Introduction

Diabetes is a leading cause of distal symmetric polyneuropathy or diabetic peripheral neuropathy (DPN) [1] with important associated health and economic implications [2]. Research into DPN has focused mainly on the peripheral nervous system with central nervous system involvement being overlooked. However, we have reported that spinal cord ‘atrophy’ is present not only in patients with established DPN, but also in those with early (subclinical) DPN [3]. This made us question whether the brain, too, may be involved.

Ascending sensory pathways of the spinal cord terminate within the ventroposterior lateral (VPL) thalamic subnucleus before high-order sensory projections are sent to the cortex [4]. The thalamus plays a crucial role by modulating information presented to the cortex [4]. Hence, in the presence of the sensory nerve dysfunction accompanying DPN, the specific aim of this study was to investigate whether thalamic function is also impaired.

The investigative technique used was proton magnetic resonance spectroscopy (H-MRS). In-vivo H-MRS offers the unique possibility of non-invasive detection of several cerebral metabolites, in particular, N-acetyl aspartate (NAA), a marker of neuronal and axonal integrity [5], creatine/phosphocreatine (hereafter creatine), which is involved in cellular bioenergetics [6], choline-containing compounds, which are involved in membrane synthesis [6] and myo-inositol, thought to be present mostly [7] in glial cells.

We conducted an H-MRS study, using NAA resonance as a neuronal marker to test the hypothesis that DPN coexists with thalamic neuronal involvement. We also characterised the relationship between thalamic neuronal biochemistry and traditional neurophysiological peripheral nerve assessments reflecting DPN severity.

Methods

Participants

We screened 28 right-handed men with type 1 diabetes from the Royal Hallamshire Hospital Diabetes Register. Exclusion factors were: cerebrovascular disease, non-diabetic neuropathies, history of alcohol consumption of more than 20 units a week, diabetic neuropathies other than DPN, painful DPN, internal diseases potentially affecting cerebral metabolism, diabetic ketoacidosis in the preceeding 6 months, hypoglycaemia in the previous 24 h and factors precluding magnetic resonance (MR) imaging. We also recruited six age-matched male non-diabetic healthy volunteers. All participants gave written, informed consent before participation in the study, which had prior approval by the South Sheffield Regional Ethics Committee.

Neuropathy assessment

Detailed neurological assessment was undertaken to identify the presence and quantify the severity of neuropathy. Neuropathic symptoms were documented by completion of the Neuropathy Symptom Score questionnaire and the outcome of a detailed neurological examination was graded using the Neuropathy Impairment Score questionnaires [8]. All participants underwent: (1) vibration detection thresholds acquired from the right foot using computer-assisted sensory evaluation (CASE IV; W. R. Electronics, Stillwater, MN, USA) [9]; (2) measurement of heart rate variability with deep breathing (O’Brien protocol) [10]; and (3) nerve conduction studies performed at a stable skin temperature of 31°C and a room temperature of 24°C, using an electrophysiological system (Medelec; Synergy Oxford Instruments, Oxford, UK).

Based on these clinical and neurophysiological assessments, diabetic participants were divided into two groups: (1) No-DPN consisting of asymptomatic diabetic patients with normal clinical and neurophysiological assessments; (2) DPN, comprising participants with both clinical and neurophysiological abnormalities (at least two abnormalities in neurophysiological assessment) [11]. In addition, a Neuropathy Composite Score (NCS), which was derived from the assessments above (Neuropathy Impairment Score of the Lower Limbs plus seven tests of nerve function), was calculated [12].

Magnetic resonance spectroscopy

All participants underwent H-MRS examination at 1.5 T (Eclipse; Philips Medical Systems, Cleveland, OH, USA). Single-voxel spectra were obtained from an 8 ml cubic volume of interest placed within the right thalamus to encompass the VPL sub-nucleus (Fig. 1a). Two spectra were acquired: (1) long echo time (TE) (TE = 135 ms, resonance time [TR] = 1600 ms) using a point-resolved technique (Fig. 1b); and (2) short TE (TE = 20 ms, TR = 5000 ms) using a stimulated-echo acquisition mode technique with a mixing time of 12 ms (Fig. 1c).

Fig. 1
figure 1

a Axial section of the brain with voxel positioned to encompass the ventroposterior thalamic subnucleus. b Example spectra obtained at long TE using a point-resolved acquisition technique and c at short TE using a stimulated-echo acquisition mode acquisition technique with a mixing time of 12 ms. Cho Choline, Cr creatine, G + sI scyllo-inositol; Glx, glutamine/glutamate; lac, lactate; lip, lipids; mI, myo-inositol; NAA, N-acetyl aspartate

Full size image

MR spectra were anonymously analysed by an experienced clinical MR physicist. All post-acquisition processing was performed using fully integrated proprietary software from the manufacturer (Twinstar workstation, Philips Medical Systems, Cleveland, OH, USA). By convention, long TE results are expressed as ratios under the three prominent resonances assigned to choline (at 3.22 ppm), creatine (at 3.02 ppm) and NAA (at 2.02 ppm), i.e. NAA:creatine, NAA:choline and choline:creatine ratios. Short TE results were calculated as the areas under the myo-inositol (3.56 ppm), choline (3.22 ppm), creatine (3.02 ppm) and NAA (2.02 ppm) resonances relative to that of unsuppressed water.

Statistical method

Statistical analyses were performed using the SPSS 11.1 (SPSS Chicago, IL, USA). Baseline characteristics and study endpoints were described as means and standard deviations for normally distributed variables and as medians and range for variables with a skewed distribution.

The appropriate tests for normality were conducted to guide subsequent analysis. Subgroup H-MRS metabolite endpoints were compared using non-parametric tests (Kruskal–Wallis). Any relationships between these endpoints and neurophysiological assessments were analysed in more detail among participants with diabetes, using Spearman’s rank correlation coefficients.

Results

Clinical characteristics of participants

The mean ages, height and weight of the DPN, No-DPN and healthy volunteer groups were not significantly different (ANOVA, p = 0.63, p = 0.29 and p = 0.24 respectively) (Table l). As expected, participants in the DPN group had a longer duration of diabetes (p = 0.02) and higher HbA1c (p = 0.02).

Table 1 Baseline characteristics and H-MRS study endpoints
Full size table

Group spectroscopy findings

The DPN group had lower mean NAA:creatine ratios at long TE (median [range] 1.55 [1.14−1.95]; p = 0.04) than No-DPN participants (1.79 [1.22−2.20]) and healthy volunteers (1.90 [1.65−2.07]) (Table 1, Fig. 2a). NAA:choline ratio was also lowest in participants with DPN (1.55 [1.37−1.93]; p = 0.02) compared with No-DPN (1.80 [1.60−2.05]) and healthy volunteers (1.76 [1.58−1.86]) (Fig. 2b). The choline:creatine ratio obtained at long TE did not vary significantly between the groups (DPN 1.04 [0.84−1.23], No-DPN 0.95 [0.52−1.26], healthy volunteers 1.06 [0.89−1.25]; p = 0.87).

Fig. 2
figure 2

a N-Acetyl aspartate/creatine (NAA/Cr) and (b) N-acetyl aspartate/choline (NAA/Cho) ratios at long echo time in healthy volunteers (HV), diabetic patients with no neuropathy (No-DPN) and neuropathy (DPN). Horizontal lines represent group medians. Subgroup H-MRS metabolite differences (DPN vs No-DPN and HV) p = 0.04 (a), p = 0.02 (b)

Full size image

Analysis of mean NAA metabolite resonances obtained at short-TE revealed no statistically significant difference between the groups (p = 0.07). No significant differences were observed with any of the other metabolites at short TE between any of the groups.

Spectroscopy and neurophysiological assessments

A significant positive association between both sural amplitude (ρ =0.61, p = 0.004) and nerve conduction velocity (ρ = 0.58, p = 0.006) and long TE NAA:creatine ratio signal was observed among participants with diabetes. In addition, the NAA:creatine ratio signal was significantly associated with peroneal latency (ρ = −0.53, p = 0.01), amplitude (ρ = 0.67, p = 0.001) and velocity (ρ = 0.37, p = 0.06) in this group. Similar correlations among participants with diabetes were noted with tibial nerve latency (ρ = −0.51, p = 0.02), but not with velocity (ρ = 0.23, p = 0.18). In this group, vibration detection threshold was related to NAA:choline ratio at a significant level (ρ = −0.70, p = 0.004), and we also observed a significant correlation between heart rate variability with deep breathing and NAA:creatine ratio (ρ = −0.46, p < 0.05). In addition, the NAA:creatine ratio was also found to relate to overall NCS (ρ = −0.53, p = 0.03) among diabetic participants.

Discussion

The main finding of this preliminary study is the significantly lower long TE thalamic NAA:creatine and NAA:choline ratios in participants with DPN compared with the No-DPN and healthy volunteer groups. The data also demonstrate significant correlations between NAA:creatine ratio and neurophysiological markers (overall NCS and individual nerve function tests) of DPN severity.

NAA is one of the highest concentrations of free amino acids in the brain and is widely used as a spectroscopic neuronal marker of intracranial disease progression. It is believed that a reduction in NAA resonance can be associated with neuronal/axonal loss or loss of neuron viability and dysfunction [13]. In the present study, spectra were acquired at two different TEs (short TE, long TE) from the same voxel. Each spectrum can provide different information regarding NAA. Reduced NAA signal obtained at short TE is thought to reflect decreased neuronal density, which may represent irreversible neuronal loss/shrinkage. Possible reversible neuronal injury/dysfunction can be implicated when the NAA:creatine ratio is reduced at long TE, as this results from changes in the NAA relaxation rates. Just as with standard spin–spin relaxation time (T2)-weighted imaging, variations in T2 imply alterations to the local neurochemical environment of NAA, which in turn affect its relaxation rates [14]. However, unlike reduction in neuronal density, alterations in T2 can be reversible which may account for restoration of metabolite ratio normality concomitantly with improvement in clinical neurological status [15, 16]. We demonstrated a significant difference in NAA:creatine ratio at long TE, but no significant difference in NAA resonance at short TE. In such a scenario, our findings may reflect thalamic neuronal dysfunction in patients with DPN rather than overt neuronal death.

The mechanism of thalamic involvement in DPN is unclear. One explanation may be that loss of afferent input, as a result of peripheral nerve damage, subsequently causes changes to occur at progressively higher levels in the central nervous system (‘dying back’ mechanism). The correlations observed between NAA and neurophysiological markers of neuropathy severity would seem to support this. It is also possible that the observed changes in the thalamus could be occurring concomitantly with changes in the peripheral nervous system. Nonetheless, thalamic neuronal involvement, whether early, late or concomitant, is likely to result in disturbed sensory gating in patients with DPN. This may have consequences for sensory perception and pain modulation. Prospective studies including a subgroup with painful DPN are required to determine at what stage during the course of the disease these abnormalities occur and what impact thalamic dysfunction has on pain perception in patients with diabetes.

Taken together with early spinal cord involvement in DPN [3], the possibility of thalamic sensory neuronal dysfunction suggests that nervous system involvement is not merely confined to the peripheral nerves, but involves the spinal cord and brain. Interestingly, various therapeutic interventions specifically targeted at peripheral nerve damage in DPN patients have thus far been ineffective [17], and it is possible that this may in part be due to inadequate appreciation of the full extent of central nervous system involvement. Recognition that DPN is a disease affecting the whole nervous system should trigger a critical rethinking of this disorder, opening a new direction for further research.