Most research focused on hyperglycaemia and hyperlipidaemia as main players to induce oxidative stress and pro-inflammatory mechanisms. Diabetes-related hyperglycaemia as well as hyperlipidaemia induce a number of pathological changes in neuronal tissue leading to oxidative stress and pro-inflammatory mechanisms, as shown in Figure 1[14,15]. Hyperglycaemia leads to elevated intracellular glucose and cellular toxicity. This glucotoxicity alters cell function in different ways leading to increased synthesis of polyols, diacylglycerol (which in turn activates protein kinase C) and hexosamines that accumulate intracellularly[16]. The exact mechanism by which these factors cause altered cell function is not yet clear, but they act in concert to induce oxidative stress[5]. Thus, levels of free radicals, such as superoxide and nitrogen species rise, especially in the mitochondria. Meanwhile, the ability to scavenge free radicals is reduced because of a depletion of the proton donor nicotinamide-adenine-dinucleotide[14].

Open in New Tab   Full Size Figure   Download Figure

Figure 1 Hyperglycaemia leads to increased hexosamines, polyols and diacylglycerol within the cell, which can cause oxidative stress leading to cell damage.

These processes may also trigger an enzyme, poly(ADP-ribose) polymerase, of great importance to deoxyribonucleic acid repair, and thereby cause break-up of the deoxyribonucleic acid strands. A further consequence is the reduction of intracellular nicotinamide-adenine-dinucleotide, exacerbated by the polyol pathway induction. As an end result, adenosine triphosphate levels have been shown to reach critical low levels in, e.g., Schwann cells, possibly resulting in cell death[17]. Finally, a key enzyme in glycolysis (glyceraldehyde-3-phosphate dehydrogenase) is inhibited when the superoxide levels rise, which cause a reduction in the substrate flux into the mitochondria. However, glucose still enters the cell, which causes even more activity in the alternative metabolic routes, leading to further production of hexosamines, polyols, poly(ADP-ribose) polymerase and advanced glycation end products, thus closing the loop of a vicious cycle[18]. For further details see[17-19].

The hyperglycaemia theory seems more valid for type 1 than type 2 DM. In line with this a Cochrane review showed that improved glucose control inhibits onset of neuropathy in type 1 DM, whereas it only had a modest, non-significant relative risk reduction in patients with type 2 DM after 4 years of follow-up. However, when patients were followed for 15-year the effect of increased glucose control showed significant risk reduction[20-22]. Although peripheral neuropathy was explored, the mechanisms are likely similar for other nerve tissues. Hence, the prevention of hyperglycemia is likely also of importance to protect the autonomic nervous system and the brain.

The altered function of the cell can also be caused by other factors. The intracellular non-enzymatic glycation of proteins gives rise to advanced glycation end-products, which in the extracellular matrix interacts with various receptors from matrix to endothelial receptors. This last mentioned interaction can lead to proinflammatory gene expression[19].

Severe and prolonged hypoglycaemia can increase release of excitatory amino acids. The release may turn uncontrolled and trigger calcium influx, thereby leading to activation of proteolytic enzymes, causing neuronal damage[23]. However, the brain may utilise other non-glucose resources such as ketone bodies and amino acids and hence be protected against the hypoglycaemic changes[7,24]. This is in contrast to the brain damage caused by ischemia and hypoxia. On the other hand, hypoglycaemia and the counter-regulatory hormonal responses are associated with an acute rise in haematocrit levels and blood viscosity, and this may influence capillary blood flow especially when structural changes of the vessels and metabolic pathways are already present[7].

Insulin receptors are found throughout the brain. The systemic insulin level is increased in most type-1 DM patients due to exogenous insulin treatment. In healthy people insulin is produced in the pancreas, released into the portal circulation and passed on to the liver, where it exerts its prime metabolic effects. When patients with diabetes are treated with exogenous insulin it is absorbed directly into the systemic circulation. This results in 200% increased insulin level in the blood, depending on the injected insulin dose[7,25]. Insulin is thought to modulate glucose utilisation in specific brain areas, such as hippocampus with a central role for memory function, and in this way it may affect cognitive functions. Furthermore, brain insulin plays a role in satiety signalling and possible neurodegenerative disorders such as Alzheimer disease[26].

Diabetes is associated with both functional and structural alterations of the cerebral vascular system, which can for example increase the risk of stroke[7]. Early changes DM is the reduced neuronal blood flow in the vasa nervorum, which can cause neurophysiological changes, such as endotheliopathy[19]. Pathogenically, reduced availability of vasodilating molecules, in particular nitric oxide due to its binding to superoxide forming peoxynitrite, play a central role[16]. Secondly, vasoconstrictive factors such as sympathetic tone and angiotensin II levels are increased. Thirdly, increased arterio-venous shunting reduces endoneurial blood flow[27]. Later on, structural changes such as pericyte degeneration, capillary membrane thickening and decreased capillary density may occur in the brain[28].

Studies regarding cerebrovascular reactivity in type-1 diabetes patients show that the normal increase in blood flow after administration of dilatory stimuli is impaired in DM patients. The impairment is more severe in patients with diabetes with longer disease duration or who have other complications[29]. Cerebral vasoreactivity and accompanying changes in blood flow are important in the preservation of adequate perfusion during abnormal events such as hypotension, hypoxia, hypercapnia and hypoglycaemia, all of which are prevalent in diabetes patients. Loss of these important regulatory mechanism may therefore have detrimental effects on the brain[7].

Besides the above described mechanisms, additional factors, such as disruption of normal endoplasmatic reticulum functioning and the role of autoantibodies may play a role, but will not be further explained here[27,30,31]. Finally, it is important to realize that the relative contribution of different factors varies between individuals (depending on characteristics such as sex, age, co-morbidity and stage of disease) and that all these factors interact[32].

Changes within the brain in DM have been investigated in both animals and humans. Although not the focus of this review, a few selected animal studies are shown in Table 1. In animals such as rats with long-term streptozotocin-induced diabetes increased abnormality in the neuron cells and blood vessels within the brain have been found. Findings were altered Golgi bodies, mitochondria and endoplasmic reticulum cisterns, concurrent with superoxide dismutase inactivation and aldose reductase accumulation[41]. Besides this, changes in dorsal motor and paraventricular nuclei of the hypothalamus have been reported in animal models, both of great importance for controlling the part of the autonomic nerves that innervate the gut[9]. Additionally, protein kinase C was found increased, which may cause altered cell functioning as explained in section (The hyperglycaemia hypothesis)[42]. However, it is not yet clear to which extend the GI mechanisms are caused by the central neuropathy. Some changes in the gut function may be explained by neurotoxicity of streptozotocin at specific locations in the brain, which could have great impact on the regulation of the gut. When having this in mind the choice of animal model is critical if wrong conclusions are to be avoided and relevance to human disease to be maintained[9]. Animal models can inform us of some of the pathophysiological and pathogenic aspects of human disease. An advantage is that they may compare to human cases even though the animal does not necessarily have the same symptoms as in a human condition. On the other hand, major differences in morphology and function between rodent and human brains do exist, making the results of animal studies difficult to translate to humans.

As outlined above, the differences between human and animal brains make it important to focus on human research. Here, methods such as quantitative sensory testing (QST), electrophysiology or imaging can be used to explore functional and structural brain changes in humans with diabetes and GI symptoms, which will be elaborated further on in the next sections.

Clinical assessment of the sensory system in patients is affected by, e.g., general malaise, additional pathophysiological influences, or co-medications. QST is a discipline to evoke sensations such as pain under controlled circumstances[43]. This is advantageous as it encompasses many of the problems seen in the clinical situation and offers the opportunity to explore the sensory system more objectively. QST can for example determine the vibration sensation threshold and the pain thresholds for cold and warm temperatures by stimulating the skin, but can also be used in the deeper tissues such as muscles and viscera. In many of the studies investigated in this review, the oesophagus has been investigated. The methods are also able to selectively activate different nerve pathways to mimic the clinical situation such as with the multimodal probe that can be used in the oesophagus and rectum[44]. Following standardised sensory stimulation there are several ways to measure the response. Psychophysical methods such as intensity ratings are most frequently used, but they give no detailed information about the brains response that can be explored with electrophysiological or imaging methods. Findings from experimental pain can be used to getting greater knowledge of a disease and pharmacological mechanisms, and clinical pain symptoms may help explaining results from experimental pain studies.

QST has been used in several studies to investigate diabetes patients with GI symptoms, as seen in Table 2. The most consistent result to painful gut stimulation in DM was hyposensitivity. Two studies found that DM patients had hyposensitivity to painful sensation evoked in the esophagus[11,45]. When using rectosigmoid electrostimulation hyposensitivity to painful stimulation was also found[4]. However in another study the authors found no significant differences in sensory or pain threshold to esophageal stimulations between DM patients and healthy controls, and findings are therefore not absolutely consistent[46]. In a study where the effect of acute hyperglycaemia was tested, increased gut sensitivity was seen. Hyperglycaemia is thought to affect the nerves, however it had no effect on the sensation in the patients[47]. The hyposensitivity seen in DM strongly supports the presence of peripheral neuropathy and it seems to be generalized to several gut segments. Hence, in comparison with healthy controls, Søfteland et al[48] found evidence of decreased cutaneous and rectal sensations in diabetes patients with sensorimotor neuropathy and increased GI symptoms. Rectal and cutaneous sensitivities were correlated and associated with abnormal heart rate variability supporting that all fibre types were affected. Another QST study in patients with longstanding DM, showed evidence of generalized neuronal damage manifested as sensorimotor, autonomic and central neuropathies, and the degree of peripheral hypoesthesia was associated to both heart rate variability and impaired conditioned pain modulation[49]. Pain and other conscious sensations are processed in the brain and traditionally a “bottom-up” model has been suggested; i.e., damage to the peripheral nerve causes central reorganization[4]. On the other hand, it cannot be excluded that spinal, brainstem or brain changes due to central neuropathy alone may contribute to the generalized sensory changes.

QST can also be used to evoke referred pain. This is partly due to convergence between visceral and somatic afferents in the dorsal horn of the spinal cord, and any central hyperexcitability will increase the size of the referred pain area[50-53]. As increases in the referred pain areas were seen following stimulation of the upper gut in DM, this indicates a widely distribution of central sensitization.

EEG is the recording of electrical activity on the scalp produced by activation of neurons in the brain. The activity can basically be recorded as either evoked potentials following an external stimulus, or in the resting state[54]. In both types of recordings, the EEG can be used to study normal pain processing and to identify alterations of pain processing in different patient groups[54]. Most studies used evoked brain potentials (EPs), the concept illustrated in Figure 3. The advantage of EPs is that they can detect neuronal activity with very high temporal resolution, thereby making analysis of the primary sensory-specific upstream activation of brain centres possible. This is of major importance as these activations take place within the first 200 ms after stimulation of the periphery[54]. Compared with methods based on hemodynamic and metabolic changes (positron emission tomografi and functional MRI) EPs have a better time resolution (ms) and with the newest models the corresponding brain sources can be modelled with a spatial resolution of a few mm[54,55].

Open in New Tab   Full Size Figure   Download Figure

Figure 3 Evoked potentials are recorded following a peripheral stimulus as indicated with the grey “lightning” and ideally a corresponding activity can be seen following 80-90 ms as an evoked potential. However, as illustrated in (2) the amplitudes of the evoked potential tend to be low and often comparable to the amplitudes of spontaneous electroencephalogram. In order to decode the evoked potentials from the background electroencephalographic activity and noise, signal averaging is necessary as illustrated in (3). Provided enough of recorded trials, the evoked potentials become bigger in amplitude and therefore visible and the random background activity cancel out. Then, the EP latencies and amplitudes of the peaks can be analyzed by visual inspection. When many (64-128) electrodes are used the corresponding brain sources can be computed based on the surface electroencephalographic recordings. EP: Evoked brain potential.

Electrical stimulation is often used to activate the nervous system but it is unspecific as it bypasses peripheral receptors and depolarizes all types of nerve fibres. In the gut electrical stimulation primarily activates Aδ fibers although the majority are unmyelinated C-fibres, but the electrical stimulus is still to be favored due to the high temporal accuracy. Recently methods such as rapid balloon distension, which is a more natural stimulus, have been developed, but they have never been used in patients with diabetes[56]. Furthermore, EPs have been recorded following non-visceral sources such as auditory, visual and somatosensory stimuli, and abnormal response in the brain has been observed in patients with type 1 and 2 diabetes[7]. There has been focus on the so-called late responses as reflected in the P300 wave of somatosensory stimuli, where the latency was increased in DM patients, and it has been hypothesized that evoked potential changes may appear before cognitive dysfunction develops[57]. The resting EEG has also been used to investigate the brain in patients with diabetes. This has revealed differences in brain connectivity and information flow, but the changes were not related to the evoked potentials or cognitive functions, and hence assess other functions of the brain[57,58]. More sophisticated analyses have confirmed that resting EEG synchronization and complexity is also related to cognitive function and blood glucose level[59,60]. However, resting EEG has not been used to explore neuropathy or GI changes and hence it will not be elaborated further in this review. Recently the neurophysiological changes using EPs to visceral organs in patients with GI symptoms have been investigated and these will be presented below.

Studies with EPs have shown evidence of altered central processing to visceral stimulation in diabetes patients with GI symptoms and the findings were repeated in several tudies[61,62]. Table 2 shows an overview of the different studies. Using evoked potentials to stimulation of recto-sigmoid and esophagus the latencies of the EPs at vertex were increased and amplitudes reduced in DM patients with major variability between patients[62]. In recent studies where multiple channels (64-128) were recorded the authors were able to model the corresponding brain sources[2]. Compared with controls, the patients had a posterior shift of the electrical sources in the anterior cingulate cortex to oesophageal stimulation, and additional sources close to the medial frontal gyrus and posterior insula[62]. Another study conducted with a similar method found that the GI symptoms correlated with characteristics of brain potentials in the diabetes patients[63]. Furthermore a study conducted by the same group showed that DM patients had an anterior shift in insular and cingulate source localizations compared to healthy controls[64]. As the insula is considered one of the main centres from where the upstream activation of visceral information is controlled, this may have major importance for our understanding of GI symptoms in DM. In fact the research showed that the more GI symptoms (vomiting, nausea, early satiety, diarrhoea, abdominal pain and/or constipation) the patients experienced, the more anteriorly the insular source was located[64].

GI symptoms and their development and maintenance were also found to correlate with reorganization within the opercular cortex localised between the insula and secondary somatosensory cortex[61]. Another study explored the communication between different brain regions. The changes in networks were correlated to severity of upper GI symptoms and life quality. It was concluded that changes in the networks could also serve as a biomarker of disturbed sensory visceral processing and GI symptoms in patients with diabetes[46].

Another method is brain imaging where magnetic resonance (MR) is the dominating method. The contrast between grey and white matter in MRI makes it the optimal choice for many conditions of the CNS including demyelination. A specific way of analysing MR images is volumetry where the total brain volume is determined by summing the grey and white matter pixels, then multiplying by the voxel dimension. With this method, a more precise estimate of atrophy can be accessed[65]. Another MRI method used in some of the studies in this review is functional magnetic resonance (fMRI) with measurement of blood oxygen level dependent contrasts (BOLD). It measures brain activity by detecting changes in blood oxygenation. It infers regional changes in brain activity, and thereby it reflects whether a specific brain region is engaged to a time-linked neurobehavioral or neurocognitive task. It has been of great importance in mapping regions in the brain linked to specific functions[66]. More specifically it is a great method to examine the superficial layers of the brain because of the excellent spatial resolution (2-5 mm), but limitations are seen in the deeper structures, such as brainstem and thalamus because of pulsation artefacts. fMRI has the possibility to take individual characteristics and anatomy into account, which can be a major advantage. Furthermore, fMRI operates in a non-radioactive and non-invasive radioactive environment, allowing subjects to be studied repetitively[54]. Though there are many advantages of fMRI there are some limitations, for example the fMRI is clearly inferior in temporal resolution compared to EEG/MEG, meaning that in pain studies fMRI is not a specific tool for investigating the primary neuronal activity directly related to the painful stimuli. Additionally, the fMRI activity to GI stimulation is not stimulus-specific. Hence, anticipation of stimuli can trigger similar activity and repeated activation can result in habituation[54]. Another imaging method is diffusion tensor imaging (DTI)[67]. It measures the directionality and magnitude of water diffusion in tissues. The mechanism behind neurostructural changes is not clear but it is the general belief that the integrity of axonal membrane and myelin sheaths is reflected by restriction of diffusion perpendicular to the fibres. The intra-axonal structures, such as microtubules, are thought reflected by diffusion parallel to the fibres. DTI measures the magnitude [described by the apparent diffusion coefficient (ADC)] and directionality [described as fractional anisotropy (FA)] of water diffusion in tissues. Reduced ADC values are seen in variety of CNS insults such as stroke and trauma, whereas reduced FA is often seen in schizophrenia, Alzheimer’s disease and depression[68].

The last method relevant for this review is the arterial spin labelling, which allows the measurement of whole brain blood flow in absolute units through the use of magnetically labelled endogenous water in blood acting as a diffusible tracer[54]. This method makes is possible to tract the temporal dynamics of the neural activation induced by pain. It is a suitable method to measure the brains response to tonic stimuli and symptoms since it is more sensitive than fMRI to changes in neural activation when stimulus duration exceeds 1 min[54].

Brain MRI conducted in diabetics has been used in different contexts. Most studies have been done in non-selected patients - i.e., without GI symptoms. In type 2 DM atrophy of the brain is mostly found in areas responsible for verbal, visual memory, executive functioning and information processing, and this may link to an increased risk for developing dementia[40]. There has also been found cortical atrophy in type-2 diabetes patients, resembling patterns in preclinical Alzheimer’s disease[69]. Cerebral dysfunction has been convincingly shown in patients with type 1 and 2 diabetes, but few neuroradiological studies have been conducted in patients with autonomic and GI symptoms[70]. Due to the sparse imaging research with focus on GI problems and diabetes this review have included few studies regarding general brain changes, as well as studies where brain changes, were comparable to those in animal studies with a correlate to GI symptoms. Generally, as shown in Table 3 structural changes with central atrophy has been reported in patients with DM. One study on type 2 diabetes patients found cortical and subcortical atrophy involving frontal and temporal brain regions, with diminished vasoreactivity and regional cerebral perfusion[71]. This supports that uncontrolled diabetes may further contribute to hypoperfusion and atrophy. More specifically MR has also been used to discover a larger lateral ventricular volume with larger white matter lesion. However, no white matter volume difference was found, which is opposite the grey matter shrinkage in the DM patients[72]. In another study with MR imaging the focus was atrophy and aging, and in midlife diabetes was associated with subcortical infarctions. More specifically a reduced hippocampal volume, whole brain volume atrophy, and mild cognitive impairment were found[73]. The reduced hippocampal volume has been confirmed in different studies including animal studies[9,69,73,74]. On the other hand a study that looked into the macrostructural brain alterations found no overall alterations with standard evaluations of the images, and it may be that the macrostructural brain changes are limited in well-treated type-1 diabetes patients[75]. It has also been suggested that the radiological appearance of the brain in patients with diabetes resembles that of normal ageing, but appears to develop at a younger age than in healthy controls[7]. In one study investigating the effect of diabetes on brain atrophy and cognitive impairment, pathological findings were only significant in women. The differences between the genders were an unexpected finding and need to be investigated in more detail[72].

Changes in cerebral blood flow in type 2 diabetes have also been investigated with arterial spin labelling[71]. Type 2 diabetes were associated with cortical and subcortical atrophy involving frontal and temporal brain regions and with diminished vasoreactivity and regional cerebral perfusion. Additionally the same study found that uncontrolled diabetes might further contribute to hypoperfusion and atrophy.

In patients with DM and GI symptoms a study used cortical volumetry and found reduced cortical thickness of the postcentral and superior parietal gyrus in patients. Those with peripheral neuropathy showed reduction in right postcentral gyrus cortical thickness compared to patients without neuropathy[74]. DTI has also been used to investigate more subtle changes in the brain. More specifically a study found that patients with long-standing DM and GI symptoms have microstructural changes in brain areas involved in visceral sensory processing. This could be related to DM-induced brain changes specific for the gut, although, e.g., insular changes may also be important in dysregulation of other functions[68]. The microstructural changes were for some areas correlated to GI parameters such as bloating and presence of gastroparesis, together with other autonomic dysfunctions and therefore may be involved in the pathogenesis of GI symptoms. However, even though the few studies in MR are consistent and microstructural alterations were found in diabetes patients with GI symptoms, they still need confirmation in other studies.