Sensorimotor diabetic neuropathy is a predominantly axonal process (Behse and Buchthal 1977; Dyck et al. 1986a, b; Johnson et al. 1986; Yagihashi and Matsunaga 1979; Schmidt and Bilbao (in press)) (Fig. 17.1a, b). Behse and co-workers (1977) observed that in predominantly sensory neuropathies, large and small myelinated fibers were equally depleted, while in sensorimotor neuropathy the large myelinated fibers (MFs) were more severely affected (Behse and Buchthal 1977). However, Dyck et al. (1986b) found no evidence of selective depletion of small or large MFs. Regenerating clusters can be prominent, but with severe neuropathy they often diminish in number. Although the neuropathy can be very chronic, actively degenerating axons may be seen, and the overall appearance is not as indolent as that of neuronal Charcot–Marie–Tooth disease (CMT). Nonetheless, studies of “early” diabetic neuropathy have reported normal myelinated fiber density, axonal area, and G-ratio in the presence of teased fiber studies showing paranodal abnormalities, segmental demyelination, and remyelination without active degeneration of myelinated axons.

We have been impressed by the frequency of focal nerve lesions in these patients, often appearing as perineurial-based regions where myelinated fibers are reduced in number, actively degenerating, or thinly myelinated (Figs. 17.2, 17.3, and 17.4). The perineurium at such sites may appear disrupted, and rarely a few randomly directed myelinated fibers may escape the confines of the nerve fascicle, forming a miniature “traumatic” neuroma (Figs. 17.2 and 17.4) which appears most common in nerves of diabetics. It is intuitive to regard such abnormalities as nerve “infarcts” although this implies an ischemic pathogenesis (vide infra). The epineurial vasculature is typically less severely involved, although vasculopathy has been reported.

Sima and colleagues have suggested a difference in the pattern of axonal pathology depending on patient age and type of diabetes: younger patients with IDDM are less likely to display multifocal nerve degeneration, and the typical axonal alteration on teased fibers is myelin wrinkling, while older patients with NIDDM demonstrate multifocal nerve lesions, with Wallerian degeneration on teased fibers (Sima et al. 1988a). In support of this observation, Llewelyn et al. (1988) found no difference in the nonuniformity of fiber loss between a group of relatively young diabetics and in control nerve taken from patients with CMT-1.

Segmental myelin changes usually take the form of inappropriately thinly myelinated axons indicating remyelination or regeneration, with cross sections revealing ongoing demyelination considerably less frequently (Figs. 17.1, 17.5, and 17.6b). Teased fibers demonstrate the spectrum of myelin alteration from paranodal widening to demyelinated segments. In asymptomatic diabetics, segmental myelin alterations may precede loss of axons (Chopra et al. 1969; Vital et al. 1974). Onion-bulb (OB) formations can be seen in diabetic neuropathy but have probably been overemphasized (Fig. 17.7). The OBs are most often rudimentary and do not dominate the histological picture as with CMT-1 or some cases of CIDP (Ballin and Thomas 1968). Indeed, onion-bulb-like structures (pseudo-onion bulbs) are more likely to be a reflection of axonal regeneration (Fig. 17.7a, b). Thomas has commented that in retrospect, the prominent onion-bulb formations described in a case of “diabetic” neuropathy in his 1966 report (Thomas and Lascelles 1966) were likely due to concurrent CIDP (Thomas and Tomlinson 1993). Some authors have reported that the myelin alterations appeared to be independent of axonal changes (Behse and Buchthal 1977), while others have emphasized clustering along certain fibers indicating secondary demyelination (Dyck et al. 1986b).

Hyalinization of endoneurial microvessels is a typical feature of diabetic nerves, regardless of the presence or absence of neuropathy (Fig. 17.1c). This may reach striking proportions, but these microvascular changes remain a nonspecific finding seen in many chronic neuropathies. Alternatively, the severity of endoneurial microvascular alterations has been correlated with the severity of the neuropathy (Malik et al. 1993). Endoneurial area may be increased (Behse and Buchthal 1977), with expansion of the endoneurial and subperineurial extracellular matrix (Ballin and Thomas 1968). Changes in epineurial microvessels are usually mild by comparison (Malik et al. 1993). A low-grade mononuclear inflammatory infiltrate is infrequently seen.

The ultrastructural correlate of vascular hyalinization is reduplication and thickening of microvascular basement membranes (Fig. 17.8a, b). Other alterations which have been emphasized in diabetic neuropathy include endothelial cell hypertrophy and hyperplasia, reduction in capillary luminal area, and microvessel “closure” (Dyck et al. 1985, 1986b; Malik et al. 1989) (Fig. 17.8c, d). However, not all workers have confirmed these findings, and they may be seen in other chronic neuropathies (Bradley et al. 1990) or be related to age (Sima et al. 1988a). Obstruction of capillary lumina with degenerate cellular material and fibrin has been described (Timperley et al. 1985; Yasuda and Dyck 1987), but is uncommon in our experience (Fig. 17.8d, e).

The full range of nonspecific axonal alterations may be seen. One lesion which has received considerable attention is “axoglial dysjunction,” referring to ultrastructural and molecular findings resulting in the disruption of the characteristic junctional complexes joining terminal myelin loops to the axolemma at the paranodal area (Sima 1993; Sima et al. 2000). Swelling of the paranodal axon is often associated with this finding. Axoglial dysjunction seems to be more significant in type 1 (younger) patients than in type 2 (Sima et al. 1988a), but its specificity is uncertain (Sladky et al. 1991). Other differences between types 1 and 2 diabetes in the neuropathologic findings of diabetic somatic nerves have also been described in animals and humans. Type 1 diabetics are reported to develop fewer foci of patchy axonal degeneration and more evidence of myelinopathy than those with type 2 diabetes, in which axonal degeneration is most prominent.

Increased numbers of denervated Schwann cell subunits give early evidence of involvement of unmyelinated fibers, and at later stages of the disease, unmyelinated fiber counts can drop below the lower limit of normal. An apparent increase in rigidity and durability of Schwann cell basal lamina has been observed in diabetics, where after axonal degeneration and regeneration the original basement membrane outline is abnormally maintained (King et al. 1989) (Fig. 17.6a). Perineurial cells may display a thickened basal lamina (Johnson et al. 1981; King et al. 1989).

The metabolic hypothesis began in earnest with the analysis of the role of the polyol pathway in the pathogenesis of diabetic neuropathy (Sima et al. 1988a; Sima 1993). Intraneural concentrations of sorbitol and fructose are increased in both experimental and human diabetic neuropathy, presumably as a consequence of increased metabolic activity along the aldose reductase pathway but do not have an osmotic effect on nerve function. The increase in intracellular sorbitol was reported to influence intracellular myoinositol content and producing a subsequent defect in Na⁺/K⁺-ATPase activity (Sima et al. 1988a). These biochemical changes were thought to cause accumulation of Na⁺ at the nodal axon where sodium channels are found, causing node swelling and detachment of the terminal myelin loops which attach to the axon, so-called axoglial dysjunction. This results in widening of the node and exposure of K⁺ channels which hyperpolarize the axonal membrane and impair conduction. Sima and co-workers have found that aldose reductase inhibitors promote regeneration of peripheral nerve in diabetic neuropathy (Sima et al. 1988b).

Against this theory stands the failure of aldose reductase inhibitors (Tsai and Burnakis 1993) to have a substantial clinical effect in many studies, although controversy continues. Salutary use of aldose reductase inhibitors would reasonably be expected to interrupt the process in its earliest phases, in order to improve diabetic neuropathy.

Recent work reveals that hyperglycemia in diabetes triggers a multitude of biochemical metabolic changes (see Table 17.3) which are currently thought to underlie the development of diabetic neuropathy. Currently, Fernyhough and co-workers (Chowdhury et al. 2010, 2011, 2013; Fernyhough et al. 2003) propose that nutrient excess in neurons mediates mitochondrial damage through alteration of the AMP-activated protein kinase (AMPK)/peroxisome proliferator-activated receptor γ coactivator-1α (PGC-1α) signaling axis. This signaling pathway is thought to affect an energy-sensing metabolic pathway which modulates mitochondrial function, biogenesis, and regeneration. The bioenergetic phenotype (maximal oxygen consumption rate, coupling efficiency, respiratory control ratio, and spare respiratory capacity) of mitochondria in diabetic neurons is aberrant as the result of changes in expression and activity of respiratory chain components brought about as a direct consequence of abnormal AMPK/PGC-1α signaling. Other investigators have confirmed and expanded these ideas, demonstrating decreased PGC-1α and downregulation of its transcription factors including TFAM and NRF1 in mouse streptozotocin-diabetic dorsal root ganglion neurons (Choi et al. 2014). A neuropathic phenotype is exacerbated in PGC-1α (−/−) diabetic mice accompanied by a severe neuropathy with reduced mitochondrial DNA and a further increase in protein oxidation. Conversely, overexpression of PGC-1α in cultured adult mouse neurons prevents oxidative stress associated with increased glucose levels (Choi et al. 2014). The result of these changes in cellular bioenergetics is a reduced adaptability to fluctuations in ATP demand, thought to be distally accentuated. This diabetes-induced maladaptive process is hypothesized to result in exhaustion of the ATP supply in the distal nerve compartment and induction of axonal degeneration, a result which can be corrected by resveratrol treatment (Chowdhury et al. 2012, 2013).

Proponents of the ischemic hypothesis have emphasized the primacy of axonal lesions, their multifocal distribution, alterations in endoneurial and epineurial vessels, and association with disease in other organs (eye, kidney) where microvascular injury is thought to play a central role (Dyck et al. 1986a, b; Dyck 1989 ; Johnson et al. 1986; Sugimura and Dyck 1982; Timperley et al. 1985). The observation that endoneurial microvascular changes differ quantitatively and qualitatively from those seen in other tissues in diabetics mitigates against the suggestion that these simply reflect nonspecific systemic microvasculopathy (Malik et al. 1989, 1993). The distally predominant symmetrical pattern of clinical involvement may result from the summation of numerous random focal insults, with the longest fibers sustaining the greatest number of injuries (Johnson et al. 1986; Waxman et al. 1976). In support of this hypothesis, reduced microvascular luminal area, increased numbers of closed endoneurial capillaries (Dyck et al. 1985; Malik et al. 1993), and plugging of vessel lumina by degenerate cellular material have been described (Timperley et al. 1985; Yasuda and Dyck 1987) and debated. The observation of increased resistance of diabetic peripheral nerve to ischemia (Thomas and Tomlinson 1993) may suggest chronic neural ischemia in these patients.

Correlation of the severity of microvasculopathy and neuropathy has supported an ischemic pathogenesis (Malik et al. 1993; Yasuda and Dyck 1987), but similar microvascular alterations are seen in chronic neuropathies in the absence of ischemia (Bradley et al. 1990). Indeed, endothelial cell hyperplasia was more prominent in CMT-1 than in diabetes and reduced vascular luminal area, and capillary “closure” has not been consistently detected (Bradley et al. 1990), or found to be no different than in age-matched controls (Sima et al. 1991). Moreover, the finding of multifocal nerve fiber loss of equal degree in CMT-1 and diabetic neuropathy weakens the assertion that this pattern suggests an ischemic neuropathy (Bradley et al. 1990; Dyck 1989; Llewelyn et al. 1988). Sima and colleagues (Sima et al. 1988a) found multifocal degeneration of axons in NIDDM (mean age 56) but not IDDM (mean age 38) patients with neuropathy, suggesting that multifocality of nerve lesions in diabetes may relate to age. Geographic regions of fibers with attenuated myelin (Fig. 17.4b) may represent foci of axonal degeneration and subsequent regeneration showing the hypomyelination typical of regenerating fibers or may indicate demyelination with remyelination. Johnson and colleagues (1986) found that the frequency of such focal lesions proximally correlated with the degree of distal nerve fiber depletion in patients with symmetrical sensorimotor neuropathy. The multifocality of the axonal lesions was verified statistically by Dyck et al. (1986b). However, multifocal loss of axons is not pathognomonic of vasculopathy or ischemia and is typical in some forms of inherited neuropathy. Endothelial cell and pericyte degeneration, endothelial cell hyperplasia/hypertrophy, decreased capillary luminal area, and altered blood–nerve barrier permeability are seen in endoneurial microvasculature in patients with diabetic symmetrical sensorimotor neuropathy, findings that correlated with electrophysiological measures of the severity of the neuropathy (Watkins and Thomas 1998; Malik et al. 1993; Giannini and Dyck 1995). Nonetheless, Theriault and colleagues (1997) measured nerve blood flow using laser Doppler flowmetry in diabetic patients with early polyneuropathy and determined that local blood flow was not decreased, even in the presence of definitive evidence of decreased sensory nerve action potential (SNAP) amplitudes and decreased numbers of axons. Finally, axon and Schwann cell nodal and paranodal changes similar to those emphasized in diabetes have been produced by chronic endoneurial ischemia alone (Sladky et al. 1991). Alterations in microvessel junctional structures, similar to “axoglial dysjunction,” have been described (Sima et al. 1991). Increased sural nerve water content due to accumulation of osmotically active solutes may cause a rise in endoneurial pressure and result secondarily in endoneurial ischemia (Griffey et al. 1988; Kalichman and Myers 1991). Nonenzymatic glycosylation of intracellular and extracellular structures may directly cause axon dysfunction or cause cellular and extracellular alterations resulting in endoneurial ischemia (Harati 1992; King et al. 1989).

In summary, the role of microvascular pathology in diabetic neuropathy has been variously interpreted as evidence for an ischemic pathogenesis in diabetic neuropathy or as unrelated but parallel changes (Zochodne 2002); however, there is evidence for tissue hypoxia in the nerve and ganglia in experimental animal models and humans (Zochodne and Ho 1992). Thus, nerve ischemia is associated with nerve dysfunction but may not represent the primary or major cause; rather, it may simply accelerate the process (Llewelyn et al. 2005).

The ischemic and metabolic hypotheses are not mutually exclusive and can be unified by suggestion that metabolic alterations in the neural milieu cause direct or indirect injury to the neural blood supply. An overview of the number of hypotheses proposed to underlie development of neuropathy in diabetes is provided in Table 17.3. It is clear that many of the hypotheses proposed may be interdependent (e.g., oxidative stress, mitochondriopathy, polyol pathway, etc.).