FUNCTIONAL MAGNETIC RESONANCE IMAGING OF THE SPINAL CORD DURING SENSORY STIMULATION IN DIABETIC RATS

Abstract

Purpose

To determine if differences exist between control and diabetic rats in functional MRI activity of the spinal cord and if fMRI can provide a means of early detection of diabetic neuropathy.

Materials and Methods

fMRI of the spinal cord, using noxious electrical stimulation (15 V (~8 mA), 0.3 ms, 3 Hz) of the hind paw, was performed in groups of control and streptozotocin (STZ)-induced type 1 diabetic rats.

Results

Diabetic rats were lighter, hyperglycemic and had lower blood pH than controls. FMRI activity at the lumbar enlargement of the spinal cord was identified in the dorsal horn ipsilateral to stimulus of all animals. Signal intensity changes across the lumbar spinal cord during periods of activity were not significantly different between control and diabetic rats, with a trend towards greater signal changes in controls. When specific regions of the spinal cord were analyzed, control rats exhibited significantly increased BOLD fMRI activity in both ipsilateral and contralateral dorsal horn compared to diabetic rats.

Conclusion

The results of this study are consistent with reports that primary afferent input to the spinal cord is diminished by diabetes, and suggest that BOLD fMRI may be useful in early detection of diabetic neuropathy.

INTRODUCTION

Neuropathy is a major secondary complication of long-term diabetes with damage to sensory, motor and autonomic peripheral nerves. Sensory dysfunction is frequently the earliest manifestation of diabetic neuropathy, with loss of sensation in the extremities being the most commonly reported symptom. Spontaneous or exaggerated stimulus-evoked pain also occurs in a sub-population of patients with diabetic neuropathy. The pathogenesis of diabetic neuropathy downstream of impaired insulin function and hyperglycemia remains largely unknown and there are currently no treatment options that either prevent or attenuate the progression of diabetic neuropathy, other than early institution and maintenance of tight glycemic control. It is therefore particularly important to identify patients who are at risk or in early asymptomatic stages of diabetic neuropathy for current treatment options to be most effective.

Animal models have been widely used to study potential mechanisms underlying diabetic neuropathy and to evaluate potential therapeutics. Insulin-deficient diabetic rats exhibit a number of structural, functional and neurochemical disorders of peripheral sensory neurons including, axonal atrophy and conduction slowing of large myelinated fibers and retraction of fiber terminals and reduced synthesis, axonal transport and release of neuropeptides in small fibers (1). These disorders are viewed as reflecting early stages of the progressive sensory loss and distal degenerative neuropathy seen in most diabetic patients. However, when investigators have applied behavioral sensorimotor tests to diabetic rats the findings have been less consistent with the phenotype of sensory loss and there are many reports that STZ-diabetic rats show allodynia and hyperalgesia rather than hypoalgesia (2). Although this may offer the opportunity to study potential mechanisms associated with diabetes-induced pain and other sensory phenomena, the incongruity of a model that displays a largely structural, electrophysiologic and neurochemical phenotype of degenerative neuropathy with concomitant behavioral phenotype of painful neuropathy remains unresolved. There is some evidence that diabetes induces exaggerated activity of primary afferents (3-5), while other studies suggest that aberrant amplification and processing of sensory information by the spinal cord and higher CNS may also contribute to the behavioral indices of allodynia and hyperalgesia (6-9). It is therefore of interest to measure neuronal activity throughout the neuraxis of diabetic rats to gain insight into potential sites of signal amplification.

Spinal blood oxygenation level-dependent functional magnetic resonance imaging (BOLD fMRI) offers the opportunity to evaluate neuronal activity in living animals by measuring the local hemodynamic response to increased cellular activity (10-15). BOLD fMRI can reliably and reproducibly detect neuronal activity in the spinal cord of normal rats during noxious electrical stimulation of the hindpaw. This activity occurs predominantly over segments L3-L5 of the spinal cord with greatest signal change in the ipsilateral dorsal horn (13, 16). As the electrical stimulus is noxious, the activity observed is believed to derive from information carried by sensory Aδ and C fibres, terminating in laminae I and II of the spinal cord, with the signal representing increased local blood flow and tissue oxygenation in response to cell activity and metabolism (10,11). This technique offers the opportunity to map areas of activity in the spinal cord of live animals and also to quantify the relative change in stimulated activity between groups of animals. We have therefore examined the potential of fMRI to detect early stages of diabetic neuropathy by comparing activity in regions of the lumbar spinal cord of control and diabetic rats during noxious electrical stimulation.

MATERIALS AND METHODS

Adult male Sprague-Dawley rats were made diabetic by a single 75 mg/kg intraperitoneal (IP) injection of STZ (Sigma, St. Louis, MO). Animals were determined to be diabetic after 2 consecutive, non-fasted blood glucose measurements of 19 mmol/L or greater using blood obtained by a distal tail prick, using a standard glucometer with glucose oxidase strip (One touch Ultra, LifeScan, Milpitas, CA). Animals were treated according to the Canadian Council for Animal Care guidelines, and the protocol was approved by local animal care committees. Animals were obtained from the St. Boniface Hospital Research Centre animal care facility.

One month after onset of the study, animals were anesthetized with isoflurane (3-4% induction, 1.5-2% maintenance) in a mixture of 40% oxygen and 60% nitrogen and intubated for mechanical ventilation (Columbus Instruments, Ohio, USA). Catheters (PE 50) were inserted into the left femoral artery so that arterial blood pressure and blood gases could be monitored, and into the left femoral vein for administration of α-chloralose anesthetic and fluids. Bupivacaine (0.25%) was administered to the wound site before suturing closed. After surgery, each animal was placed in the supine position within an animal holder, isoflurane anesthesia discontinued and replaced by α-chloralose (30 mg/mL, 80 mg/kg) administered over approximately 5 minutes with maintenance doses (40 mg/mL) every 90 minutes. Ventilation volume was adjusted in order to maintain normal arterial blood gases (pO₂ of 100-120 mmHg, pCO₂ of 35-45 mmHg) while keeping the ventilation rate constant at approximately 60/minute. Blood gases, pH, blood pressure, heart rate, and rectal temperature were monitored every ten minutes throughout the experiment to ensure a normal physiological range.

For electrical stimulation, two small needle electrodes were placed subcutaneously into the right hind paw (opposite to the side of surgery) on either side of the middle digit. Electrical stimulation (15 V (approximately 8 mA), 0.3 ms duration, 3 Hz) was triggered by the spectrometer to allow current from the stimulator (Grass S48, Mass., USA) to either be delivered to the animal during the active stimulation condition or not be delivered during the resting condition. Prior to imaging, the stimulus was tested briefly to ensure that a toe twitch could be seen. The stimulation presentation followed a block paradigm with 6 images obtained at rest alternated with 6 images acquired during electrical stimulation, and this sequence was repeated for a total of 10 blocks (60 images). At the conclusion of the experiment, animals were euthanized with pentobarbital (120 mg/kg, i.v.).

Images were acquired on a 7 T horizontal bore magnet (Magnex, UK) and a Biospec (Bruker, Germany) console with a Magnex, SGRAD205/120S (12 cm) gradient insert. The animal was placed supine in the animal holder so that the lumbar enlargement was centered over the surface coil. In order to correctly identify the lumbar enlargement, the animal was positioned with the 13^th rib (attached to vertebra T13) located over the centre of the surface coil. The positioning was verified using scout images. The quadrature arrangement of a butterfly and a surface coil were tuned and matched to 300 MHz. Scout images were acquired in axial and sagittal directions to locate the lumbar spinal cord regions of interest. The thickest part of the spinal cord was used to landmark lumbar spinal cord segment L3. Six - 2 mm thick slices were chosen for imaging, corresponding to lumbar spinal cord segments L1 to L6, with a 2 cm FOV and 128 × 64 data matrix, resulting in a voxel size of 0.156 × 0.312 × 2.0 mm. Functional imaging experiments were conducted using a 2-shot fast spin-echo sequence with effective echo time of 54.6 ms, repetition time gated to respiration (approximately 3 seconds), and 60 repetitions, resulting in a total acquisition time of approximately 6 minutes. The experiment was repeated three times for each animal. After the animal was sacrificed, high-resolution T₁-weighted images of the same six slices were acquired (128 × 128 matrix) to overlay the fMRI activations.

For data analysis, a region of interest was selected around the spinal cord and data analyzed by direct correlation on a pixel-by-pixel basis to the paradigm convolved with the hemodynamic response function, using custom-made software in IDL (Interactive Data Language, Research Systems Inc., Boulder, CO). A correlation coefficient of R=0.332 was used to give a p≤0.01. The activity maps generated for each of the three data sets for each animal were then combined and overlaid on the T1-weighted anatomical image using custom-made software developed in MatLab. These images were then manually overlaid to produce a combined activity map for each group.

Signal intensity was graphed in a time course that represents the convolution of the hemodynamic response function and the stimulus paradigm. Time course data were averaged first for each animal then across control and diabetic groups. Statistical analysis was performed using repeated measures analysis of variance (ANOVA) using Statistica software (Statsoft, USA), or unpaired t-test, as appropriate. Differences were considered significant at the p ≤ 0.05 level.

For analysis of signal intensity in specific anatomic regions, cross sectional images were zero filled to 128 ×128 and resized during the analysis program to 160 × 160. The spinal cord was divided into right and left dorsal and ventral regions and central canal using the combined data set for each animal. The number of active pixels was counted in each of these regions in each of the six slices acquired. The number of pixels was then compared between control and diabetic animals in each of the regions of interest using unpaired t-tests.

RESULTS

Diabetic animals were significantly (p<0.001) lighter, hyperglycemic and had lower blood pH than control animals (Table 1). Blood gases (pO₂ and pCO₂) were maintained within normal physiological range throughout the data acquisition (Table 1). One diabetic animal was excluded from the electrical stimulation experiment because of an exaggerated abdominal motion in response to the stimulus and one control animal was excluded due to MR acquisition problems. All animals produced a clear foot twitch when the stimulus was turned on prior to positioning the animal inside the magnet. The current applied to the paw was consistent between control and diabetic animals (Table 1). Resting (baseline) blood pressure (BP) was not significantly different between control and diabetic rats (Table 1). An increase in BP was observed during stimulation periods, with the greatest changes from resting BP observed during the first period of stimulation, followed by either a diminished increase from baseline or no change in BP during subsequent stimulation periods. The increase in mean arterial pressure from baseline during the initial periods of electrical stimulation was not significantly different between diabetic (16 ± 4) and control animals (19 ± 4).

Table 1.

Mean physiological and electrical stimulation parameters obtained during fMRI experiments in diabetic and control rats

Group	Weight (g)†	pO2 (mm Hg)	pCO2 (mm Hg)	Blood pH†	Final Blood Glucose (mmol/l)†	MAP (mm Hg)	Applied Current (mA)
Diabetic	362 ± 59	114.3 ± 16.6	37.7 ± 6.44	7.29 ± 0.12	28.9 ± 3.2	85.0 ± 8.3	7.9 ± 0.5
Control	530 ± 63	114.9 ± 24.9	34.5 ± 6.08	7.48 ± 0.03	7.08 ± 0.86	80.6 ± 13.6	8.1 ± 1.3

Data is shown as group Mean ± SD. The cross illustrates significant differences between groups at

^†p<0.001.

Mean arterial pressure (MAP) is the measure at rest prior to electrical stimulation. Applied current is the calculated amount of current (mA) applied by the electrical stimulator set at 15V for fMRI experiments.

To validate the temporal association between peripheral nerve stimulation and fMRI signal in the spinal cord, total fMRI signal was quantified in 60 consecutive scans of L1-L6 spinal cord for each control and diabetic rat. Images were acquired following the paradigm of 6 × 6 second data collection periods during the resting (non-stimualted) phase followed by a similar period of data collection during the stimulated phase and this sequence repeated for 5 cycles of rest and stimulus. Baseline signal during the resting phases was similar in both groups and an increase in signal above baseline was detected in both groups during all stimulation cycles (Figure 1). Repeated measures analysis of variance across all five stimulation phases for control (n=9) and diabetic (n=9) animals indicated no significant difference between the two groups although there was a trend towards greater signal change in control rats during the first stimulation phase that was absent in the subsequent stimulation phases.

Figure 1.

Averaged time courses of fMRI signal response to noxious electrical stimulation in diabetic (n=9) and control (n=9) rats. Data are group mean (N=9/group) ± SEM. Black horizontal bars represent periods of electrical stimulation with a total duration of approximately 36 seconds. Each data point represents a summation of signal collected over a 6 second imaging period. See text for details of the stimulation and imaging paradigm.

Individual fMRI maps demonstrated activity in the dorsal horn ipsilateral to the site of stimulus of all animals (individual data not shown). Figure 2 shows merged data for control rats; a pixel is labeled as “active” only if a minimum of three animals demonstrate activity at that point. The colour coding demonstrates greater overlap of activity within the group increasing from yellow through red in the colour spectrum. In these merged images, the dominant activity in control rats was located in the ipsilateral dorsal horn at spinal cord segments L3 to L5 (Figure 2a). Activity was also noted contralateral to the stimulus side in segments L2-L5 of control rats (Figure 2a) and in both the ipsilateral and contralateral ventral horn (data not shown). In diabetic rats, notable activity was restricted to the ipsilateral dorsal horn in the L3 segment. The distribution of activity was quantified using pixel counts of the merged images (Table 2). Diabetic rats showed a significant reduction in activity in both the ipsilateral and contralateral dorsal horn when compared to control rats (both p<0.05 by paired t test) and there was also a trend towards reduced activity in the ipsilateral and contralateral ventral horn.

Figure 2.

Combined fMRI activity maps of lumbar segments L1 through L6 in a) control and b) diabetic animals following noxious electrical stimulation of the right hind paw. Pixels are labeled as active only if a minimum of 3 of the 9 animals displayed fMRI activity at a level of p < 0.01 in each individual analysis. The colour coding demonstrates greater overlap of activity within the group increasing from yellow through red in the colour spectrum. The right side of the image is the right side of the animal.

Table 2.

Comparision of number of active voxels of the spinal cord gray matter.

Gray Matter Region of Interest	Control Pixel Count (mean ± SD)	Diabetic Pixel Count (mean ± SD)
Right Dorsal*	218 ± 103	86 ± 60
Right Ventral	124 ± 120	56 ± 39
Left Dorsal*	163 ± 78	61 ± 59
Left Ventral	125 ± 85	66 ± 65
Central Canal	14 ± 16	11 ± 9

The asterisk represents significant difference at a level of p<0.05

DISCUSSION

Whereas the hemodynamic response in fMRI studies of the brain is well established, spinal fMRI is a relatively new technique. Several technical issues make fMRI of the spinal cord particularly challenging, including motion of the body and CSF, vascular activity and magnetic field distortions due to tissue interfaces (17). In the present study we have mitigated these issues by a number of technical modifications. In order to reduce motion due to respiration, image acquisition was gated to the respiratory cycle and confined to the lumbar region, which is more stable than the cervical and thoracic segments because it is farther from the lungs. In an effort to reduce the amount of signal noise derived from the large blood vessels that supply the cord but are located external to the cord itself, we took advantage of the well-demarcated anatomic and functional boundaries of the spinal cord to select regions of interest when performing data analysis. Finally, susceptibility artifacts due to tissue interfaces, particularly bone/tissue interfaces and that can produce field inhomogeneities were obviated by the use of a spin echo sequence.

The temporal pattern of fMRI signal change in rat spinal cord correlated well with alternating periods of rest and electrical stimulation of the hind paw, suggesting that the increased signal during stimulation periods was initiated by paw stimulation rather than other events. The amount of signal change detected in spinal cord of control rats during stimulation periods was also consistent with previous studies that used similar conditions of alpha-chloralose anesthesia and a fast spin echo sequence at 7 T (13-15) or 9.4 T (12). A lower value has been reported in one study performed at 9.4 T (18), possibly due to differences in experimental design, depth of anesthesia or method of data analysis. The greatest signal change was in the dorsal horn ipsilateral to the site of stimulus, which agrees with other BOLD fMRI studies involving noxious electrical stimulation of a hind limb (13, 16, 19) and is consistent with the anatomic and neurochemical localization of central terminals of nociceptive primary afferents and their synapses onto second order neurons. A recent study has demonstrated that selective stimulation of non-noxious A fibers does not evoke detectable fMRI activity (16), implying that the change in signal measured during our stimulation conditions represents that initiated by activity of smaller nociceptive sensory fibers. Increased signal during stimulation was also noted in the ventral horn; this is consistent with previous findings involving fMRI of the spinal cord with hindpaw stimulation (12-15, 18) and is likely due to secondary activation of spinal reflex pathways that result in motor activity. Signal in the contralateral grey matter of the spinal cord suggests activity of interneurons connecting to neurons associated with ascending pathways. Signal located outside the boundaries of the grey matter likely reflect increases in the oxygenated blood content of the dorsal spinal blood vessels and was excluded from analysis.

The signal that achieved the statistical threshold for activity was measured over all voxels in the spinal cross-section and averaged over 5 periods of paw electrical stimulation was similar in control (14.3%) and diabetic (14.1%) rats. However, it may be of interest to note that peak signal declined over the 5 stimulation periods in control rats, with the signal in period 5 being significantly less than that obtained during the first period (p<0.001 by paired t test and representing the mean signal during data collection periods 2-6 for each of stimulation periods 1 and 5). Diabetic rats did not show any progressive change in signal during the 5 periods of electrical stimulation, so that a trend towards a separation in the responses to electrical stimulation during the initial stimulation period was obviated by the declining values of control rats over time. The sequential decline in signal of control rats has been noted in prior studies (18) and may have a physiologic basis arising from adaptations to sustained nociceptive sensory input, such as induction of descending inhibitory systems. Interestingly, recent electrophysiologic studies measuring spinal modulation of sensorimotor reflexes have indicated that induction of GABA_A receptor-mediated inhibitory systems by repeated electrical stimulation is lost in diabetic rats (8). Whether a similar phenomenon is being detected in our present fMRI study remains to be confirmed. It is also interesting to note that percentage signal changes are similar in diabetic and control animals, despite the substantially greater number of active pixels observed in control compared to diabetic rats. This suggests that the hemodynamic and BOLD responses are similar in regions showing fMRI activity in both groups. It is unknown whether there are vascular changes in the spinal cord of diabetic animals at the level of the cord examined in this study. The present results would suggest that differences are more likely to be neurological than vascular.

Secondary data analysis of defined anatomic quadrants of the spinal cord indicated that only 4 weeks of diabetes was sufficient to reduce signal in the ipsilateral and contralateral dorsal horns of the cord. This indicates that fMRI can be used to detect early diabetes-induced changes in the spinal physiologic response to peripheral sensory stimulation and suggest the potential for use in identifying early stages of neuropathy in diabetic subjects. The physiologic or anatomic disorders that promote reduced fMRI signal detected in the spinal cord of diabetic rats are not known. There is no notable loss of sensory neuron cell bodies in the dorsal root ganglia of STZ-diabetic rats at the short durations of diabetes used in the present study (20), and while there are reports of retraction of the peripheral terminals of C fibers from the epidermis, this also requires longer periods of diabetes to develop (21). It is plausible that primary afferents are less responsive to peripheral stimulation or that activity of primary afferents is not transmitted to second order neurons due to diminished neurotransmitter release or post-synaptic responses. No clear consensus has yet emerged regarding changes in expression and function of receptors for excitatory neurotransmitters in the spinal cord during diabetes (22), but the release of excitatory amino acids and neuropeptides in the spinal cord after noxious stimulation of the footpad is reduced in short-term STZ diabetic rats (23, 24) and could contribute to reduced fMRI signal by diminishing activity of post-synaptic neuronal pathways.

While it is tempting to equate the reduced spinal release of excitatory neurotransmitters and subsequent reduced neuronal activity in the spinal cord with diminished fMRI signal change in the dorsal horn of diabetic rats, other possibilities relating to the nature of the signal that is measured by BOLD fMRI must also be considered. As increased BOLD fMRI signal indicates increased tissue oxygen content arising from the hemodynamic response to increased cellular metabolism, disruption of the hemodynamic response by diabetes may contribute to the diminished fMRI signal recorded. It has been suggested that substance P acting as a vasodilator may contribute to the BOLD fMRI signal by increasing local blood flow and volume (16). The reduced synthesis, axonal transport and stimulus-evoked release of substance P and CGRP from primary afferents (23, 25-27) therefore has potential to impair local blood flow as well as excitatory stimulation of spinal neurons and glia. Both substance P and CGRP have vasoactive actions that are compromised by diabetes in the periphery (28) and while this appears to be unrelated to vascular reactivity to substance P and CGRP per se, recent studies have suggested that the neurovasculature of diabetic rats has diminshed endogenous vasodilator capacity (29). Further studies will be required to establish whether a loss of vasodilator function arising from reduced signals and/or capacity to respond occurs in the spinal cord and the extent of any contribution to the reduced fMRI signal. Finally, it should be noted that the analysis paradigm that we applied to individual spinal cord quadrants identified sites that consistently showed increased signal across the group of animals. It is plausible that diabetes induces a migration of central terminals of the stimulated primary afferent population, so that there is dispersion of their terminals throughout the dorsal horn rather than concentration in the superficial laminae, with a consequent apparent loss of focal signal. Diminished isolectin IB4 immunostaining intensity has been reported in the dorsal horn of diabetic mice (30) and although it is not known whether this represents reduced IB4 expression, retraction of the central terminals of non-peptidergic primary afferents that are labeled by IB4 or dispersal of the synaptic region associated with these neurons, the potential for diabetes to disrupt the spinal anatomy of nociceptive processing pathways clearly requires further evaluation.

Although we cannot yet identify the precise nature of the reduced fMRI signal in the spinal cord of diabetic rats, it is particularly notable that our data do not suggest marked general or focal increase in tissue activity, despite the many reports of tactile allodynia and thermal hyperalgesia after paw stimulation in such animals. Both increased peripheral nerve activity and spinal sensitization have been evoked as amplification sites for the apparent allodynic and hyperalgesic behavioral responses to sensory stimuli in diabetic rats (3,6-8) but if either or both were the case, increased fMRI signal would be predicted. The absence of increased spinal fMRI signal upon peripheral stimulation may point towards dysfunction in the higher CNS, as has been suggested by recent imaging studies (9). Nevertheless, while the precise physiologic mechanisms and consequences of the decreased fMRI signal seen in the spinal cord of diabetic rats remain to be clarified, our data indicate the potential for using fMRI as a tool for early detection of diabetes-induced disruption of the peripheral/spinal neuraxis and possibly early stages of diabetic neuropathy.