OUP user menu

Motor pathway injury in patients with periventricular leucomalacia and spastic diplegia

Jong Doo Lee, Hae-Jeong Park, Eun Sook Park, Maeng-Keun Oh, Bumhee Park, Dong-Wook Rha, Sung-Rae Cho, Eung Yeop Kim, Jun Young Park, Chul Hoon Kim, Dong Goo Kim, Chang Il Park
DOI: http://dx.doi.org/10.1093/brain/awr021 1199-1210 First published online: 8 March 2011

Summary

Periventricular leucomalacia has long been investigated as a leading cause of motor and cognitive dysfunction in patients with spastic diplegic cerebral palsy. However, patients with periventricular leucomalacia on conventional magnetic resonance imaging do not always have motor dysfunction and preterm children without neurological abnormalities may have periventricular leucomalacia. In addition, it is uncertain whether descending motor tract or overlying cortical injury is related to motor impairment. To investigate the relationship between motor pathway injury and motor impairment, we conducted voxelwise correlation analysis using tract-based spatial statistics of white matter diffusion anisotropy and voxel-based-morphometry of grey matter injury in patients with periventricular leucomalacia and spastic diplegia (n = 43, mean 12.86 ± 4.79 years, median 12 years). We also evaluated motor cortical and thalamocortical connectivity at resting state in 11 patients using functional magnetic resonance imaging. The functional connectivity results of patients with spastic diplegic cerebral palsy were compared with those of age-matched normal controls. Since γ-aminobutyric acidA receptors play an important role in the remodelling process, we measured neuronal γ-aminobutyric acidA receptor binding potential with dynamic positron emission tomography scans (n = 27) and compared the binding potential map of the patient group with controls (n = 20). In the current study, white matter volume reduction did not show significant correlation with motor dysfunction. Although fractional anisotropy within most of the major white matter tracts were significantly lower than that of age-matched healthy controls (P < 0.05, family wise error corrected), fractional anisotropy mainly within the bilateral corticospinal tracts and posterior body and isthmus of the corpus callosum showed more significant correlation with motor dysfunction (P < 0.03) than thalamocortical pathways (P < 0.05, family-wise error corrected). Cortical volume of the pre- and post-central gyri and the paracentral lobule tended to be negatively correlated with motor function. The motor cortical connectivity was diminished mainly within the bilateral somatosensory cortex, paracentral lobule, cingulate motor area and visual cortex in the patient group. Thalamovisual connectivity was not diminished despite severe optic radiation injury. γ-aminobutyric acidA receptor binding potential was focally increased within the lower extremity homunculus, cingulate cortex, visual cortex and cerebellum in the patient group (P < 0.05, false discovery rate corrected). In conclusion, descending motor tract injury along with overlying cortical volume reduction and reduced functional connectivity appears to be a leading pathophysiological mechanism of motor dysfunction in patients with periventricular leucomalacia. Increased regional γ-aminobutyric acidA receptor binding potential appears to result from a compensatory plasticity response after prenatal brain injury.

  • corticospinal tract
  • functional connectivity
  • motor dysfunction
  • periventricular leucomalacia
  • tract-based spatial statistics

Introduction

Periventricular leucomalacia has long been investigated as a leading cause of motor and cognitive impairment in subjects who were born prematurely with a low birth weight or very low birth weight. The main neuropathological feature is a focal or diffuse necrosis of cerebral white matter with loss of all cellular elements. Although necrosis can be macroscopic in size and readily seen on MRI or ultrasonography as a cystic lesion (cystic periventricular leucomalacia), it is observed in <5% of infants with very low birth weight. More commonly, microscopic necrosis with glial scars is diffusely involved in cerebral white matter (diffuse periventricular leucomalacia). Although cystic periventricular leucomalacia is more correlated with motor impairment, the diffuse type of periventricular leucomalacia accounts for the vast majority of periventricular leucomalacia cases (Volpe et al., 2009), and it is known to be a high risk factor of spastic diplegic cerebral palsy. Conventional brain MRI study can be used in the detection of structural abnormalities, and it provides information on the timing and extent of the hypoxic-ischaemic insult in patients with cerebral palsy (Bax et al., 2006). Previous MRI studies in adolescents born with prematurity and very low birth weight demonstrated high incidence of structural abnormalities that are compatible with periventricular leucomalacia such as dilatation of the occipital horn of lateral ventricles, periventricular white matter volume reduction or thinning of corpus callosum. Among them, ∼10% had spastic diplegia symptoms (Vangberg et al., 2006; Skranes et al., 2007). In contrast, white matter injury cannot be seen in up to 29% of patients with spastic diplegic cerebral palsy on conventional MRI (Bax et al., 2006). A recent study reported that the incidence of cerebral palsy was 24% in moderate periventricular leucomalacia and 67% in severe periventricular leucomalacia cases (Woodward et al., 2006). Therefore, the severity of white matter injury shown on conventional MRI does not appear to be closely correlated with the severity of motor impairment.

Currently, diffusion tensor imaging is widely used for the evaluation of the white matter tract injury in various brain diseases since diffusion tensor imaging is more sensitive than conventional MRI in the detection of microstructural damage. Previous data demonstrated white matter injury on diffusion tensor imaging while conventional MRI failed to demonstrate any structural abnormality in preterm infants (Arzoumanian et al., 2003; Anjari et al., 2007). Nevertheless, until recently there has been a lack of voxelwise cross-correlation study between motor pathway injury and motor dysfunction. In this study, we assessed the correlation, on a voxelwise basis, between fractional anisotropy values of the major white matter tracts and the severity of motor dysfunction using tract-based spatial statistics. Our primary purpose was to investigate whether descending motor tract injury is closely correlated with motor dysfunction in patients with periventricular leucomalacia.

Grey matter injury is also commonly associated with periventricular leucomalacia (Inder et al., 1999, 2005; Pierson et al., 2007). However, it remains to be further elucidated whether regional grey matter injury is involved in the pathogenesis of motor dysfunction in patients with periventricular leucomalacia. Therefore, we also performed correlation analysis between motor dysfunction scale and regional grey matter integrity such as regional cortical volume reduction and alterations in motor cortical or thalamocortical functional connectivity. We evaluated regional γ-aminobutyric acid (GABAA) receptor binding potential in patients with spastic diplegic cerebral palsy and compared the results with those of healthy controls, which could provide valuable information regarding to brain reorganization process after prenatal hypoxic-ischaemic insult since GABAergic neural transmission plays an important role in the repair process of the glutamatergic system (Schwartz-Bloom and Sah, 2001).

Materials and methods

Subjects

In total, 43 children or adults (26 males and 17 females, age range 6–29 years, mean age 12.98 ± 5.26 years, median 12 years) with spastic diplegia symptoms and diffuse periventricular white matter injury on conventional MRI were included. Subjects with structural abnormalities such as a cortical infarction, malformation or any other type of congenital anomalies in the cortical and subcortical grey matter regions on MRI were not included in the current study. All spastic diplegic cerebral palsy subjects were born prematurely (<37 weeks, range 26–34 weeks, mean 30.91 ± 2.10, median 32 weeks) with a low birth weight (n = 32), very low birth weight (n = 9) or extremely low birth weight (n = 2) (range 0.93–2.48 kg, mean 1.69 ± 0.36 kg) but born appropriate for gestational age. Three subjects were born from a twin pregnancy. Twelve subjects had history of perinatal asphyxia, however, grey matter injury such as parasagittal ulegyria or cortical infarction that might produce permanent motor dysfunction was not observed on conventional MRI. Other risk factors for periventricular leucomalacia such as chronic lung diseases, intrauterine infection or prolonged rupture of membranes were not evident in their medical records. Three subjects had eye problems such as strabismus (n = 2) or decreased visual function (n = 1), five subjects had speech problems and three subjects had neurocognitive delay. All patients received orthopaedic surgery, such as heel cord lengthening to reduce spasticity symptoms, 3–58 months prior to this study. The motor dysfunction scale was evaluated using the gross motor function classification system (GMFCS). Thirty-three patients had a mild degree of motor dysfunction (17 subjects with Level 1 and 16 subjects with Level 2), and 10 patients had moderate-to-severe dysfunction (Level 3 in six, Level 4 in three and Level 5 in one subject).

Conventional magnetic resonance imaging acquisition

For structural imaging, we obtained high-resolution T1-weighted MRI volume data using a Philips 3T scanner (Intera Achieva, Philips Medical System, Best, The Netherlands) with a SENSE head coil. The scanning parameters of the 3D T1-TFE sequence were an axial acquisition with a 224 × 256 matrix, field of view = 220 mm, voxel size = 0.98 × 0.98 × 1.2 mm3, echo time = 4.6 ms, repetition time = 9.6 ms, flip angle = 8° and slice gap = 0 mm. The T2-weighted images were acquired axially with a 400 × 319 matrix, field of view = 230 mm, voxel size = 0.45 × 0.45 × 5 mm3, echo time = 80 ms, repetition time = 3000 ms, flip angle = 90° and slice gap = 2 mm. The T2-weighted and fluid attenuation inversion recovery images were acquired axially with a 352 × 238 matrix, field of view = 230 mm, voxel size = 0.45 × 0.45 × 5 mm3, echo time = 125 ms, repetition time = 11 000 ms and slice gap = 2 mm.

Diffusion tensor image acquisition

We obtained diffusion tensor images using a single-shot echo-planar acquisition with the following parameters: 112 × 112 acquisition and a 128 × 128 reconstructed matrix, field of view = 220 mm, voxel size = 1.72 × 1.72 × 2 mm3, SENSE factor 2, echo time = 70 ms; shortest repetition time = ∼13 000 ms, flip angle = 90°, slice gap = 0 mm, two averages per slice, b-factor = 600 s/mm2, non-cardiac gating and ∼70 axial slices. We acquired diffusion-weighted images from 45 non-collinear, non-coplanar directions with a baseline image without diffusion weighting. The spatial distortions induced by eddy currents and motion artefacts in diffusion-weighted images were corrected by registering the diffusion-weighted images to the non-diffusion-weighted b0 image using a low order non-linear transformation (Kim et al., 2006).

Resting state functional magnetic resonance imaging acquisition

Among the 43 patients, we performed resting state functional MRI in 13 subjects and obtained data from 11 (mean age of 10.17 ± 2.69 years); two subjects were excluded due to motion artefacts. Functional MRI data of age-matched normal controls (n = 11, mean age of 12.36 ± 4.13 years) without neurological and structural abnormalities; risk factors for periventricular leucomalacia were used for group comparison. We acquired 165 axial volume scans from each subject using a Embedded Image single shot echo planar imaging sequence using a 3T Philips MRI scanner with the following parameters: voxel size = 2.75 × 2.75 × 4.8 mm3, slice number = 29 (interleaved), matrix = 80 × 80, slice thickness = 4.8 mm, repetition time = 2000 ms, echo time = 30 ms, field of view = 209 × 220 mm2. Foam pads were used to reduce head motion during echo planar imaging data acquisition. During each scan, subjects were instructed to keep their eyes closed, rest without movement and attempt to sleep for ∼5 min.

Tract-based spatial statistics of white matter fractional anisotropy

We performed statistical analysis of the fractional anisotropy values with a tract-based spatial statistics technique using FMRIB’s Diffusion Toolbox implemented in the FMRIB’s software library (http://www.fmrib.ox.ac.uk/fsl/tbss/index.html) as described previously (Smith et al., 2006; Anjari et al., 2007). A mean fractional anisotropy skeleton with a threshold fractional anisotropy >0.2 from all patients was generated to include the major white matter tracts. Fractional anisotropy data of age-matched normal controls (n = 43) without neurological and structural abnormalities or any risk factors for periventricular leucomalacia were derived from our normal database. The MRI acquisition methods used for normal controls were not different from those of patients. Fractional anisotropy differences between the patient and control group were evaluated with a permutation-based randomized test (Nichols et al., 2002) and inference using threshold-free cluster enhancement (Smith et al., 2009) implemented in FMRIB’s software library to correct multiple comparison problems.

To investigate white matter injury sites that are associated with motor dysfunction, we performed voxelwise correlation analysis between fractional anisotropy values within the major white matter tracts and GMFCS levels using threshold-free cluster enhancement and results were obtained after 10 000 random permutations (nuisance parameters: age and sex). The voxels with significant correlation (P < 0.05, family-wise error corrected) were projected onto the mean fractional anisotropy skeleton on tract-based spatial statistics.

Assessment of white and grey matter volume reduction

The white matter volume was calculated after extraction of grey matter volume using Freesurfer software v.4.0.5 (http://surfer.nmr.mgh.harvard.edu). The white matter volume adjusted for intracranial volume (white matter volume/intracranial whole brain volume) was compared with GMFCS level after age correction.

We performed voxel-based morphometry in patient group and age- and gender-matched healthy controls (n = 43) using DARTEL (Ashburner, 2007) implemented in the Statistical Parametric Mapping 8 software (http://fil.ion.ucl.ac.uk/spm, Wellcome Department of Cognitive Neurology, London, UK). We generated a group grey matter template to which the entire individual grey matter segmentation map was spatially normalized. To adjust the volume changes during the non-linear transformations, spatially normalized grey matter segmentation maps were modulated by the amount of expansion or shrinkage during the spatial normalization process by applying the Jacobian determinant of the deformation field (Good et al., 2001).

These modulated grey matter maps were smoothed using a 6 mm full-width half-maximum isotropic Gaussian kernel. The regional volume differences between spastic diplegic cerebral palsy subjects and healthy controls were determined using t-statistics for every voxel in the grey matter and correlation analysis between grey matter volumetric density of each voxel and GMFCS level was performed in patients with spastic diplegic cerebral palsy using Statistical Parametric Mapping 8.

Resting state functional connectivity study

In consideration of the magnetization equilibrium, the first 10 images from the functional MRI data were excluded. The remaining 155 echo planar images were corrected for the acquisition time delay between different slices. We adjusted for head motions by realigning 155 consecutive volumes to the first image of the session. The realigned images were coregistered to T1-weighted images that were used to spatially normalize functional data to a template space using non-linear transformation. The normalized data were then spatially smoothed using an 8 mm full-width half-maximum Gaussian filter.

Functional connectivity maps between motor cortex or thalami and whole brain grey matter were calculated using a correlation approach. The time series of each of the motor cortex and thalami were calculated by averaging the activity of the bilateral Brodmann area (BA) 4 for motor cortex, which was defined by the cytoarchitectonic map (Geyer et al., 1996), and by averaging the activity of the bilateral thalami in automated anatomical labelling (Tzourio-Mazoyer et al., 2002). Time series of all voxels within the grey matter were quadratic detrended and band-pass filtered (0.01–0.08 Hz) (Weissenbacher et al., 2009). For each subject, the correlation coefficients between time series of seed regions and those of each voxels within the grey matter were transformed to a z-value using Fisher r–z transformation (Weissenbacher et al., 2009). Statistical analysis of the motor cortical or thalamic connectivity map was performed by one sample t-test. The connectivity map between patients with spastic diplegic cerebral palsy and normal control groups was also compared using two-sample t-test with Statistical Parametric Mapping 8.

18F-fluoroflumazenil positron emission tomography imaging and statistical analysis of binding potential using statistical parametric mapping

18F-fluoroflumazenil used in this study is a newly developed radiotracer that binds to central benzodiazepine receptors. The Bmax, Kd and IC50 values of this radiotracer are comparable to those of 3H-flumazenil (Mitterhauser et al., 2004; Chang et al., 2005; Lee et al., 2007).

Twenty-seven patients (14 males and 13 females, mean age 13.89 ± 4.79 years, median 14 years) and 20 normal volunteers (10 males and 10 females, mean age 21.20 ± 1.32 years, median 21 years) completed a dynamic PET scan without motion artefact using a DSTe PET scanner (GE, Milwaukee, WI, USA) to measure the GABAA receptor binding potential. Among the 27 patients with spastic diplegic cerebral palsy, eight subjects (mean age 22.8 ± 4.09 years) were age-matched to normal controls. A transmission CT scan was performed for attenuation correction and dynamic emission data were obtained in a sequence of 150 frames in 3D mode (60 × 10 s, 40 × 15 s, 20 × 30 s, 30 × 60 s) for a total acquisition time of 60 min after injection of ∼5.5 MBq (0.15 mCi)/kg of 18F-fluoroflumazenil. The attenuation corrected images were reconstructed in a 128 × 128 × 35 matrix with a voxel size of 1.95 × 1.95 × 4.25 mm using the Hanning and Ramp filters.

To calculate the GABAA receptor binding potential, we used a multi-linear reference tissue method (Millet et al., 2002) that was implemented in PMOD software (PMOD Technologies Ltd., Zurich, Switzerland). The mean activity within a manually delineated the region of interest at the pons was used for reference tissue activity because the density of GABAA receptors in the pons is negligible.

Spatial pre-processing and statistical analysis of the 18F-fluoroflumazenil PET data were performed using Statistical Parametric Mapping 8 for the group comparison of the binding potential map. The GABAA receptor binding potential data of the patient group (n = 27) and the age-matched subgroup of patients with spastic diplegic cerebral palsy (n = 8) were compared with those of normal control group after scaling the binding potential map images proportionally by the global whole brain GABAA receptor binding potential. Proportional scaling was performed to minimize the individual variation of the global binding potential resulted from wide range of age distribution in patient group. This study was conducted in accordance with institutional guidelines and was approved by the institutional review board (No. 4-2006-0311). We obtained informed consent from all subjects and normal controls.

Results

White matter injury in patients with spastic diplegic cerebral palsy

Conventional MRI showed various degrees of ventricular dilatation and decreased white matter volume, mainly within the posterior thalamic radiation, posterior part of the corona radiata and thinning of corpus callosum in all patients with spastic diplegic cerebral palsy (Fig. 1A). However, increased signal intensity representing white matter injury extended almost to the entire bilateral periventricular white matter regions including the anterior portion of the corona radiata on fluid attenuated inversion recovery images (Fig. 1B and C). Group comparison of fractional anisotropy values between patients with spastic diplegic cerebral palsy and healthy controls using tract-based spatial statistics revealed lower fractional anisotropy values within almost all white matter tracts in the patient group, although white matter tracts within the posterior part of the brain were more extensively involved than anterior white matter tracts (Fig. 1D, P < 0.05 family-wise error corrected).

Figure 1

Periventricular white matter injury in patients with periventricular leucomalacia on conventional MRI and tract-based spatial statistics. (A–C) Transaxial fluid attenuated inversion recovery images show dilatation of the lateral ventricles with angular deformity of the lateral margins, approximation of the grey matter and sulci to the ventricular margins secondary to white matter volume loss (curved arrows) and thinning of the corpus callosum (red arrow heads). These are typical MRI findings of diffuse type periventricular leucomalacia. Increased signal intensity, representing white matter injury, can be seen within the posterior limb of the internal capsule (black arrows) and also within the periventricular white matter regions. Periventricular white matter injury is present not only within the posterior corona radiata but also within the anterior corona radiata (black arrow heads). (D) Group comparison of fractional anisotropy values between patients with periventricular leucomalacia and age-matched healthy controls using tract-based spatial statistics shows decreased fractional anisotropy values in almost all major white matter tracts. Regions of the mean fractional anisotropy skeleton in green represent undamaged tracts, while regions in red represent damaged white matter tracts in patient group (P < 0.05, family-wise error corrected). ALIC = anterior limb of the internal capsule; CbP = cerebellar peduncle; CC = corpus callosum; EC = external capsule; PLIC = posterior limb of the internal capsule; SLF = superior longutudinal fascicle.

Correlation of the severity of white matter injury with GMFCS levels

White matter volume reduction in each hemisphere did not show significant correlation with motor dysfunction (P > 0.1). Fractional anisotropy values within the posterior white matter tracts and corpus callosum showed a negative correlation with GMFCS levels (P < 0.05 family-wise error corrected). However, fractional anisotropy values within the bilateral corticospinal tracts and posterior body of the corpus callosum were more strongly correlated with motor dysfunction (P < 0.03 family-wise error corrected) (Fig. 2).

Figure 2

Voxel-wise correlation analysis between fractional anisotropy values and GMFCS levels on tract-based spatial statistics. Green lines represent a mean fractional anisotropy skeleton of undamaged white matter tracts. Areas in blue represent regions showing significant negative correlation between fractional anisotropy values and GMFCS levels (P < 0.05, family-wise error corrected). Fractional anisotropy values of most of the white matter tracts within the posterior brain areas show a significant negative correlation with GMFCS levels (A–C). In particular, fractional anisotropy values within the bilateral corticospinal tracts (C–F) and the posterior body of the corpus callosum (E) also show significant correlation with motor dysfunction (red clusters, P < 0.03, family-wise error corrected). CC = corpus callosum; CST = corticospinal tract; SLF = superior longitudinal fascicle.

Grey matter injury on voxel-based morphometry

Grey matter volume of the bilateral sensorimotor cortex, parietal, temporal and occipital lobes, posterior cingulate cortex, basal ganglia, thalami and cerebellum was significantly reduced in the patient group compared with the age- and gender-matched normal controls (Fig. 3A, P < 0.005 false discovery rate corrected). In particular, cortical volume of the pre- and post-central gyri, and the paracentral lobule tended to be negatively correlated with GMFCS levels at the threshold level of uncorrected P < 0.005 (q < 0.22 after false discovery rate control) (Fig. 3B).

Figure 3

Grey matter volume reduction in patients with periventricular leucomalacia and correlation with motor dysfunction. (A) Voxel-based morphometry. Grey matter volume is reduced mainly within the posterior part of the cerebral cortex including the sensorimotor cortex, basal ganglia and thalami in patients with periventricular leucomalacia (P < 0.005, false discovery rate corrected, cluster size ≥100). (B) Grey matter volume reduction versus motor dysfunction. Grey matter volume within the pre- and post-central gyri and paracentral lobule is negatively correlated with GMFCS levels (uncorrected P < 0.005 or q < 0.22 after false discovery rate control, cluster size ≥100). CS = central sulcus; PCL = paracentral lobule.

Motor cortical and thalamic connectivity

The motor cortex in normal control subjects had extensive functional connections to the adjacent somatosensory cortex, visual cortex and cerebellum on one-sample t-test. In the patient group, motor cortical connections appeared to have connections more expanded to the adjacent parietal area (family-wise error corrected P < 0.05). However, group comparison study showed decreased connectivity within the bilateral somatosensory cortex, paracentral lobule, presupplementary motor area, cingulate motor area, visual cortex, superior and inferior parietal lobules in the patient group at the threshold level of uncorrected P < 0.005 (q < 0.06 after false discovery rate control) (Fig. 4A).

Figure 4

Functional connectivity of the motor cortex and thalami. (A) Motor connectivity. The patient group shows more expanded motor cortical connections to the adjacent parietal area on one-sample t-test (arrows, P < 0.05, family-wise error corrected, cluster size ≥100). However, group comparison revealed decreased sensorimotor and visuomotor connections in patient group (uncorrected P < 0.005, or q < 0.06 after false discovery rate control). (B) Thalamic connectivity to the parieto-occipital area, cingulate cortex and frontal area appears to be more expanded in patient group based on one-sample t-test (P < 0.05, family-wise error corrected, cluster size ≥100); however, connectivity is decreased within the cingulate cortex and the caudate nucleus on group comparison (uncorrected P < 0.005, or q < 0.12 after false discovery rate control). ACC = anterior cingulate cortex; CMA = cingulate motor area; PCC = posterior cingulate cortex; PCL = paracentral lobule; PoCG = post-central gyrus; VC = visual cortex; SPL = superior parietal lobule.

Thalami are connected to the basal ganglia, cerebellum and overlying cortex especially to the cingulate cortex in normal controls. In the patient group, parietal, occipital, prefrontal and posterior cingulate connectivity appeared to be more expanded on one-sample t-test (P < 0.05, family-wise error corrected). A group comparison study showed that thalamic connectivity to the caudate nucleus, anterior and posterior cingulate cortex, and cerebellum appeared to be diminished in patient group at the threshold level of uncorrected P < 0.005 (q < 0.11 after false discovery rate control) (Fig. 4B). Thalamo-visual connectivity was not affected despite severe optic radiation damage.

Statistical analysis of GABAA receptor binding potential

GABAA receptor binding potential was focally increased within the bilateral lower extremity homunculus including the paracentral lobule, cingulate cortex, visual cortex and cerebellum despite significant grey matter volume reduction, whereas receptor binding potential was diffusely decreased in the prefrontal, temporal, parietal and subcortical nuclei in patient group (Fig. 5, P < 0.05, false discovery rate corrected) (Table 1). Group comparison study between the age-matched spastic diplegic cerebral palsy subgroup (n = 8) and normal control group confirmed increased receptor binding pattern within these regions as seen in the total patient group (Fig. 6).

Figure 5

GABAA receptor binding pattern in patients with periventricular leucomalacia and spastic diplegic cerebral palsy compared with healthy controls. (A) The GABAA receptor binding potential is focally increased within the paracentral lobule, lower extremity honunculus, cingulate cortex, visual cortex and cerebellum in patient group (red colours, P < 0.05, false discovery rate corrected). (B) The GABAA receptor binding potential is diffusely decreased within the prefrontal, parietal, temporal and subcortical nuclei (blue colour, P < 0.05, false discovery rate corrected). Arrow heads and arrows denote central sulcus. ACC = anterior cingulate cortex; CS = central sulcus; Cbll = cerebellum; PCC = posterior cingulate cortex; PCL = paracentral lobule; VC = visual cortex.

Figure 6

GABAA receptor binding pattern in age-matched patient subgroup (n = 8) compared with normal control group. The receptor binding pattern is almost the same as the binding pattern of the total patient group (uncorrected P < 0.05). CMA = cingulate motor area; PCL = paracentral lobule; VC = visual cortex.

View this table:
Table 1

Increased GABAA receptor binding potential sites in spastic diplegic cerebral palsy (P < 0.05, false discovery rate corrected, cluster ≥100)

SideRegionBrodmann areaMNI (mm)Zmax
xyz
RightAnterior cingulateBA 24103263.70
Posterior cingulateBA 2910−42125.07
Cingulate gyrusBA 2412−16363.52
Lingual gyrusBA 1826−8645.11
CuneusBA 1926−82165.42
Fusiform gyrusBA 2036−72−44.85
Paracentral lobuleBA 612−28624.78
Precentral gyrusBA 412−28625.38
InsulaBA 1344−42205.18
Cerebellar tonsil*18−60−423.88
Declive of vermis*2−70−203.39
LeftAnterior cingulateBA 24−228104.40
Posterior cingulateBA 29−6−4644.96
Cingulate gyrusBA 31−16−20423.32
Lingual gyrusBA 18−14−9063.84
CuneusBA 19−18−82143.84
Fusiform gyrusBA 20−22−88−124.02
Medial frontal gyrusBA 6−14−24563.76
Precentral gyrusBA 4−26−20483.26
InsulaBA 13−40−44203.23
Superior temporal gyrusBA 41−42−4264.01
Cerebellar tonsil*−24−52−463.75
  • * = not applicable.

Discussion

The diffuse form of periventricular leucomalacia is known to be a leading aetiological factor of cerebral palsy, especially in infants with very low birth weight or those born prematurely (Bax et al., 2006; Platt et al., 2007). The underlying pathophysiological mechanisms of diffuse periventricular leucomalacia include hypoxic-ischaemic insult with excessive glutamatergic excitotoxicity and free radical injury to the immature oligodendrocytes or oligodendrocyte precursor cells. Maternal-fetal infections (Rezaie and Dean, 2002; Back et al., 2007) or chorioamnionitis are independent risk factors for cerebral palsy as well as leading causes of periventricular leucomalacia (Wu et al., 2003).

The severity of white matter injury as seen on conventional MRI is considered to be closely correlated with motor dysfunction in patients with periventricular leucomalacia. However, this relationship is still debatable since patients with periventricular leucomalacia do not always have neurological abnormalities, and structural abnormalities may not be seen on conventional MRI in patients with spastic diplegic cerebral palsy (Olsen et al., 1997). In addition, it is still controversial whether corticospinal tract injury is related to motor dysfunction in patients with periventricular leucomalacia and spastic diplegic cerebral palsy. Previous reports demonstrated a significant correlation of corticospinal tract injury with neurological impairments (Kidokoro et al., 2008; Murakami et al., 2008), whereas Hoon et al. (2009) reported that motor dysfunction is more related to thalamocortical injury with abnormal sensorimotor connections rather than corticospinal tract injury itself. Therefore, pathophysiological mechanisms involved in motor impairment in patients with periventricular leucomalacia need to be further elucidated. However, these controversial results might have been resulted from limitations of the analytic method used; conventional MRI technique is not sensitive to detect subtle white matter injury and it is also difficult to identify individual white matter tract injury precisely, in particular, corticospinal tract injury. Although diffusion tensor imaging with tractography enables correct localization of individual tract injury, most of the previous results were derived from region of interest based analysis method, which is less sensitive than voxelwise analysis.

Recently, tract-based spatial statistics is being widely used for voxelwise group comparison of white matter diffusion properties for correlation analysis between white matter micro-structural abnormalities on diffusion tensor imaging and various functional abilities (Counsell et al., 2008). We compared group difference of fractional anisotropy values between patients with spastic diplegic cerebral palsy and controls using tract-based spatial statistics, and further analysed correlation between motor function scales and fractional anisotropy values at every voxel along the whole white matter tract trajectories. In the current study, fractional anisotropy values within most of the white matter tracts were significantly lower in patients with periventricular leucomalacia than those of age-matched normal controls (P < 0.05, family-wise error corrected). Although the severity of white matter injury within most of the posterior white matter tracts and corpus callosum showed a significant negative correlation with motor function (P < 0.05, family-wise error corrected), bilateral corticospinal tract injury and corpus callosal injury (posterior body) were more significantly correlated with motor dysfunction (P < 0.03, family-wise error corrected) (Fig. 2). These diffusion tensor imaging results clearly demonstrate that focal injury within the corticospinal tract is closely related to motor dysfunction in patients with periventricular leucomalacia. Negative correlation between GMFCS level and the severity of white matter injury within the posterior body and isthmus of the corpus callosum is also congruent with previous diffusion tensor imaging data showing close correlation with neurodevelopmental abilities (Counsell et al., 2008). In fact, these callosal regions are pathways for interconnecting the bilateral motor cortex (Park et al., 2008).

Grey matter injury such as neuronal degeneration, gliosis (Inder et al., 1999, 2005; Northington et al., 2005; Pierson et al., 2007; Stone et al., 2008) and subplate neuronal damage (Leviton and Gressens, 2007) is known to be commonly associated with periventricular leucomalacia following prenatal hypoxic-ischaemic insult. These neuronal injuries produce regional volume reduction and the geographic pattern of neuronal damage is close to the predilection sites of white matter injury, predominantly involving the posterior brain region (Leviton and Gressens, 2007) as shown on our voxel-based morphometry and tract-based spatial statistics studies. Although grey matter volume can be reduced within the predilection site of periventricular leucomalacia even in adolescents born very prematurely without motor disabilities (Nosarti et al., 2008), we found that cortical volume of the specific regions, especially somatosensory and motor cortices, and paracentral lobule appeared to be negatively correlated with GMFCS levels (Fig. 3). Therefore, our correlation analysis between grey matter volume reduction and GMFCS levels along with the current fractional anisotropy correlation analysis data support the fact that motor pathway injury could be a leading pathophysiological mechanism of motor dysfunction in patients with periventricular leucomalacia since fibres of the corticospinal tract arise from the precentral and postcentral gyri including the paracentral lobule, premotor and parietal areas.

In addition to the morphological abnormalities, alteration of functional connectivity within the affected cortical areas could be an important determinant for final motor function. In a previous animal study, increased cortico-cortical connections within the visual cortex were observed after deafferentation of thalamocortical inputs (Kingsbury et al., 2002), and a human study demonstrated expanded sensorimotor cortical connections in patients with spastic diplegic cerebral palsy (Burton et al., 2009). Despite local expansion of the cortico-cortical connectivity, our patient group showed a diminished motor cortical connectivity especially within the bilateral paracentral lobule and cingulate motor area as compared with the age-matched control group. Sensory-motor and visuomotor connections, thalamic connections to the caudate nuclei and cingulate cortex were also diminished in the patient group. Based on these results, diminished motor cortical connectivity within the motor control areas, along with focal injury in the corticospinal tracts could be a relevant pathophysiological mechanism producing motor dysfunction. Although thalamic connectivity to the anterior cingulate cortex is diminished, it might be more related to impaired cognitive function in patients with spastic diplegic cerebral palsy. Further studies of the correlation between cognitive function scales and anterior cingulate connectivity are required.

In terms of alterations in functional connectivity in patients with spastic diplegic cerebral palsy, it could be resulted from subplate neuronal injury following hypoxic-ischaemic insult with subsequent repairing process (Kingsbury et al., 2002; Burton et al., 2009) since subplate neurons are highly vulnerable to hypoxic-ischaemic insult (McQuillen and Ferriero, 2005; Leviton and Gressens, 2007). Under normal circumstances, subplate neurons receive inputs from the thalamus and send excitatory glutamatergic projection to cortical layer IV before cortical neurons directly receive thalamocortical inputs (Pinon et al., 2009; Kanold and Luhmann, 2010). When subplate neurons are damaged before maturation of the thalamocortical circuit, these projections bypass the injured areas, but project to the adjacent areas containing intact subplate neurons (Ghosh et al., 1990). In addition, descending cortical connections to the internal capsule, corpus callosum and deep subcortical nuclei also can be injured after subplate neuronal damage (De Carlos and O’Leary, 1992; McConnell et al., 1994).

After subplate neuronal injury, expression of the genes or proteins involved in neurotransmitter regulation, neuronal survival or axon extension is subsequently altered (Lein et al., 1999; Osheroff et al., 2009). Among the various neurotrophins and neurotransmitters, we focused on GABAA receptor expression because inhibitory GABAA receptor plays a substantial role in brain remodelling processes after injury (Schwartz-Bloom and Sah, 2001). Usually, regional GABAA receptors within the ischaemic regions are transiently down-regulated or internalized into the cytoplasm (Schwartz-Bloom and Sah, 2001). Reduction of the inhibitory GABAA receptor expression is necessary to induce or maintain the excitability of perilesional and functionally connected neurons and to expand receptive field size for an optimal practice-dependent plasticity (Pascual-Leone et al., 2005; Benali et al., 2008; Cramer, 2008). In fact, long-term decreased uptake of 11C-flumazenil within the damaged areas after ischaemia has been reported previously (Yamauchi et al., 2007).

A post-mortem human study also revealed significant loss of GABAergic neurons and GABAA receptor density within the prefrontal cortex in infants with periventricular leucomalacia (Robinson et al., 2006). In the current study, the patients with spastic diplegic cerebral palsy group showed decreased GABAA receptor binding potential within the prefrontal, parietal and temporal areas as previous reported. In contrast, receptor binding potential was focally increased within the lower extremity homunculus, cingulate cortex, visual cortex and cerebellum despite significant grey matter volume reduction (Fig. 5). Our previous report using delayed PET images also demonstrated a significant higher uptake of 18F-fluoroflumazenil within the bilateral motor and visual cortices in patients with spastic diplegic cerebral palsy (Lee et al., 2007). However, focally increased GABAA receptor binding potential, especially within the motor control areas, is contradictory to the classical GABAergic response to ischaemic insult, and also contradictory to radiolabelled flumazenil uptake pattern within ischaemic areas because the uptake of radiolabelled central benzodiazepine receptor agents within damaged neurons is decreased (Yamauchi et al., 2007). Our contradictory results could be explained by the fact that immature forms of the GABAA receptors consisting of α2 and α3 subunits with a depolarizing activity are upregulated after prenatal ischaemic injury until adulthood, whereas mature form α1 subunits of GABAA receptors are down-regulated (Paysan et al., 1997; Kanold and Shatz, 2006; Kanold and Luhmann, 2010). Therefore, we hypothesized that GABAA receptors were upregulated in patients with spastic diplegic cerebral palsy as an adaptive neuronal plasticity process in compensation for reduced GABAergic neuronal input and diminished functional connectivity due to white matter injury or overlying grey matter volume reduction or both. Further studies regarding to the subunit composition of the GABAA receptors and their function are required.

Limitations of the study

This study has several limitations. First, resting state functional MRI data were obtained from only 11 patients and majority of our patients with spastic diplegic cerebral palsy had lower GMFCS levels, which might result in low statistical power for group comparison of motor and thalamic connectivity between normal controls and patients with spastic diplegic cerebral palsy (uncorrected P < 0.005). However, the threshold of uncorrected P < 0.005 is estimated to be q < 0.06 for motor and q < 0.11 for thalamic connectivity, respectively, after false discovery rate control. Therefore, statistical significance of our group comparison data in functional connectivity study can be acceptable since q-value in the range 0.1–0.2 after false discovery rate control is known to be acceptable for multiple comparison (Genovese et al., 2002). Future works with a large series of patient group would produce better results with higher statistical power.

Second, our patients with spastic diplegic cerebral palsy group had a wide range of age distribution. However, it would not significantly affect on our results since cerebral palsy is a permanent and non-progressive disorder (O’Shea, 2008). Nevertheless, further studies with a narrow age range patient group are needed.

Another limitation may be diffusion tensor imaging data acquisition errors due to motion artefact. In this study, we obtained diffusion-weighted images from 45 directions to increase the tensor estimation reliability of the tensor estimation (Jones et al., 2004). Although this acquisition method might produce motion artefacts, we used a low order non-linear transformation between diffusion-weighted images and the b0 image to reduce motion artefacts as well as eddy current effects (Kim et al., 2006). Therefore, statistical results in the current study might not be significantly affected by tensor estimation errors.

Conclusion

Our data demonstrate that descending motor pathways injury along with overlying cortical volume reduction and decreased functional connectivity could be a leading pathophysiological mechanism for motor dysfunction in patients with periventricular leucomalacia. The severity of optic radiation injury was less correlated with motor impairment than corticospinal tract injury. Increased GABAA receptor binding potential, especially within the decreased connectivity areas could be related to compensatory plasticity process after brain injury.

Funding

Korea Healthcare Technology R&D Project, Ministry for Health, Welfare and Family Affairs, Republic of Korea (A080446).

Acknowledgements

We thank Mr Hoon Hee Park, Han Sang Lim, Dong Wook Rho and Hyuk Namgung for their excellent technical assistance in data acquisition and processing; Drs Jung Young Kim and Tae Hyun Choi for the production of 18F-fluoroflumazenil. The production techniques and the quality control of 18F-fluoroflumazenil were supported by the Korea Institute of Radiological and Medical Sciences project.

Footnotes

  • *These authors contributed equally to this work.

Abbreviations
BA 
 Brodmann area
GABA 
 γ-aminobutyric acid
GMFCS 
 gross motor function classification system

References

View Abstract