Introduction

Parkinson’s disease (PD) is characterized by loss of dopaminergic neurons and is effectively treated by the dopamine precursor, levodopa. Chronic levodopa treatment, however, leads to involuntary rhythmic motions called levodopa-induced dyskinesia (LID) in 30–80% of patients.

Several experimental studies were performed to examine the role of the blood–brain barrier (BBB) in the pathogenesis of LID. In a Gd-DTPA MRI study in the 1-methyl-4-phenyl-1,2,3,6-tetrahyropyridine (MPTP) primate model, no evidence was found for increased BBB permeability in LID animals [1]. The images, however, were acquired in the un-medicated state, using a relatively high-molecular-weight contrast agent. Subsequent work in the 6-hydroxydopamine (6-OHDA) rat dyskinesia model revealed levodopa-related increases in striatal BBB permeability using autoradiographic techniques [2] as well as microPET [3]. In line with observations in the rodent model, human postmortem tissue study also disclosed evidence of potential BBB leakage in advanced PD patients with dyskinesia [4].

While these data are consistent with alterations in BBB permeability in human PD, objective evidence to this effect is currently lacking. To determine whether permeability is locally increased by levodopa administration, we used PET with the potassium (K+) analogue rubidium-82 (82Rb), half-life 76 s, which is sensitive to the disruption of BBB tight junctions. We hypothesized that levodopa causes an increase in striatal BBB permeability, which in addition to the normal large neutral amino acid (LNAA) transport process, results in net influx of the drug into the extravascular space. We also hypothesized that this change is greater in patients with LID compared to their non-LID (NLID) counterparts.

Methods

Subjects

We studied 19 PD patients and 12 healthy control subjects as described elsewhere [5]. The PD patients were divided into two groups based on the presence of LID during an intravenous levodopa infusion (see below). The LID group comprised eight subjects in whom levodopa infusion caused sustained dyskinesia. Study details for these subjects are provided in Table 1. The non-LID (NLID) group comprised 11 subjects: nine had stable motor responses to levodopa infusion without; two were drug-naïve. The clinical characteristics of the two PD groups are provided in Table 2. The study was approved by the Institutional Review Board of Northwell Health. Written informed consent was obtained from all participants.

Table 1 Study details for individual LID subjects
Full size table
Table 2 Clinical profiles of patients with Parkinson’s disease
Full size table

Positron emission tomography

The PD patients and the control subjects underwent baseline 82Rb PET (saline infusion) after fasting overnight and/or consuming a very low-protein breakfast at least 4 h prior to the PET study. Oral dose of 200 mg carbidopa was administered one hour prior to the levodopa infusion. In the chronically treated PD patients, antiparkinsonian medications were withheld at least 12 h before imaging. UPDRS motor ratings were obtained prior to imaging in all PD patients. Those on daily oral medication (n = 17) were rescanned on the same day during an intravenous levodopa infusion as detailed elsewhere [5, 6]. In each PD subject, drug infusion was titrated to achieve 20–40% improvement in UPDRS motor ratings for at least 30 min before the start of imaging, with < 10% change in UPDRS ratings obtained at subsequent 30-min intervals. In the LID subjects, the rate of levodopa infusion was determined by the initiation of dyskinesia, which was usually within 30 min. The levodopa infusion rate was then adjusted to reduce the induced, involuntary movements to a minimum (average infusion rate: 1.2 ± 0.6 mg/kg/h). PD and healthy control subjects were additionally scanned with [15O]-water (H215O) and [18F]-fluorodeoxyglucose (18FDG) PET before and during the infusion as part of a larger protocol as described elsewhere [5, 7]. A flowchart of the radiotracer studies that were conducted is provided in Fig. 1.

Fig. 1
figure 1

Flowchart for the PET studies. Day1. a [15O]-water (H215O) PET studies for cerebral blood flow and 82Rb bolus injection for BBB permeability performed in the baseline off-state and during levodopa infusion. Day2. b H215O PET bolus studies in the off-state during normocapnia (room air) and during hypercapnia (5% CO2 rebreathing) [7]. [18F]-FDG PET for cerebral glucose metabolism before and during levodopa infusion was conducted as described previously [5]

Full size image

A 10-min transmission scan was performed before the PET scan. 740 MBq (20 mCi) of 82Rb, eluted from a 82Sr/82Rb generator (Bracco Diagnostics Inc., Monroe Township, NJ) in normal saline was injected as an intravenous bolus for approximately 20 s. Serial PET images were acquired for 6 min (1 × 20 s, 1 × 10 s, 6 × 5 s, 4 × 7 s, 7 × 30 s, and 1 × 62 s; Fig. 2) in three-dimensional mode using the GE Advance PET scanner (GE Medical Systems, Milwaukee, WI). This eight-ring bismuth-germanate scanner provided 35 two-dimensional image planes with transaxial resolution of 4.2 mm in all directions. All PET studies were performed in the rest state with the subjects’ eyes open in a dimly lit room with minimal auditory stimulation.

Fig. 2
figure 2

Rubidium-82 time–activity curves in sagittal sinus and putamen. a We obtained serial rubidium-82 scans for 6 min after bolus injection for each subject and condition. The images shown correspond to those at 43, 48, 53, and 58 s in a subject. The radioactivity in the superior sagittal sinus (SSS, arrows) reached a peak at approximately 1 min. b Time activity curves show rubidium-82 activity during OFF (saline infusion) and ON (levodopa infusion) in a subject. The time course was similar in the right putamen (black) and the SSS (blue). The scaled activity in the putamen (gray) largely overlapped with the SSS activity (blue), suggesting that there was no transfer of rubidium-82 across the blood–brain barrier

Full size image

Data analysis

Preprocessing of all image data was performed using Statistical Parametric Mapping (SPM) 5 (http://www.fil.ion.ucl.ac.uk/spm/) run on MATLAB 7.3.0. Dynamic 82Rb PET images were realigned for motion corrections across time series and then spatially normalized using the same spatial transformation parameters determined between individual transmission PET image and transmission PET image template in a standard brain space. The PET images were smoothed with an isotropic Gaussian kernel full-width at half maximum of 10 mm. The concentration in tissue (Ct) was measured in anatomical regions-of-interest (ROIs), as well as predefined volumes-of-interest (VOIs) based on a priori information using Scan Analysis and Visualization Processor software (http://feinsteinneuroscience.org). We evaluated Ct in the bilateral anterior cingulate cortex, posterior putamen, and thalamus as aromatic L-amino acid decarboxylase-rich (AADC) regions, as well as in the right superior temporal gyrus and inferior parietal lobule (which have low AADC expression) as control regions. The pre-determined VOIs were 3-mm-radius spheres centered at four coordinates in the putamen and thalamus, in which significant levodopa-mediated dissociation of cerebral blood flow and glucose metabolism was discerned in the same PD patient sample [5]. The concentration in blood (Cb) was measured in a 3-mm-radius sphere centered in the superior sagittal sinus. Additionally, we calculated the volume of distribution (Vd) for each of the regions as the average of Ct/Cb of the last 7 frames spanning 2–6 min (see Appendix. Modeling approach, power calculations), which was compared between groups and conditions.

Comparison between normal controls and patients with and without LID was performed using ANOVA with post hoc Tukey–Kramer HSD correction for multiple comparisons. The difference between values measured on and off levodopa was assessed separately for PD patients in the LID and NLID groups using paired t tests. Statistical analyses were conducted using SPSS version 22 (IBM Corporation, Armonk, NY). Results were considered significant at p < 0.05.

Results

Given that K1 and Ki could not be estimated in the 82Rb PET data (see discussion of the modeling approach in the Appendix), we concluded that substantial alterations in BBB were not present on- or off-levodopa infusion. We did, however, use the data to estimate Vd as a measure of the changes in distribution volume in AADC-rich regions in the posterior putamen, thalamus, and anterior cingulate cortex (see Methods). Vd measurements were then used to determine whether distribution volume is altered in one or more of these regions, particularly in areas found previously to exhibit significant levodopa-mediated flow-metabolism dissociation in the same population [5]. We found no significant difference in baseline Vd measured in these regions (Fig. 3) in the PD group as a whole, or in the LID and NLID subgroups compared to healthy control subjects (p > 0.1). A sole exception is the left thalamus in which a significant increase in the off-state was observed in LID compared to NLID (F[2,28] = 3.85, p = 0.033; Tukey–Kramer HSD). Likewise, no significant Vd changes were seen (Fig. 3) for within-subject differences between baseline and on-state values in these brain regions (see Fig. 4).

Fig. 3
figure 3

Volume of distribution in various regions in normal, LID, and NLID. We compared apparent volume of distribution (Vd), which served as a measure of tissue vascular space, between the normal controls (NL) and patients with Parkinson’s disease in the OFF state (saline infusion). The patients were classified into two groups according to the absence or presence of levodopa-induced dyskinesia (LID). a We measured Vd in the atlas-based regions-of-interest that included aromatic l-amino acid decarboxylase-rich regions, and found no significant increases except in the left thalamus of LID subjects (one-way ANOVA, p = 0.032 with post-hoc correction for multiple comparisons using Tukey–Kramer HSD). b We also measured Vd in the volumes-of-interest in which we previously reported flow-metabolism dissociation, and found no significant differences. NLID non-LID

Full size image
Fig. 4
figure 4

Volume of distribution during saline and levodopa infusion in aromatic l-amino acid decarboxylase-rich regions. We compared volume of distribution (Vd) between saline infusion (OFF) and levodopa infusion (ON) conditions in Parkinson’s disease patients with or without levodopa-induced dyskinesia (LID). We detected no significant Vd changes in the aromatic l-amino acid decarboxylase (AADC)-rich regions-of-interest as well as ROIs devoid of AADC (a) or the volumes-of-interest of flow-metabolism dissociation (b). NLID non-LID

Full size image

Discussion

Abnormal increases in local BBB permeability were not seen with 82Rb PET in LID or NLID PD patients, whether scanned in the un-medicated baseline state or during levodopa infusion. The current results contrast with the striking changes seen under comparable pharmacological conditions in the 6-OHDA rodent dyskinesia model [2, 3]. We note that the rodent dyskinetic model differs from human LID in a number of respects. First, while the experimental model is based upon acute and local injury of the ascending dopamine fiber bundle by 6-OHDA injection, human PD develops insidiously with widespread pathology that is not limited to the nigrostriatal pathway. More important perhaps is the dosage of levodopa which was substantially larger in rodent compared to the human [3]. Lastly, the method used to assess BBB permeability in the experimental model is different than that used in the human PD study. In the rodent model, α-aminoisobutyric acid (AIB), a neutral amino acid, was used. This tracer can cross intact endothelial cells as well as disrupted tight junctions of the BBB. It is primarily transported across BBB endothelial cells by the amino acid transporter A system, and also by the L and ASC systems [8]. By contrast, 82Rb does not depend on amino acid transport and therefore is not likely to be influenced by levodopa in the same way. That being said, substantial evidence exists apart from AIB uptake measurements in the rodent to implicate dysfunction of the striatal neurovascular unit in the pathogenesis of LID in human PD [4, 9, 10].

Our study was based upon animal data suggesting tight junction opening as a cause for BBB permeability. Therefore, 82Rb, a potassium analog that crosses the BBB independently of the LNAA transporter system, was selected to detect changes at tight junctions induced by levodopa. That said, pharmacological modulation of LNAA transporters may be another cause for permeability changes as seen with radiolabeled AIB in rodent studies. The LNAA transport system has been extensively studied, and up-and-down regulation has been reported. In particular, (1) plasma amino acid levels compete with and can decrease levodopa uptake in the brain [11, 12]; (2) endotoxin-induced inflammation can down-regulate L-type amino acid BBB transporter 1 (LAT1) in rodents [13]; (3) Up-regulation of LAT1 is well established in oncology and has been successfully used in imaging tumors [14]. However, up-regulation of LAT1 for the chronic levodopa treatment has not been reported in PD.

Other radiotracers can be used to assess LAT1 transporter function, such as [18F]-3-O-methyl-dopa (3-OMFD) and [11C]-methylaminoisobutyric acid (MeAIB). In an [18F]-3-OMFD PET study in human PD, we found that the kinetics of this tracer were similar to that of [18F]-FDOPA, but with small regional differences [15]. Studies using 3-OMFD can be performed in the early disease stage (off-levodopa), but data analysis is complicated in the on-levodopa dyskinesia state by interactions between levodopa metabolites and the radiotracer. [11C]-MeAIB, as with [11C]-AIB, would be subject to LNAA transporter modulations (especially the A, ASC, and L systems) in the course of levodopa administration.

Future off-state studies can be undertaken with either of these tracers, although long-duration dynamic scans, particularly in LID patients, may be difficult to perform. Methods to differentiate changes in tight junctions from LNAA transporters will be helpful, such as a recently developed non-contrast MRI technique [16].

Estimation of the influx constant K1 as a direct measure of BBB permeability was not possible in the PD or the healthy control subjects. Nonetheless, using Vd as a measure of distribution volume changes and possibly an indirect reflection of angiogenesis, we found no evidence of change in the apparent volume of distribution in PD patients with LID, relative to control subjects or subjects without LID except for the left thalamus in the “off” state. The enhanced cerebral blood flow response to levodopa in the left thalamus observed in LID subjects [17] may therefore reflect a change in Vd, at least in part.

Even though a number of radiotracers for angiogenesis have been developed for oncologic use, none to date are suitable to detect the localized, low magnitude changes observed in experimental LID. Vasodilation resulting from levodopa infusion [18] can be confirmed using C15O PET, but these studies were not performed because of concerns over radiation dose. The issues attendant to absolute values of K1 and Vd for 82Rb PET are further discussed in the Appendix.

Finally, we speculate that for very low K1 values the 82Rb bolus technique may not be able to estimate very small changes in influx accurately because of the short half-life of the tracer and the correspondingly short scan duration. This issue may be addressed using a constant infusion technique. That, however, would require a separate blood volume study and additional radiation exposure. The possibility of a small, highly localized levodopa-mediated permeability change in PD generally, and in LID in particular, cannot be excluded based on the current data.

A number of technical issues attendant to 82Rb PET are also worthy of mention. We did not sample the tracer in the blood during scanning because of the complexity of the necessary protocol. For a primarily intravascular radiotracer, such as 82Rb, a curve derived from radioactivity measurements in one of the large cerebral blood vessels is theoretically a better representation of the arterial input function than actually sampling radial arterial blood (with associated smearing). Deriving input function from images also makes the procedure more tolerable to the subjects. In human 18FDG PET studies, a close relationship between carotid artery image-derived function (IDF) and actual arterial sampled curves has been demonstrated [19]. However, the absolute scale of the IDF cannot be accurately determined without actual sampling of peripheral blood for calibration, which was not done here. Lastly, to reduce motion artifacts, we titrated the levodopa infusion to just below the dyskinesia threshold. Thus, the possibility exists that a significant increase was not observed in the absence of LID. Indeed, given the limited number of subjects in the current study and the complex nature of the imaging protocol, the results may not be generalizable to the PD population at large.

Conclusion

Our findings do not support the presence of significant alterations in BBB permeability in PD patients with or without LID. The possibility of small, highly localized permeability changes in PD cannot be completely excluded using our 82Rb PET method. Increased off-state Vd in the left thalamus of LID subjects in the off-state is an interesting finding, which will require further confirmation.