OUP user menu

Changes in dopamine availability in the nigrostriatal and mesocortical dopaminergic systems by gait in Parkinson's disease

Yasuomi Ouchi, Toshihiko Kanno, Hiroyuki Okada, Etsuji Yoshikawa, Masami Futatsubashi, Shuji Nobezawa, Tatsuo Torizuka, Keisei Tanaka
DOI: http://dx.doi.org/10.1093/brain/124.4.784 784-792 First published online: 1 April 2001

Summary

The basal ganglia play a role in controlling movement during gait. The aim of the present study was to investigate changes in dopamine transporter (DAT) availability in the striatum and extrastriatal region in association with walking exercise in six normal subjects and seven age-matched unmedicated patients with Parkinson's disease. This was done by comparing DAT radioligand uptake in the dopaminergic projection areas after gait with that under the resting condition using a DAT probe, 11C-labelled 2-β-carbomethoxy-3β-(4-fluorophenyl) tropane ([11C]CFT) and PET. Physiological parameters were stable during and after gait in both groups. The regions of interest method for measuring differences in [11C]CFT uptake level and voxel-based statistical parametric mapping (SPM96) showed that [11C]CFT uptake in the striatum (specifically the putamen) was decreased by gait to a greater extent in normal subjects, whereas a significant reduction in [11C]CFT uptake was not found in the putamen but in the caudate and orbitofrontal cortex in Parkinson's disease patients. These results are the first in vivo evidence that DAT availability is reduced in the nigrostriatal projection area by basic human behaviour, i.e. gait. Alterations in this availability in Parkinson's disease suggested that shifted activation in the medial striatum and the mesocortical dopaminergic system might reflect the pathophysiology of parkinsonian gait.

  • gait
  • dopamine availability
  • Parkinson's disease
  • PET
  • statistical parametric mapping
  • ABP = arterial blood pressure
  • AC–PC = anterior–posterior intercommissural line
  • ANCOVA = analysis of covariance
  • ANOVA = analysis of variance
  • BA = Brodmann area
  • [11C]CFT = 11C-labelled 2-β-carbomethoxy-3β-(4-fluorophenyl) tropane
  • DAT = dopamine transporter
  • FWHM = full width at half maximum
  • PR = pulse rate
  • RI = ratio index
  • ROI = region of interest
  • SPM = statistical parametric mapping
  • UPDRS = Unified Parkinson's Disease Rating Scale

Introduction

It has been reported that the basal ganglia modulate motor movement during gait (Marsden, 1982; Mink and Thach, 1991). In particular, striatal dopaminergic neurones are activated by behavioural manipulations during the playing of video games (Koepp et al., 1998). This behaviour-induced dopamine release was first reported in humans using the postsynaptic D2 receptor antagonist [11C]raclopride. In rodents, locomotive movement significantly increases extracellular dopamine level in the striatum as measured by microdialysis (Heyes et al., 1988; Hattori et al., 1994). Dopamine released at the synapses is transported for reutilization through presynaptically located dopamine transporter (DAT), which plays an important role in regulation of extracellular levels of dopamine (Jones et al., 1999). It has been reported that a high dose of extracellular dopamine (l-dopa) or significantly elevated endogenous dopamine levels reduces the binding of [3H]WIN35,428 [equivalent to 2-β-carbomethoxy-3β-(4-fluorophenyl) tropane, or CFT] in mice (Scheffel et al., 1996); a reduction in DAT tracer availability by strenuous motor behaviour should, therefore, indicate a high level of dopamine release in the dopaminergic projection areas as shown in rodents.

Mapping of neural substrates for bipedal gait in the human brain has been examined using PET with [18F]2-fluoro-2-deoxy-D-glucose ([18F]FDG) (Ishii et al., 1993) and single photon emission tomography (SPECT) with a perfusion tracer (Fukuyama et al., 1997; Hanakawa et al., 1999), which illustrated broader neuronal activation chiefly in the `basal ganglia–frontal cortex' motor loop regions. However, these perfusion-based imaging studies on gait control in the human brain precluded dopaminergic information in the basal ganglia, which is regarded as a motor control modulator. Therefore, we explored the basal ganglia with and without loss of dopamine input using dopaminergic tracer to illustrate altered activation of dopamine neurones by gait in humans. We expected such studies to shed light on the in vivo contribution of the dopaminergic nigrostriatal and mesocortical systems to execution of gait in humans.

In the present study, we performed PET with a DAT probe, 11C-labelled 2-β-carbomethoxy-3β-(4-fluorophenyl) tropane ([11C]CFT) to investigate changes in [11C]CFT availability after 1 h of strenuous walking in healthy subjects and unmedicated Parkinson's disease patients with hypokinetic gait. Irreversibly bound tracers such as [18F]FDG and [11C]CFT might be suitable to use for activation studies demanding longer periods of time to accomplish tasks, because accumulation of [11C]CFT in the specific binding region increases with time (Ouchi et al., 1999b), resembling the pattern of tissue [18F]FDG uptake (Greenberg et al., 1981; Ginsberg et al., 1987).

Methods

Subjects and patients

Six healthy subjects (five males, one female; mean age ± SD = 65.3 ± 5.9 years) and seven age-matched unmedicated Parkinson's disease patients (five males, two females; 66.3 ± 6.6 years) at Hoehn & Yahr stage 2–3 gave their written informed consent to participation in the present study approved by the local ethics committee of the Hamamatsu Medical Center. Clinical assessment of each Parkinson's disease patient was performed with the Unified Parkinson's Disease Rating Scale (UPDRS) (Fahn et al., 1987) and the Mini-Mental State Examination (Table 1). MRI revealed no brain morphological abnormalities in any of the subjects. l-Dopa treatment after PET examination was effective for parkinsonian symptoms in all Parkinson's disease patients.

View this table:
Table 1

Details of Parkinson's disease patient characteristics

PatientsAge (years)SexDDH&YMMSEUPDRS (me/a/mo)
DD = disease duration (years); H&Y = modified Hoehn & Yahr disability score (1–5); MMSE = Mini-Mental State Examination (max = 30); UPDRS = Unified Parkinson's Disease Rating Scale (me = mentation, behaviour and mood, a = activities of daily living, mo = motor examination).
168M1.522711/19/38
274M232611/24/53
371M1.522810/17/34
465M0.9229 8/14/24
564F1.5227 7/16/28
656F1.8227 8/16/36
759M2.1328 7/18/42

MRI and PET procedures

MRI with 3-dimensional mode data sampling was performed just before PET measurement using a static magnet (0.3 T MRP7000AD; Hitachi, Tokyo, Japan) to determine areas of the striatal nuclei and the orbitofrontal cortex to set regions of interest (ROIs). In reference to measurements of tilt angle and spatial coordinates obtained in the procedure for determining the anterior–posterior intercommissural line (AC–PC) on each subject's sagittal MR images, a PET gantry was set parallel to the AC–PC line by tilting and moving the gantry for each study, which permitted reconstruction of PET images parallel to the AC–PC line without reslicing (Ouchi et al., 1998).

Each participant underwent two PET measurements on the same day. Serial scans (time frames: 4 × 30 s, 20 × 60 s, 14 × 300 s) lasting 92 min were first performed under the resting (supine) condition. Three hours after the first PET measurement, the participant was instructed to walk for 50 min just after tracer injection and the second PET scan (60–92 min postinjection) was performed in the supine position after a 50 min walk. At the first measurement, arterial blood samples were collected periodically for 92 min immediately after a slow bolus injection of 450-MBq of [11C]CFT to measure [11C]CFT binding potential quantitatively (Ouchi et al., 1999b). At the second scan, each subject received 300 MBq of [11C]CFT. These two PET scans parallel to the AC–PC line were performed using a gantry-mobile high-resolution PET camera (SHR2400, Hamamatsu Photonics KK, Hamamatsu, Japan) with 2.7 mm in-plane and 5.5 mm axial spatial resolution (Yamashita et al., 1990). A thermoplastic face mask designed for radiosurgery was used for head fixation to the same place during the two PET measurements.

Task performance

Participants walked at their own pace along a white line on the corridor back and forth with the cuff of a sphygmomanometer around the left arm, which measured time course of changes in arterial blood pressure (ABP) and pulse rates (PR) during walking. Normal subjects were, however, requested to walk more slowly than usual. Cadence (steps/min) was calculated on the straight part of the corridor, but behaviour during turning was not assessed. All Parkinson's disease patients, with mild to moderate hypokinetic gait, were able to accomplish a 50 min walk without rest or assistance. Each participant was instructed to look at the white lines as much as possible while walking.

Data analysis

ROI method

Irregular ROIs were drawn bilaterally over the caudate, putamen, orbitofrontal cortex and cerebellum on the MR images (Ouchi et al., 1999b). The border of the orbital frontal cortex was outlined in the prefrontal region from the rectus gyrus to the bottom of the genu of the corpus callosum (Mai et al., 1997; Buchanan et al., 1998). The borders for the caudate and putamen were delineated on all planes on which they appeared (Backman et al., 1997). Tissue time activity curves were derived for the caudate, putamen and orbitofrontal cortex in PET measurement under resting condition. With these tissue time activity curves and metabolite-corrected plasma tissue time activity curves, binding potential (equivalent to k3/k4; Mintun et al., 1984) of [11C]CFT for each region was estimated quantitatively based on the 3-compartment 4-parameter model (Ouchi et al., 1999b; Wong et al., 1993). Simultaneously, tissue [11C]CFT uptake relative to the cerebellar uptake (ratio index, RI) was calculated using the following formula: (Countregion – Countcerebellum)/Countcerebellum, using late integrated [11C]CFT data of both rest and gait PET images (Wong et al., 1993). Regression analysis was performed to investigate whether RI values could reflect estimated k3/k4 levels. A positive correlation between these two parameters would allow us to compare RI values on the late integrated images (Frost et al., 1993), and to generate RI-based parametric images for voxel-by-voxel comparisons [see statistical parametric mapping (SPM) analysis]. Differences in RI levels at rest and gait were tested with repeated ANOVA, with respect to one inter-subject factor (Parkinson's disease, Normal) and intra-subject factors (orbitofrontal cortex, caudate and putamen). Spearman rank correlation analysis was performed to examine the relation between RI changes and magnitudes of cadence during gait in each dopaminergic projection area in Parkinson's disease group.

SPM analysis

The parametric RI images were analysed using SPM96 (Friston et al., 1995). Spatial normalization of the RI image to the standard stereotaxic brain atlas (Talairach and Tournoux, 1988) was performed using transformation parameters for early integrated images of [11C]CFT 0–20 min post injection (Ito et al., 1999). After smoothing images with an isotropic Gaussian kernel of 8 mm, t statistics were performed on a voxel-by-voxel basis for contrast of the within-group condition, resulting in t-statistic maps [SPM(t)] which were subsequently transformed to the unit normal distribution [SPM(Z)]. The threshold was set at Z > 3.09 without a correction for multiple comparisons, corresponding to P < 0.001 at each voxel because areas of interest were expected a priori.

Results

Physiology and performance during gait

There were no significant changes in systolic ABP or PR during or after gait in either the normal or the Parkinson's disease group (P > 0.05, repeated ANOVA) (Fig. 1A). However, a slight drop in blood pressure from the level in the supine position, followed by a gradual increase, was observed in both groups. A moderate difference in PR level between healthy subjects (mean 62.7/min) and Parkinson's disease patients (mean 69.0/min) might reflect a state of anxiety in the Parkinson's disease group because their mean PR level was reduced to 63.1/min without changes in systolic blood pressure after their return to the waiting room. There were few changes in these physiological parameters during measurement confirming that neither orthostatic nor gait-induced haemodynamic changes were major contributory factors to the tracer uptake in the brain. The length of stride and the cadence were not significantly different between Parkinson's disease and normal groups (Fig. 1B), possibly because normal subjects were instructed to walk slowly as mentioned above. Although no significant changes were observed in performance evaluated at the straight part of the corridor, all Parkinson's disease patients had difficulty in turning at each end of the corridor.

Fig. 1

Time course changes in systolic arterial blood pressure (systolic ABP: open circles = normal, closed circles = Parkinson's disease) and pulse rate (PR: open squares = normal, closed squares = Parkinson's disease) during gait (A), which reveal no significant changes with time (repeated ANOVA; P > 0.3), and performance data (B) showing a tendency of gait difficulty in the Parkinson's disease group (closed bars) compared with the normal group (open bars) (one-way ANOVA; P < 0.1).

Correlation between estimated k3/k4 values and ratio indices in the striatal and extrastriatal regions

Estimated k3/k4 values were significantly correlated with calculated RI levels both in the striatum (r = 0.977, P < 0.001, f(x) = 1.817x + 0.077) and in the orbitofrontal cortex (r = 0.799, P < 0.05, f(x) = 0.949x + 0.032) (Fig. 2). These results indicated that it was appropriate to use late integrated images normalized by the cerebellar radioactivity for SPM analysis.

Fig. 2

Positive correlations between estimated k3/k4 values and calculated relative ratio index (RI) levels in both striatum (closed circles, r = 0.977) and orbitofrontal cortex (open circles, r = 0.799).

Changes in [11C]CFT binding by gait in normal subjects

Repeated ANOVA on changes in RI values between rest and gait conditions showed that [11C]CFT uptake was significantly reduced in the putamen (P < 0.005) and the caudate (P < 0.03) by gait (Table 2). Although percentage reduction [(Gait RI – Rest RI)/Rest RI × 100] in the orbitofrontal cortex failed to reach a significant level possibly due to a large standard deviation and the small number of participants, the magnitude of percentage reduction was greater than that in the caudate, indicating that dopamine release was enhanced in the mesocortical projection region by gait in normal subjects. SPM96 illustrated that [11C]CFT uptake in the bilateral putamen was significantly lower during gait than at rest (P < 0.001 uncorrected) (Fig. 3A). There was no significant correlation between cadence and percentage reduction in [11C]CFT uptake in the normal group, possibly due to unnatural stepping during walking as instructed prior to each PET measurement.

View this table:
Table 2

Estimated [ 11 C] β-CFT binding (at rest) and changes in ligand uptake (ratio index, RI) in two groups

GroupRegionRestGait% reduction
k3/k4 (min)RIRI
Values are shown as mean ± SD.*P < 0.05 versus control group (two-way ANOVA with multiple comparison); P < 0.05 versus rest condition (repeated ANOVA with multiple comparison).
NormalCaudate3.54 ± 0.581.90 ± 0.351.65 ± 0.33 15.0 ± 6.8
Putamen3.44 ± 0.921.94 ± 0.511.48 ± 0.29 24.1 ± 6.7
Orbitofrontal0.15 ± 0.030.11 ± 0.040.09 ± 0.03–16.7 ± 10.5
PDCaudate1.39 ± 0.34*0.76 ± 0.140.56 ± 0.10–25.8 ± 12.4
Putamen1.11 ± 0.37*0.52 ± 0.030.51 ± 0.02 2.3 ± 1.7
Orbitofrontal0.09 ± 0.01*0.09 ± 0.020.06 ± 0.01–27.4 ± 3.5
Fig. 3

Brain regions (transverse and coronal glass brain views) in which [11C]CFT uptake was significantly reduced by voluntary walking in healthy subjects (within-group comparison, P < 0.001 uncorrected) [putamen: x, y, z = –22, 6, 2 (Z = 4.67); 26, –2, –4 (Z = 4.27)] (A) and there was a significant correlation between [11C]CFT uptake reduction and task performance (cadence) in the Parkinson's disease group (correlation analysis, P < 0.001 uncorrected) [orbitofrontal area: x, y, z = 24, 20, –18 (Z = 3.93); –16, 12, –16 (Z = 3.78); 6, 4, –10 (Z = 3.78)] (B)

Changes in [11C]CFT binding by gait in Parkinson's disease patients

Consistent with our previous report (Ouchi et al., 1999b), static levels of [11C]CFT binding in the Parkinson's disease group were significantly lower in all projection regions (see k3/k4 values at rest in Table 2). Repeated ANOVA did not show any significant RI reduction in the putamen by gait, possibly due to the small differences in the RI magnitude between RI at rest and RI during gait, but significant decreases in RI level were detected in the orbitofrontal cortex and caudate possibly due to the small deviation and small number of patients in the Parkinson's disease group (Table 2). Spearman rank correlation analysis indicated a significant correlation between cadence and percentage reduction in [11C]CFT uptake in the orbitofrontal cortex (P < 0.001), with a slight, but not statistically significant, correlation in the caudate (P > 0.05) in Parkinson's disease (Fig. 4), which was confirmed by SPM correlation analysis showing that the orbitofrontal [11C]CFT uptake was inversely correlated with cadence (P < 0.001 uncorrected) (Fig. 3B).

Fig. 4

Percentage change in [11C]CFT uptake reduction against the performance level (magnitude of cadence) in Parkinson's disease patients. Spearman rank correlation shows a significant inverse correlation in the orbitofrontal cortex (r = –0.859) with a tendency towards inverse correlation in the caudate (r = –0.391).

Discussion

The present DAT study on gait performance had several caveats regarding physiology and performance of participants during gait, and data analysis. Although physiological parameters such as PaCO2, PaO2 and pH were not measured, there were no significant time course changes in systolic ABP or PR in the normal or Parkinson's disease group during gait. This finding allowed us to compare data between the two groups without global physiological effects. Regionally, however, it is possible that changes in local cerebral blood flow (CBF) may affect levels of tracer binding because an increase in regional CBF could alter distribution volume (k1/k2) in both target regions and the cerebellum. Previous SPECT and PET studies related to gait (Hanakawa et al., 1999; Mishina et al., 1999), or maintenance of standing posture (Ouchi et al., 1999a) illustrated mild or negligible increases in CBF in the striatum along with no significant CBF increase in the lateral cortex of the cerebellum (in contrast to augmentation of regional CBF in the vermal or midsagittal area), which we selected as the ROI in the present study. This indicated that striatal [11C]CFT uptake could not be influenced by normalization of cerebellar counts obtained from these cerebellar ROIs, which was based on SPM analysis in the present study. Our preliminary PET study, in which three normal subjects performed a one foot sequential flexion–extension movement with a frequency of ~1 Hz in the supine position for 60 min and then discontinued the movement, showed reduction in [11C]CFT uptake with time in the striatum contralateral to the movement side with time during the strenuous motor exercise. This did not correlate with changes in the cerebellar radioactivity (no hemispheric differences). This preliminary result agreed with the prediction that the effect on CBF of striatal [11C]CFT uptake would be negligible in the present study, though the task modality was different. This preliminary study also showed that [11C]CFT uptake reduction continued after task cessation, suggesting that elevated dopamine release occurred after gait in the present study. Although the present performance results on cadence measured in the straight part of the corridor were the same in the normal and Parkinson's disease groups, all Parkinson's disease patients showed difficulties in turning at each end of the corridor. This suggested that Parkinson's disease patients might have been coerced to continue to walk with more physical and mental effort than normal subjects. Although ratio values were lower than the estimated k3/k4 levels, the good correlation between these parameters at rest (Fig. 2) indicated that the ratio index could reflect the level of binding potential with a risk of underestimating true k3/k4 levels especially during walking. Gait was reported to increase local CBF in specific brain regions. However, as mentioned above, a gradual decrease in [11C]CFT uptake and use of cerebellar lateral hemisphere ROI permitted us to perform further analyses, not only by the ROI method, but also by SPM analysis with these calculated ratio values in the present study. The magnitude of reduction in [11C]CFT uptake (15–27%) in the present study was considerably greater than the reported level of within-subject test–retest variability (9.3 ± 4.7%; Villemagne et al., 1998), which confirmed that the results from the present ROI-based analysis reflected pathophysiological changes in DAT activity.

To our knowledge, the present study based on changes in dopamine transporter availability is the first to attempt to elucidate the in vivo contribution of dopamine in the striatal and extrastriatal projection systems to execution of basic motor exercise, i.e. gait in humans. The present study, however, could not exclude the possibility that standing posture itself induces dopamine release in the basal ganglia because dopamine-deprived parkinsonian patients exhibit postural abnormalities. ROI analysis showed a significant reduction in [11C]CFT uptake in the nigrostriatal region with dominance of the putamen (which was confirmed by SPM, Fig. 3A) and moderate reduction in the mesocortical region by gait in normal subjects. This mesocortical activation might be partly ascribed to fulfilment of constrained walking in which natural stride was deliberately reduced in normal subjects. In contrast, such gait-induced significant reduction in [11C]CFT uptake was not observed in the putamen, but in the orbitofrontal cortex and caudate in Parkinson's disease patients, although there was a possibility of statistical errors due to the small number of patients included. The present results suggested that, under normal conditions, persistence of gait execution activated dopaminergic neurones chiefly in the putamen, while the mesocortical and medial striatal dopaminergic neurones were involved in continuance of hypokinesic voluntary gait in Parkinson's disease.

The role of dopamine neurones in the dopaminergic projection regions during gait has yet to be elucidated, but our results showed that the striatal dopaminergic neurones were involved in execution of automatic movement in long-lasting human gait. In the striatum, the putamen is specifically believed to receive information about ongoing cortical motor activity from the motor cortex and the premotor cortex (Kunzle, 1975, 1978). Electrophysiological and histochemical studies indicated that dopaminergic nigrostriatal input inhibits the indirect pathway involving putamino–external pallidal–subthalamic–internal pallidal–thalamic relay to suppress negative feedback to cortical motor fields and stimulates the direct pathway involving putamino–internal pallidal–thalamic relay to provide positive feedback (Albin et al., 1989; Chevalier and Deniau, 1990; DeLong, 1990; Gerfen et al., 1990; Parent, 1990). These two pathways are thought to play separate roles; the direct pathway facilitates appropriate cortically initiated movements, while the indirect pathway suppresses conflicting undesired motor patterns (Alexander and Crutcher, 1990). Thus, putaminal activation by strenuous gait in the present study might reflect persistent excitation of the direct pathway to provide a net effect facilitating cortically initiated movement, i.e. voluntary gait. As previous functional imaging studies have shown that the cerebral precentral motor fields connecting reciprocally with the putamen were activated during bipedal gait in normal subjects (Fukuyama et al., 1997; Hanakawa et al., 1999), the present gait-induced putaminal dopaminergic activation suggested that this putaminocortical motor circuit was in a state of positive feedback facilitating the `automatic' movements during normal gait. Therefore, lesions in the putamen cause development of hypokinesia as shown in MPTP-induced akinesic monkeys with severe dopamine depletion in the putamen (Kish et al., 1988).

This was also supported by the present results showing severe reduction in [11C]CFT uptake in the putamen of Parkinson's disease patients with hypokinetic gait at rest. Interestingly, in Parkinson's disease, marked and moderate reductions in [11C]CFT uptake were found in the orbitofrontal cortex and the caudate, respectively, caused by strenuous voluntary gait in the present study. The dopaminergic projections to these areas are different from the nigrostriatal `motor' system both histologically and functionally (Scatton et al., 1982; Uhl et al., 1985; Oeth and Lewis, 1992). Electrophysiological studies indicated that dopaminergic neurones in the primate A10 area innervating the ventral striatum, amygdala and the prefrontal cortex (Szabo, 1980; Porrino and Goldman-Rakic, 1982) responded to alerting external stimuli with behavioural significance (Thierry et al., 1976; Schultz et al., 1993) and that neurones in the caudate head, which received input from the prefrontal regions, responded to environmental events that were cues for the initiation of behavioural responses (Rolls et al., 1983). In addition, the ventral striatal neurones receiving input from the inferior temporal visual cortex were also reported to respond to novel visual stimuli (Caan et al., 1984). Thus, significant activation in the orbitofrontal and caudate dopamine neurones in parkinsonian gait suggested that the mesocortical and ventrostriatal dopaminergic systems play important roles in sustaining execution of independent gait with the aid of external stimuli, specifically through visual cues in Parkinson's disease.

The contributions of these mesocortical and ventrostriatal systems to execution of parkinsonian gait could be viewed from a symptomatological view point. Spearman rank correlation analysis showed a significant correlation between cadence and reduction in [11C]CFT uptake in the caudate and the orbitofrontal cortex in Parkinson's disease patients (Fig. 4), which was confirmed by correlation analysis with SPM (Fig. 3B). The brain areas showing statistically significant SPM results chiefly encircled the bilateral orbitofrontal cortex. This indicated that more severe bipedal gait performance in Parkinson's disease patients was associated with a greater degree of activation of the mesocortical system. The finding that postural geometry and postural responses to unpredictable motion during gait were impaired in Parkinson's disease patients (Horak et al., 1996) suggested that psychomotor reactions might be dominant in support-free gait in Parkinson's disease patients.

In conclusion, this study suggests that dopaminergic neurones in the putamen play important roles in execution of routine motor movement in human gait and that gait-associated stress might also affect changes in [11C]CFT availability in the dopaminergic projection area. Abnormal activation of the mesocortical dopaminergic system and ventral striatum might be related to the pathophysiology of parkinsonian gait.

Acknowledgments

We wish to thank Dr Masanobu Sakamoto (Hamamatsu Medical Center) for advice on clinical data.

References

View Abstract