion is supported by observations of recovery of NCV and reversion of atrophy of myelinated sensory axons in the sural nerve of STZ rats treated with intrathecal IGF-I[89]. In addition, the defect in IGFs or IGF-receptor expression could also add to the pathogenesis of early symptoms of DPN either directly or through modulation of insulin production or nerve sensitivity to insulin. Indeed, down-regulation of IGF-I receptors, which is observed in nerves of STZ-diabetic rats, occurs comparatively early, within 1 week after the onset of hyperglycemia[181]. Furthermore, continuous subcutaneous infusion of IGF-II was shown to recover pain pressure thresholds to a normal level after 6 wk of diabetes in the ZDF rat model[86]. The latter observation is interesting because, in STZ-diabetic rats, a similar magnitude and early decrease in pain pressure thresholds seems to result from insulinopenia (see previous section). Affinity of IGF-II binding to brain type insulin receptors or to IGF-I-R is two-three times lower than that of respective natural ligands[145]. Therefore, the effect of IGF-II on mechanical hyperalgesia may still be explained within the framework of our insulin signaling hypothesis of early neuropathy. However, the possibility of a more complex regulation of pain pressure thresholds cannot be excluded. It is important also that in all the examples above the effects of treatments with IGFs occurred with no measurable changes in the glycemic status of studied animals; once again suggesting that the pathogenesis of DPN is multifactorial.

C-peptide

C-peptide is a segment of the proinsulin molecule sliced off to form insulin. Acting through both its own receptors and modulating activity of insulin receptors, C-peptide produces multiple insulin and IGF-like effects[27,183,184]. C-peptide also enhances autophosphorylation of IR and effects of insulin, and treatment with C-peptide reverses decreased expression of IGF-I, NGF and neurotrophin-3 receptors in type 1 spontaneously diabetic rats[27,183,185]. Considering these effects and the fact that C-peptide is secreted in equimolar concentrations with insulin, it can be concluded that C-peptide insufficiency may have an important role in the pathogenesis of type 1 diabetes[75,183,184]. In agreement with this implication, in patients with recently diagnosed type 1 diabetes, C-peptide treatment was shown to correct sensory NCV and vibration perception[186]. C-peptide treatment also corrects skin microcirculation in diabetic patients[187] and endoneurial blood flow and NCV slowing[75], thermal hyperalgesia, atrophy and degeneration of C-fibers[185] in BB/Wor type 1 diabetes rat model.

As expected, no differences in plasma C-peptide levels were found in pre-diabetic patients and patients with type 2 diabetes of short duration (less than 5 years from diagnosis), even though the number and severity of the signs and symptoms of DPN differed substantially between these groups[128]. What is less clear in the same study is a lack of detectable deficiency of C-peptide in advanced type 2 diabetes patients (more than 5 years) who could have been expected to start developing insulinopenia. In general, however, multiple questions remain to be resolved in relation to C-peptide and its role and mechanisms of action in DPN. Similar to conditions of chronic or postprandial hyperglycemia or impaired insulin- or IGF-signaling, C-peptide deficiency appears to affect activity of Na,K-ATPase, NO production and neurotrophism, but the C-peptide mediated regulation does not appear to depend on oxidative stress, which is apparently important in the pathogenesis of these conditions[75,188]. It is also not clear whether any of the symptoms may be directly attributed to C-peptide insufficiency in DPN. Endoneurial blood flow, NCV and heat nociception appear to be the foremost candidates[186,187], but antioxidant treatments and insulin replacement correct these abnormalities in STZ rats at least as efficiently as does C-peptide replacement in the BB/Wor rat model of type 1 diabetes.

CONCLUSION

DPN is a frequent and troublesome complication of diabetes mellitus. Diabetes manifests in a case-specific variety of signs and symptoms, and associates with complex biochemical, functional and structural abnormalities of the peripheral nervous system. While the obvious hyperglycemia present in diabetes can explain the development of these abnormalities, data suggest that other factors may also contribute. We have discussed the evidence for insulinopenia in type 1 diabetes or insulin resistance in type 2 diabetes as causal factors in the development of DPN. We have suggested that these two cases actually represent a single cause of impaired insulin signaling. Considering the role of insulin signaling in DPN more completely explains the changes in nerve function in pre-diabetic or early diabetic patients and animal models. This does not preclude hyperglycemia also as a factor in DPN, but allows a more complete picture of the disease process (Figure 11). In addition to insulin signaling, some evidence also exists for a role of IGF and C-peptide in mediating DPN. The pathogenesis of DPN is obviously multifactorial, and despite long-standing efforts, remains poorly understood. This work has also suggested that insulin signaling has effects that occur with different levels of receptor occupancy. Thus insulin action on nerve function requires a higher level of receptor occupancy than does insulin action on glycemic control. This could be explained by different levels of receptor coupling to second messenger signaling pathways in nerve versus liver or muscle. Future studies should elucidate these mechanisms so that a clearer picture of DPN can be obtained, especially in pre-diabetes where early detection could improve therapeutic outcomes.

Figure 11 Modified scheme of pathogenesis of diffuse diabetic neuropathy with inclusion of insulin signaling. According to this view, the disease process starts before impairment of glucose metabolism becomes apparent. Derangement of insulin signaling in PNS triggers and maintains the neuropathy at this stage. Postprandial and chronic hyperglycemia is not the least important factor of DPN, but they become involved at relatively advanced stages of DPN. C-peptide and IGFs insufficiencies represent another important set of pathogenic mechanisms; however, the role of these mechanisms in early DPN remains undetermined.