Reduced norepinephrine transporter binding in Parkinson's disease with dopa responsive freezing gait

Freezing of gait (FOG) is a major cause of falling and leads to loss of independence in Parkinson's disease (PD). The pathophysiology of FOG is poorly understood although there is a hypothesized link with NE systems. PD-FOG can present in levodopa-responsive and unresponsive forms. We examined NE transporter (NET) binding via brain positron emission tomography (PET) to evaluate changes in NET density associated with FOG using the high affinity selective NET antagonist radioligand [1 1 C] MeNER (2S,3S)(2-[a-(2-methoxyphenoxy)benzyl]morpholine) in N=52 parkinsonian patients. We used a rigorous levodopa challenge paradigm to characterize patients as non-freezing PD (NO-FOG, N=16), levodopa responsive freezing (OFF-FOG, N=10), levodopa-unresponsive freezing (ONOFF-FOG, N=21), and primary progressive freezing of gait (PP-FOG, N=5). Linear mixed models identified significant reductions in whole brain NET binding in the OFF-FOG group compared to the NO-FOG group (-16.8%, P=0.021). Additional contrasts tested post-hoc identified trends toward increased NET expression in ONOFF-FOG vs. OFF-FOG (~10%; P=0.123). Linear mixed models with interaction terms identified significantly reduced NET binding in right thalamus in the OFF-FOG group (P=0.038). A linear regression analysis identified an association between reduced NET binding and more severe NF0G-Q score only in the OFF-FOG group (P=0.022). This is the first study to examine brain noradrenergic innervation using NET-PET in PD patients with and without FOG. Based on the normal regional distribution of noradrenergic innervation and pathological studies in the thalamus of PD patients, the implications of our findings suggest that noradrenergic limbic pathways may a play a key role in OFF-FOG in PD. This finding could have implications for clinical subtyping of FOG as well as development of therapies.

This study showed that decreased FMT uptake in LC was strongly correlated with increased self-reported FOG severity in PD patients 33 . This observation must be confirmed through more comprehensive studies in larger cohorts of patients. Finally, the relationship between FOG and anxiety is also consistent with the link between FOG and NE 34,35 . FOG can be elicited in situations of elevated stress 36 , which increases LC activity, NE release and arousal in animal models 6,[37][38][39] . This suggests the hypothesis that a dysregulated LC-NE system may have aberrant responses to stress in PD that could, in turn, trigger FOG episodes.
In this study we examined NE transporter (NET) binding via positron emission tomography (PET) to evaluate changes in NET density associated with FOG. NET is a member of the Na+/Cl-dependent neurotransmitter transporter gene family located on the nerve terminals as well as the somatodendritic field of NE neurons and serves as a proxy for LC terminal integrity. We used the radioligand [11C]MeNER (2S,3S) (2-[α-(2-methoxyphenoxy) benzyl]morpholine), which has high affinity and is a selective antagonist for NET 40,41 , to compare NET availability across PD patients with and without FOG, as well as in a small group of patients with non-PD FOG (referred to as primary progressive Freezing of Gait) 42 . We also separated PD-FOG patients into those responsive and non-responsive to levodopa 43 .

Study population
Study patients were recruited from the Emory Movement Disorders Clinic and provided written informed consent according to procedures approved by the Institutional Review Board of Emory University.
Inclusion criteria for PD were: Age≥18 years; PD diagnosis defined by the United Kingdom Brain Bank criteria 44 ; Hoehn & Yahr stage IV or higher in the OFF state; levodopa responsive; capable of signing a consent document and willing to participate in all aspects of the study. Inclusion criteria specific to participants with FOG were FOG noted in medical history and confirmed visually by examiner. Patients with primary progressive freezing gait (PP-FOG) diagnosed based on previously reported criteria 45 were included; while F o r P e e r R e v i e w patients previously treated with dopamine receptor blocking medications; or with neurological or orthopedic disorders interfering with gait including vascular parkinsonism; dementia or other medical problems precluding completion of study protocol were excluded.

Levodopa challenge paradigm
The details of the levodopa paradigm challenge have been published elsewhere 43 . In brief, patients came to clinic in the practically defined "OFF" state >12 hours after last intake of antiparkinsonian medications 46 .
They were assessed for motor symptoms using the Movement Disorder Society-Unified Parkinson's Disease Rating Scale part III motor examination (MDS-UPDRS-III). After the assessment, patients were administered a levodopa equivalent dose (LED) greater than their typical morning dose. The assessment was repeated after patients reached their full "ON" state.

FOG group assignment
We classified each PD patient into one of three study groups based on history of the presence of FOG along with scores on the MDS UPDRS-III item 11 "Freezing of Gait" exam in each of the "OFF" and "ON" medication states. Participants who had no history of FOG and received a score of zero on item 11 in both medication states were classified as "no freezing," or "NO-FOG." Participants who received a nonzero score on item 11 in the "OFF" medication state but a zero score in the "ON" medication state were classified as "OFF-FOG." Participants who received a nonzero score on item 11 in both medication states were classified as "ONOFF-FOG." Patients in the primary progressive Freezing of Gait group were classified as "PP-FOG." [ 11 C]MeNER-PET Imaging Patients were assessed with NET-PET imaging on a separate testing day. The typical interval between levodopa challenge testing and NET-PET imaging was 2.2 ± 2.8 months. Urine pregnancy tests were obtained within 24 hours prior to injection of [ 11 C]MeNER for female subjects, unless no menses during the past CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 15, 2022. ; https://doi.org/10.1101/2022.03.14.22272365 doi: medRxiv preprint F o r P e e r R e v i e w consecutive 12 months. Any noradrenergic medications were held for a duration based on the half-life. A telephone contact seven days post imaging session was made to assess for any adverse events. Patients maintained their typical medication schedule for NET-PET imaging.

Imaging Procedures
PET data was collected on a High-Resolution Research Tomography (CTI, Inc. Knoxville, TN) at the Emory Center for Systems Imaging. The PET system has a 2.5mm spatial resolution and 10-fold higher sensitivity than current clinical PET cameras 48 . Subjects had a venous catheter placed in their forearm and were administered [ 11 C]MENeR in a quiet uptake room for 60 minutes. Subjects were then placed supine on the imaging table with their head lightly restrained using Velcro straps, and 30 min of PET data was collected. PET data were divided into six frames, each five minutes duration and reconstructed with an ordinary Poisson ordered-subset expectation maximization (OP-OSEM) algorithm (6 iterations, 16 subsets) including scatter and attenuation correction.
Structural MRI images were acquired on a 3T Siemens Prisma scanner with a 32-channel receiver array head coil. A standard whole-brain 3D T1-weighted MPRAGE sequence 49 recommended for brain structure morphometric analysis 50 (FOV = 256 mm; TR/TI/TE/FA = 2530 ms/1100 ms/3 ms/8°; 1 mm X 1 mm X 1 mm resolution; time of acquisition ~ 9 minutes) was acquired to provide anatomic detail. Foam padding was utilized

Image Analysis
PET data was corrected for inter-subject motion and co-registered to their structural MRI scan by optimization of the mutual information metric. MRI structural data were then spatially normalized to the Montreal Neurological Institute (MNI T2 152) template with SPM8's segment option and default parameters. The inverse non-linear transform was used to map regions of interest from the Anatomical Automatic Labeling template 51 including the frontal lobe, left and right thalamus, temporal lobe, LC and cerebellum. Regional standardized uptake value ratios (SUVR) were extracted from the reconstructed PET data normalized to the cerebellum 33 . We have shown that this approach is highly correlated with NET binding density in structurally similar analogue based on reboxitine 52 . Imaging technologists and those assessing results were blinded to the their group assignment.

Statistical Analysis
All summary statistics are presented as mean (standard deviation) or frequency (percentage). Clinical and demographic variables were compared across study groups using univariate tests (Chi-squared, ANOVA).
Variation in NET binding across study groups was assessed with a linear mixed model with fixed effects for study group (NO-FOG, OFF-FOG, ONOFF-FOG, and PP-FOG, with NO-FOG as the reference group) and brain region (Frontal cortices, left thalamus, right thalamus, locus ceruleus, temporal lobes, with frontal cortices as the reference group) and a random effect for patient. Subsequent models evaluated sensitivity of main results to inclusion of covariates identified as marginally significant in univariate tests (sex, duration). Region-specific group effects were assessed with an additional linear mixed model with region by group interaction terms.
Linear mixed models were implemented in LmerTest::lme4 in R software and fitted using restricted maximum likelihood. Linear regression models were applied to evaluate associations between NET expression and self- Data availability: The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Clinical and demographic characteristics
A total of N=60 patients were enrolled and assessed with levodopa challenge. Of these, PET data was subsequently obtained for 52: Sixteen were classified as NO-FOG, 10 as OFF-FOG, and 21 as ONOFF-FOG.
The remaining five patients were PP-FOG. Clinical and demographic characteristics are presented in Table 1.
No statistically significant differences in frequency of missing data were observed across study groups (Supplemental Information). No statistically significant differences in clinical or demographic variables were observed between groups; with the exception of N-FOGQ score and MDS-UPDRS-III FOG item scores (expected from the design) and marginally significant effects of sex (P=0.05) and disease duration (P=0.08).
Higher levels of anxiety were reported among those with freezing (average scores of 10.8±8.2 vs. 8.4±6.1 on the Beck Anxiety Inventory); however, this contrast was not statistically significant (P=0.69, ANOVA).

NET binding
In general, a similar pattern of variation in NET binding across study groups was observed across all brain regions of interest. The lowest levels of NET SUVR values were observed in the OFF-FOG group, followed by PP-FOG, ONOFF-FOG, and NO-FOG, respectively. (Region-specific values are summarized in Supplemental Information; Table S1). Linear mixed models identified significant reductions in whole brain NET binding in the OFF-FOG group compared to the NO-FOG group (change in SUVR -16.8%, P=0.021; CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) Linear mixed models with interaction terms between study group and brain region identified a significant reduction in NET binding in right thalamus in the OFF-FOG group (P=0.038). A similar reduction of smaller magnitude was observed in left thalamus (P=0.157).

Right thalamus NET expression and NFOG-Q score
Because of the strong interaction observed between NET binding and OFF-FOG group in the right thalamus, we tested whether NET abundance in this area was associated with self-reported FOG severity as measured on NFOG-Q. The PP-FOG group was excluded from this analysis due to the small number of cases.
A linear regression of NFOG-Q score onto right thalamus NET availability identified a statistically significant negative relationship, with reduced NET associated with more severe NFOG-Q score in the OFF-FOG group (P=0.022), but not in the other groups ( Figure 2). The overall variation in NFOG-Q score explained by NET binding with interaction terms for group was R 2 adj = 0.88. Similar relationships were observed in a model with the square root of NFOG-Q score as the dependent variable to stabilize variance (Supplemental Information).

Discussion
Page 10 of 41 ScholarOne, 375 Greenbrier Drive, Charlottesville, VA, 22901 Support (434) 964 4100 Brain . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) There is mounting evidence supporting a role for extranigral noradrenergic (NA) systems in the development of FOG in PD. This is a first look at the use of NET-PET imaging using the ligand [ 11 C]MeNER as a tool to examine the role of NA denervation in PD-FOG. We examined whole brain NET binding via PET to evaluate changes in NET density associated with FOG; and particularly, in relation to FOG responsiveness to levodopa. We scanned 47 PD patients (16 NO-FOG, 10 OFF-FOG, 21 ONOFF-FOG) and 5 patients with primary progressive freezing gait (PP-FOG). We found significantly decreased whole-brain NET binding in the OFF-FOG group compared to NO-FOG (P=0.021). Importantly, we showed that this deficit was specific to OFF-FOG; patients with ONOFF-FOG or PP-FOG showed NET binding comparable to those without FOG.
Additional region-specific comparisons revealed decreased binding in the right thalamus (P=0.038) with a trend for the left thalamus (P=0.16), in the PD OFF-FOG group. These findings support the hypothesis that a loss of NA innervation, perhaps particularly to the thalamus, may contribute to levodopa responsive FOG (OFF-FOG) in PD patients. The results also add to previous findings suggesting a connection between NA dysfunction and FOG. Further, this decreased NA innervation may relate to the recently reported increase in CSF Aβ42 amyloid in PD-FOG patients 53 as previously suggested by studies in non-human primates and several transgenic rodent models where the reduction of NE contributed to increased Aβ pathology 54-56 . Interestingly, the increased amyloid aggregation in these models was reversed by using droxidopa 57 .

Structural and functional changes in the thalamus in FOG
This work adds to a growing body of knowledge demonstrating structural and functional changes in the thalamus -and in the connections of the thalamus to other brain regions -in PD patients with FOG. In a prospective MRI neuroimaging study, D'Cruz et al 58 showed local inflations in bilateral thalami in patients who presented with or would later develop FOG. These inflations involved medial thalamic sub-nuclei volumes which are highly innervated by NA pathways. Resting-state analyses at baseline further revealed that patients is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 15, 2022. ; https://doi.org/10.1101/2022.03.14.22272365 doi: medRxiv preprint F o r P e e r R e v i e w who would go on to develop FOG had thalamo-cortical coupling with limbic and cognitive regions initially stronger than that among patients who would not go on to develop FOG. This initial increased coupling exhibited a marked reduction over the two years. The authors hypothesized that many of these changes may have been compensatory in nature, with increased reliance on non-motor thalamic networks, particularly limbic connections, in PD patients with FOG; after initial success, decompensation occurred, and FOG would emerge over the course of disease progression.
Other studies have shown several anatomical changes in connections between the thalamus and other brain regions in FOG, including some known to be crucial for balance and gait. These include diminished structural connectivity between the thalamus and mesencephalic locomotor region and reduced thalamic fiber tracts in the right hemisphere in FOG 59 and more pronounced white matter abnormalities in multiple tracts connecting to the thalamus in FOG 60 . It has also been shown that FOG episodes are associated with decreased anterior thalamic Blood-Oxygen-Level-Dependent (BOLD) signal, potentially associated with paroxysmal reductions in mesencephalic locomotor region activity 61,62 .
Our results imply that there is decreased NA innervation of the thalamus in patients with levodopa responsive FOG, which strongly suggests that interaction between NA sensory signaling and dopaminergic movement regulation is particularly important for OFF-FOG phenotype. It is possible that in the presence of dopamine replacement, some patients are able to successfully leverage non-motor regions of basal-gangliathalamocortical loops to compensate for impaired NA motor -specifically gait and balance -pathways. It is conceivable that this could lead to compensatory "inflation" of these areas as described by D'Cruz and colleagues, and that additional damage to other tracts impinging on the thalamus would impact this compensation. However, because the majority of these studies did not classify patients by FOG levodopa responsive state, it is difficult to draw firm conclusions. What appears clear is that there is a class of patients with greater NA deficits in the thalamus who do not experience FOG in the presence of adequate dopamine. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 15, 2022. ; https://doi.org/10.1101/2022.03.14.22272365 doi: medRxiv preprint Noradrenergic innervation of the thalamus Noradrenergic innervation of the thalamus has been examined in monkey and human in some detail. In a study of two nonhuman primates, NA axon maps were generated with dopamine β-hydroxylase (DβH) and NET immunohistochemistry, and in vitro quantitative autoradiography for alpha-1 and alpha-2 adrenergic receptor densities 63 . The distribution patterns of DβH, NET, and adrenergic receptors were very similar with some modest variation noted. The most densely innervated thalamic regions were the midline, caudal intralaminar (paracentral and parafascicular), and medial mediodorsal nuclei. NA axons were present at moderate levels within the anterior nuclei, while the ventral motor thalamic nuclei exhibited moderate to low NA innervation. The sensory relay nuclei were generally less innervated. The visual relay nucleus, dorsal geniculate nucleus displayed the lowest NA innervation. These findings were concordant with postmortem human results 13,64,65 .
Although NET-PET does not provide the spatial resolution to distinguish the distribution of NET binding within the various thalamic nuclei, the impact of decreased thalamic NET binding can be speculated based on these findings and considering the systems thought to be involved in freezing, including motor, visual, cognitive, and limbic circuits 66 . It is noteworthy that the highly NA-innervated caudal intralaminar regions include predominantly thalamic neurons that project heavily to associative and limbic striatum 67,68 . The dense NA axonal innervation of these thalamic nuclei may allow for an indirect NA modulation of basal ganglia activity across motor and non-motor functional domains 63 . Furthermore, these densely innervated regions are known to receive extensive cholinergic innervation from the pedunculopontine tegmental area 64 , a cholinergic brain network that is also severely disrupted in PD 69 .  These findings indicate that the burden of the PD-related pathology is seen in nuclei representing essential parts of the motor and limbic circuit 70 .

Limbic circuits and FOG
These results are consistent with the hypothesis that abnormal limbic-striatum connectivity may interfere with processing specific anxiety provoking situations that lead to FOG -particularly OFF-FOG. As shown, the thalamic NA system is closely linked to limbic circuits and, in turn, we have shown in this study a link between NA changes and FOG. Other studies have shown a connection between FOG and limbic circuit changes. The limbic circuit plays a critical role in generating adequate emotional and behavioral responses, regulation of motivational processes and social behavior, and emotional processing, particularly of stimuli that induce anxiety and depression, including in PD 71-75 , all of which are essential for learning and memory. It has been suggested that this circuit is connected to the development of FOG based on several reports of worsening anxiety and depression in PD-FOG patients compared to PD-NOFOG 34,35 . Further, greater situational anxiety is also associated with worse FOG severity 35 . Imaging studies 66, 76 have found that connectivity between the amygdala and striatum was significantly increased in freezers compared to non-freezers. It is suggested that the increased connectivity between the limbic circuit and striatum represents over-integration between these normally segregated networks which may lead to transient overload of motor network processing, that is the already  77,78 . In addition, anti-coupling between the cortical and subcortical limbic network and within the subcortical network was marked during freezing 66 . This might suggest there is an abnormality in processing specific anxiety provoking situations that lead to FOG, particularly OFF-FOG, and that this relates to abnormal limbic-striatum connectivity. Considering that anxiety and panic frequently occur as manifestations of off episodes 79 it makes sense that the limbic circuitry would relate to the OFF-FOG group specifically.
ONOFF vs OFF FOG in PD: Different pathophysiology?
The results of this study provide additional evidence supporting the hypothesis that OFF-FOG and ONOFF-FOG result from distinct pathophysiology. We have shown that OFF-FOG relates to hallucinatory phenomena while ONOFF-FOG is associated with distinct cognitive features -with the implication that these two types are associated with distinct pathophysiology in regions outside the basal ganglia 80 . We have also shown that axial and lower limb parkinsonian signs are also less severe and more levodopa responsive in OFF-FOG than ONOFF-FOG -with the implication that these types are associated with distinct pathophysiology within the basal ganglia 43 . The present results also support the notion that OFF-FOG and ONOFF-FOG are likely governed by separate pathophysiological mechanisms.
We have now shown a potential neurochemical marker separating FOG levodopa responsive subtypes, whereby only the OFF-FOG group demonstrates significantly lower NA innervation. Although some investigators have suggested that FOG progresses through a continuum from OFF-FOG to ONOFF-FOGindicating that they are not separate entities 81 -others support the idea that ONOFF-FOG is an independent form that can develop without prior OFF-FOG 82,83 . In fact, we previously showed through detailed examination of medical records that ONOFF-FOG can appear without first transitioning through OFF-FOG 43 . Our results  Based on these results, we speculate that trials of NA agents for FOG would likely benefit from including only those patients with OFF-FOG (those with greatest decrease in NET PET signal), as it stands to reason that only these patients will likely benefit. Previous trials may have underestimated efficacy of agents if ONOFF-FOG cases were included. We consider this to be relatively likely, as patients with more severe FOG often have unresponsive FOG. Therefore, those who might have been the most likely to be recruited for trials might have been those least likely to benefit.  89 , and it has been suggested that PP-FOG may result from extranigral neuronal degeneration.
Based on our results, we suggest that disruption of the NA system may not contribute significantly to this type of FOG, because NET-PET binding was similar to the ONOFF-FOG and NO-FOG groups but differed from the OFF-FOG group. However, because of the small sample size, additional subjects will be needed to confirm this result, taking into consideration that multiple pathophysiological bases exist for ONOFF-FOG.

Limitations
There are some limitations to this study. The sample size was moderate, but due to the well-recognized demographic and clinical imbalances in PD patients with and without FOG, it remains difficult to completely control for potential confounds related to increased disease progression and age usually observed in FOG patients. We are encouraged by the fact that the main deficit seen in OFF-FOG was not seen in more severe ONOFF-FOG, which makes confounding by overall disease severity less likely. Because of the nature of observational studies, the potential for confounds remain, and these results should be replicated in a larger sample and preferably in a prospective context. In addition, spatial resolution of PET has reduced our ability to image certain regions such as specific nuclei of the thalamus and the LC. Conclusion This is the first study to examine brain NA innervation using NET-PET in PD patients with and without FOG. We found significantly decreased whole-brain NET binding in the OFF-FOG group compared to NO-FOG. Additional region-specific comparisons revealed decreased binding in the thalami, right greater than left and more notable for the PD OFF-FOG group than the ONOFF-FOG and NO-FOG groups. Based on the normal regional distribution of NA innervation and pathological studies in the thalamus of PD patients, our findings suggest that limbic pathways may play a key role in OFF-FOG in PD. This finding could have implications for clinical subtyping of FOG as well as development of therapies.  Brain . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 15, 2022. ; https://doi.org/10.1101/2022.03.14.22272365 doi: medRxiv preprint Brain . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 15, 2022. ; https://doi.org/10.1101/2022.03.14.22272365 doi: medRxiv preprint Brain . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)     CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)  CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

Incomplete imaging data
Of N=60 participants originally enrolled in the study, NET-PET imaging was not completed for N=8 (N=1, NO-FOG; N=1, OFF-FOG; N=5, ONOFF-FOG; N=1, PP-FOG). Only partial documentation for missing data was available. Reasons for missing data were: patient moved away, N=1 NO-FOG; inability to place head in scanner, N=1 OFF-FOG, no documentation available, N = 5 ONOFF-FOG, N=1 PP-FOG. No statistically significant difference in frequency of missing imaging data was observed across study groups (P=0.61, chisquared test).

Numerical values of NET expression
Regional standardized uptake value ratios (SUVR) extracted from the reconstructed PET data normalized to the cerebellum are shown in Table S1. Numerical values of 1.0 correspond to average expression equal to that observed in the cerebellum; numerical values ≥ 1.0 correspond to average expression greater than that observed in the cerebellum, which is considered to have negligible NET expression.

Linear regression of NFOG-Q score onto right thalamus NET expression with square root transform
Because of the potential that non-normal distribution of residuals might invalidate the assumptions of linear regression, we repeated the regression analysis using square-root transformed NFOG-Q scores as the Brain . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 15, 2022. ; https://doi.org/10.1101/2022.03.14.22272365 doi: medRxiv preprint F o r P e e r R e v i e w outcome variable. The results were very similar to those of the linear regression model presented in the main body of the manuscript, with a significant variation in NFOG-Q score with right thalamus NET expression in the OFF-FOG group (P=0.012) but not in the other groups. The overall variation in square root transformed NFOG-Q score explained by the model was R 2 adj = 0.96.

Association between MDS-UPDRS-III FOG score and right thalamus NET expression
We also examined whether right thalamus NET expression was associated with increased FOG severity as measured on the MDS-UPDRS-III. We dichotomized MDS-UPDRS-III FOG score in each of the ON and OFF states as 0-1 or ≥2 and dichotomized right thalamus NET expression about its median. Logistic regression models with terms for NET expression category and Group failed to identify statistically significant relationships in either the OFF or ON state.