Mapping neuromodulatory systems in Parkinson’s disease: lessons learned beyond dopamine

Parkinson’s disease (PD) is the second most common neurodegenerative disease with mixed motor and non-motor symptoms. Dopaminergic drugs remain the mainstay of therapy for PD. However, some motor symptoms (e.g., gait problems) and a broad range of non-motor symptoms (e.g., cognitive impairment and depression) are often unresponsive to dopaminergic drugs. It is because the pathology of PD is not limited to the loss of midbrain dopamine neurons. Recent in vivo human brain imaging studies have provided novel insights into the dysfunction of multiple non-dopaminergic systems in PD. I review positron emission tomography, single photon emission computed tomography, neuromelanin-sensitive magnetic resonance imaging, and functional magnetic resonance imaging studies demonstrating the parallel alteration of noradrenergic, serotonergic, and cholinergic systems in PD. These studies shed light on the relationship between neuromodulators and vulnerable cognitive, affective, and motor functions. I finally discuss open questions in the field. More human pharmacological imaging studies are needed to reach a mechanistic understanding of the non-dopaminergic modulation of human brains.


Introduction
Parkinson's disease (PD) is the second most common neurodegenerative disease with mixed motor and nonmotor symptoms. The pathological hallmark of PD involves the presence of Lewy body, progressive loss of dopamine neurons in the substantia nigra pars compacta, and consequent dopamine depletion in the nigrostriatal pathway. Some motor symptoms of the disease, including bradykinesia, rest tremor, and rigidity, can be alleviated through dopamine replacement therapy and functional neurosurgery (e.g., subthalamotomy, subthalamic nucleus deep brain stimulation). However, other motor symptoms (e.g., gait and balance problems) and a broad range of non-motor symptoms (e.g., cognitive impairment and dementia, mood disturbances, and sleep disorders) are frequently unresponsive to dopaminergic drugs and striatal dopamine neuron grafts (Politis et al. 2012). Certain motor and non-motor symptoms may even worsen after subthalamic nucleus deep brain stimulation (Alomar et al. 2017;Wang et al. 2016). The underlying mechanisms of these motor and non-motor symptoms are poorly understood. Mechanism-based therapeutic options do not exist.
Pathology of PD is not limited to dopaminergic dysfunction. Although the neurotoxin 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) is widely used to induce dopaminergic lesions in animal models, MPTP often fails to produce PD-like impairment in motor coordination without concurrent loss of other monoamines (e.g., noradrenaline, Rommelfanger et al. 2007). Parallel alterations in the noradrenergic, serotonergic, and cholinergic systems have been observed in post-mortem PD brains (Goldstein et al. 2011;Rinne et al. 2002). As Braak et al. (2003) pointed out, the pathological process of PD begins in the dorsal motor nucleus and progresses in an ascending fashion to the midbrain and forebrain. Noradrenaline neurons in the locus coeruleus and serotonin neurons in the raphe nuclei may even be affected before the substantia nigra dopamine neurons (Braak et al. 2006). Human brain imaging has improved the understanding of neuromodulatory systems in various brain disorders in the past two decades. It is superior in characterizing the functional role of each neuromodulator. In particular, it can demonstrate in vivo the progressive alteration of neuronal terminals and reveal how a specific neuromodulator regulates a given cognitive, affective, or motor function. I will review recent advances in the non-dopaminergic modulation of cognition, mood, and motor, with evidence from positron emission tomography (PET), single photon emission computed tomography (SPECT), neuromelanin-sensitive MRI (NM-MRI), and functional MRI (fMRI) studies.

Loss of locus coeruleus noradrenaline neurons
The locus coeruleus is a primary source of ascending noradrenergic innervations in the central nervous systems of humans and other mammalians. In post-mortem PD brains, the locus coeruleus neurons show severe morphological changes, including swollen cell bodies containing Lewy body and reduced branched dendrites (Chan-Palay and Asan 1989). There is a mean loss of 83% locus coeruleus neurons (Zarow et al. 2003), resulting in a substantial reduction of noradrenaline levels in the frontal cortex (-50%) and striatum (putamen, -72%, caudate nucleus, -60%) (Goldstein et al. 2011).
The structural integrity of locus coeruleus can be measured in vivo using NM-MRI. Neuromelanin is a by-product of catecholamine synthesis, existing in locus coeruleus noradrenaline neurons and substantia nigra dopamine neurons (Fedorow et al. 2005;Zecca et al. 2001). In patients with PD, neuromelanin signals of the locus coeruleus and substantia nigra are remarkably diminished (Sasaki et al. 2006;Wang et al. 2018), consistent with the loss of noradrenaline and dopamine neurons in these nuclei (Isaias et al. 2016;Ito et al. 2017).

Noradrenaline and cognitive impairment
Cognitive impairment is a common non-motor symptom in PD. About 36% of de novo patients with PD exhibit mild cognitive impairment (Broeders et al. 2013;Foltynie et al. 2004). As the disease progresses, 46% of patients with PD develop dementia in ten years (Williams-Gray et al. 2009. Eventually, 83% of patients with PD develop dementia if they survive twenty years (Hely et al. 2008). A long-held debate concerns whether dopamine or noradrenaline (or both) drives cognitive impairment at early stages of PD.
Executive dysfunction occurs as the initial manifestation of cognitive impairment in PD. Executive function is a set of general-purpose control mechanisms that regulate thoughts and actions. There are three core executive functions: inhibition of prepotent responses, updating of working memory representations, and shifting between tasks or mental sets (Miyake and Friedman, 2012;Miyake et al. 2000). Noradrenaline has been linked to response inhibition and attentional shifting (Robbins 2007), whereas dopamine has been linked to working memory updating (Cools and D'Esposito 2011;Liu et al. 2021).

Noradrenaline and response inhibition
The ability to suppress prepotent responses (response inhibition) is a core executive function impaired in PD (Di Caprio et al. 2020;Gauggel et al. 2004). Response inhibition is thought to be supported by fronto-basal ganglia loops, including the fronto-subthalamic hyperdirect and fronto-striatal indirect pathways (Jahanshahi et al. 2015;Nambu et al. 2002). It is often measured as the stop-signal reaction time (SSRT) derived from the stop-signal task (Logan and Cowan 1984;Verbruggen et al. 2019).
Preclinical evidence from animal and human pharmacological studies suggests a crucial role of noradrenaline in response inhibition. In rats, the selective noradrenaline reuptake inhibitor atomoxetine can produce a dosedependent speeding of SSRT (Bari et al. 2009(Bari et al. , 2011aRobinson et al. 2008b). In contrast, the α2 adrenergic agonist guanfacine can prolong SSRT (Bari et al. 2011b). Neither selective serotonin nor dopamine reuptake inhibitors significantly impact response inhibition. In healthy adults and patients with PD, oral administration of atomoxetine can speed SSRT (Chamberlain et al. 2006;O'Callaghan et al. 2021), which is associated with enhanced stopping-related activity over the right inferior frontal gyrus (Chamberlain et al. 2009;Ye et al. 2015).
Our recent brain imaging studies confirm the role of the locus coeruleus noradrenergic system in response inhibition. First, we showed the therapeutic effect of atomoxetine on response inhibition in patients with PD using pharmacological fMRI (Ye et al. 2015. The study used a double-blind, randomized crossover design. Thirty-four patients with PD completed the stop-signal task during fMRI scanning after a single oral dose of 40-mg atomoxetine or placebo. Thirty-eight matched healthy control subjects completed the stop-signal task during fMRI without drugs. Under placebo, patients showed a selective impairment in response inhibition, reflected as longer SSRT, lower stopping-related activity over the right inferior frontal gyrus and pre-supplementary motor area, and weaker functional connectivity between the right inferior frontal gyrus and right striatum. Two hours after the intake, atomoxetine significantly enhanced the stopping-related right inferior frontal activity and strengthened the functional connectivity between the right inferior frontal gyrus, right striatum, and right subthalamic nucleus, leading to shorter SSRT and improved response inhibition in patients with PD Ye et al. 2015). The therapeutic effect was more prominent in patients with more severe motor symptoms .
Second, we showed that damage to the locus coeruleus might regulate response inhibition by modulating the fronto-subthalamic hyperdirect pathway in patients with PD using NM-MRI and fMRI (Xu et al.: Locus coeruleus integrity correlates with inhibitory functions of the fronto-subthalamic 'hyperdirect pathway' in Parkinson's disease, unpublished). The cross-sectional study included 29 patients with PD and 29 matched healthy control subjects. All participants completed an NM-MRI measurement of locus coeruleus integrity and an fMRI scanning with the stop-signal task. In healthy control subjects, locus coeruleus integrity correlated with the stoppingrelated activity of the right inferior frontal gyrus and subthalamic nucleus, which further correlated with SSRT. Patients showed significantly reduced locus coeruleus integrity, longer SSRT, and lower stopping-related right inferior frontal activity than healthy control subjects. The relationship between SSRT and the fronto-subthalamic hyperdirect pathway was preserved in most patients. However, locus coeruleus no longer correlated with the stopping-related activity over the right inferior frontal gyrus or the subthalamic nucleus. That is, locus coeruleus might modulate inhibitory functions of the fronto-subthalamic hyperdirect pathway in healthy brains. Damage to the locus coeruleus might impact the fronto-subthalamic hyperdirect pathway, leading to response disinhibition in PD.

Noradrenaline and attentional shifting
Noradrenaline may also play a role in attentional shifting (Robbins 2007). For example, Li et al. (2019) showed that locus coeruleus integrity correlated with shifting functions in de novo patients with PD. The cross-sectional study included 23 patients with mild cognitive impairment, 48 patients without cognitive impairment, and 32 healthy control subjects. All participants completed an NM-MRI measurement of locus coeruleus integrity and a comprehensive battery of neuropsychological tests. In particular, attentional shifting was measured as the completion time of the Trail Making Test B (TMT-B). Patients with mild cognitive impairment showed a significant reduction in the locus coeruleus integrity than healthy controls subjects. Across all patients (with and without cognitive impairment), locus coeruleus integrity negatively correlated with TMT-B completion time. Patients with better locus coeruleus integrity performed better in the shifting task. This observation has been confirmed by another NM-MRI study from Prasuhn et al. (2021). However, the exact mechanisms remain unclear on how the locus coeruleus noradrenergic system modulates attentional shifting. Future research is needed to address this issue systematically.

Noradrenaline and depression
Depression is common in PD: the frequency increases from 13% in de novo patients (Ravina et al. 2007;Wu et al. 2015) to 37-50% in medicated patients as the disease progresses (Maeda et al. 2017;Martinez-Martin et al. 2007). Even though the D2/D3 receptor agonist pramipexole has shown an antidepressive effect in patients with PD (Barone et al. 2010), the pathophysiology of PD depression is still poorly understood. Current human brain imaging evidence is against the simple serotonin hypothesis of depression (see the section 3.2 "serotonin and depression").
It is recently proposed that the combined depletion of dopamine, noradrenaline, and serotonin is underlying PD depression (Delaville et al. 2011;Maillet et al. 2016). In an 11 C-RTI-32 PET study, Remy et al. (2005) measured the loss of subcortical noradrenergic and dopaminergic terminals in 8 PD patients with depression and 12 PD patients without depression. 11 C-RTI-32 has similarly high affinities to dopamine and noradrenaline transporters but not serotonin transporters. Non-depressed patients showed less 11 C-RTI-32 binding than healthy control subjects in the substantia nigra, putamen, caudate nucleus, and ventral striatum, confirming the characteristic dopamine depletion in the nigrostriatal pathway. However, depressed patients showed less 11 C-RTI-32 binding than non-depressed patients in the locus coeruleus, amygdala, ventral striatum, and thalamus, suggesting a role of locus coeruleus noradrenergic projections in depressive symptoms. This study represents one of the initial attempts to test the new hypothesis. The role of each monoamine in regulating mood has yet to be specified.

Noradrenaline and gait
Outcomes of dopaminergic drugs for gait and balance problems are mixed. In particular, freezing of gait is unresponsive to dopaminergic drugs (Grimbergen et al. 2009). A recent 18 F-FMT PET study suggested that noradrenergic dysfunction may contribute to the pathophysiology of freezing of gait (Ono et al. 2016). This cross-sectional study measured the aromatic L -amino acid decarboxylase (AADC) activity in 30 PD patients with freezing of gait, 10 PD patients without freezing of gait, and 11 healthy control subjects. The AADC activity in the locus coeruleus negatively correlated with the severity of freezing of gait, regardless of disease duration and the severity of other motor symptoms.
However, evidence from open-label pharmacological studies in PD patients with freezing of gait was less consistent. Toghi et al. (1993) showed that droxidopa, a synthetic amino acid precursor of noradrenaline, could increase central noradrenaline levels and improve the freezing of gait in 7 out of 13 patients with advanced PD. Nutt et al. (2004) reported that a high dose of the noradrenaline and dopamine reuptake inhibitor methylphenidate (0.4 mg/kg) could increase walking speed and time by 20% when combined with a two-hour levodopa infusion (1.0 mg/kg/h). Devos et al. (2007) found that a three-month treatment of higher dose methylphenidate (1 mg/kg) could improve gait symptoms (e.g., shorter total times and fewer steps in the Stand-Walk-Sit test, and higher scores in the Tinetti Scale) in PD patients on subthalamic nucleus deep brain stimulation and off dopaminergic medication. Pollak et al. (2007) reported the effect of low dose methylphenidate on parkinsonism gait in an '8 trajectory' test (walking around two chairs in a figure of 8).
The beneficial effect of methylphenidate on gait control might be due to an interaction between dopaminergic and noradrenergic systems rather than the noradrenergic system alone. When the selective noradrenaline reuptake inhibitor atomoxetine was applied to patients with advanced PD, no effect was observed on the freezing of gait (Revuelta et al., 2015). Noradrenaline may not regulate gait control directly. It is more likely other neuromodulators are engaged (see the section "acetylcholine and gait").

Loss of raphe nuclei serotonin neurons
The ascending serotonergic pathway to the forebrain and the descending serotonergic pathway to the brainstem nuclei and spinal cord arise from the rostral and caudal clusters of the raphe nuclei, respectively. In PD, Lewy neurites and Lewy bodies that contain an aggregated form of α-synuclein are present in the raphe nuclei. Braak et al. proposed that the caudal raphe serotonin neurons are affected at an early stage (stage 2, before the substantia nigra dopamine neurons), and the rostral raphe serotonin neurons are affected at a later stage (stage 4, after the substantia nigra dopamine neurons but before the neocortex) (Braak et al. 2006(Braak et al. , 2003. Recent PET studies do not fully support the Braak staging hypothesis. Instead, the PET studies showed that both the caudal and rostral clusters of the raphe complex were relatively preserved until later stages of the pathological process (Albin et al. 2008;Politis et al. 2010a). For example, Politis et al. (2010a) assessed the integrity of presynaptic serotonergic terminals in PD patients at early (disease duration 0-5 years, mean Hoehn-Yahr stage 1.9), established (disease duration 6-10 years, mean Hoehn-Yahr stage 3.1), and advanced stages (disease duration 10 + years, mean Hoehn-Yahr stage 3.4) using the serotonin transporter marker 11 C-DASB. A progressive nonlinear pattern of serotonin neuron loss was observed. Early-stage patients showed a marked reduction of DASB binding in the caudate nucleus, thalamus, hypothalamus, and anterior cingulate cortex. Established-stage patients showed an additional reduction of DASB binding in the putamen, insular cortex, prefrontal cortex, and posterior cingulate cortex. Advanced-stage patients showed further decreased DASB binding in the caudal and rostral raphe nuclei, ventral striatum, and amygdala. One possibility is that α-synuclein aggregation begins in the raphe nuclei at early stages but does not affect the function of serotonin neurons until later stages. Further research is needed to understand precisely how α-synuclein aggregation leads to the loss of physiological functions of serotonin neurons in the raphe nuclei.
The degeneration of serotonin neurons may not be influenced by chronic exposure to dopaminergic drugs or the replacement of striatal dopamine neurons. For example, Politis et al. (2010a) found no correlation between DASB binding in the raphe nuclei and individual patients' lifetime total dopaminergic medication intake in medicated patients with PD. Politis et al. (2012) found a profound reduction of DASB binding in the raphe nuclei, prefrontal cortex, and posterior cingulate cortex in three PD patients who received intrastriatal transplantation of dopamine-rich fetal mesencephalic tissue 13-16 years previously. In contrast, these patients showed a normal 18 F-Dopa uptake in the striatum. That means the loss of serotonin neurons continues in the raphe nuclei and regions receiving ascending serotonin projections after successful transplantation of fetal dopamine neurons.

Serotonin and depression
The serotonin hypothesis of depression has provided an impetus to numerous studies on the pathophysiology of depression since the 1960s. In its original (and simplistic) form, the serotonin hypothesis postulates that a lowering of brain serotonin is responsible for depression. If the serotonin hypothesis holds in PD, one would expect a decreased density of pre-synaptic serotonin reuptake transporters and compensatory up-regulation or supersensitivity of post-synaptic serotonin receptors.
A simple serotonin deficiency could hardly explain PD depression. In a pilot 11 C-WAY 100635 PET study, Doder et al. (2003) found a mean reduction of 27% 5-HT 1A receptor availability in the raphe nuclei in patients with PD compared to healthy control subjects, confirming the loss of raphe nuclei serotonin neurons. However, the authors observed no difference in the 5-HT 1A receptor availability between depressed and non-depressed patients with PD. This observation was consistent with findings from depressed PD patients who had not received any antidepressants. In a 123 I-β-CIT SPECT study, Kim et al. (2003) found the density of midbrain serotonin transporters in depressed patients with PD was similar to that in healthy control subjects. Moreover, there was no correlation between the density of midbrain serotonin transporters and the severity of depressive symptoms.
More counterevidence comes from recent PET studies (Ballanger et al. 2012;Boileau et al. 2008;Politis et al. 2010b). Several 11 C-DASB PET studies consistently showed a global increase of serotonin transporter binding in the caudal raphe nuclei, limbic system, and prefrontal cortex in depressed PD patients compared with non-depressed PD patients and healthy control subjects. The increase of serotonin transporter binding correlated with the severity of depressive symptoms. Moreover, an 18 F-MPFF PET study showed a lower availability of postsynaptic 5-HT 1A receptors in the limbic system in PD depressed patients compared with non-depressed PD patients.
Consistent with the molecular imaging results, the latest Movement Disorder Society Evidence-Based Medicine Review addressed conclusions with caution regarding using selective serotonin reuptake inhibitors for treating depression or depressive symptoms in PD. Only the serotonin and noradrenaline reuptake inhibitor venlafaxine is considered effective and clinically useful. Selective serotonin reuptake inhibitors, including citalopram, sertraline, paroxetine, and fluoxetine, are regarded as having insufficient evidence for their efficiency (Seppi et al. 2019(Seppi et al. , 2011.

Serotonin and cognitive impairment
The serotonergic system strongly connects with a specific type of response inhibition, i.e., the prevention of a prepotent action (action restraint or NoGo inhibition). For example, rats with serotonin reuptake transporter knockout had lower levels of central serotonin and its primary metabolite 5-hydroxyindoleacetic acid. These rats showed less premature response than wild-type rats (Homberg et al. 2007). In rats and mice, blocking 5-HT 2A receptors in the nucleus accumbens can slow responding speed and reduce premature responding. In contrast, blocking 5-HT 2C receptors can increase response speed and reduce response accuracy, resulting in more premature responses (Fletcher et al. 2007;Robinson et al. 2008a). In healthy adults, oral administration of the selective serotonin reuptake inhibitor citalopram can amplify the NoGo-related activity over the right inferior frontal gyrus (Del-Ben et al., 2005;Macoveanu et al., 2013;Völlm et al. 2006), even though behavioral effects are often absent at the group level and may depend on individual differences in the serotonergic system (e.g., a functional promoter polymorphisms of the serotonin transporter gene or the baseline density of central 5-HT 2A receptors) (Fischer et al. 2015;Macoveanu et al. 2013).
We recently demonstrated the therapeutic effect of citalopram on response inhibition in patients with PD using pharmacological fMRI (Ye et al. 2014. The study used a double-blind, randomized crossover design. Thirty-four patients with PD completed the Go-NoGo task during fMRI scanning after a single oral dose of 30-mg citalopram or placebo. Citalopram reduced NoGo errors in proportion to disease severity, accompanied by enhanced NoGo-related activity over the right inferior frontal gyrus. PD patients with more severe motor symptoms responded better to citalopram. The relationship between serotonin and other types of response inhibition (e.g., stop-signal inhibition) is relatively weak. Although we observed a positive effect of citalopram on SSRT , such an effect was not present in other human pharmacological studies using citalopram (Drueke et al. 2010;Nandam et al. 2011), acute tryptophan depletion (Clark et al. 2005) or the 5-HT 1A receptor agonist buspirone (Chamberlain et al. 2007).

Serotonin and tremor
Tremor is a core motor symptom of PD that has been linked to the loss of raphe nuclei serotonin neurons. In a 11 C-WAY 100635 PET study, Doder et al. (2003) found a correlation between the availability of 5-HT 1A receptors in the raphe nuclei and the severity of resting tremor rather than the severity of postural tremor, bradykinesia, or rigidity, suggesting a post-synaptic mechanism of resting tremor. In a recent 11 C-DASB PET study, Loane et al. (2013) found a reduction of serotonin transporters in the raphe nuclei, caudate nucleus, and putamen in tremorpredominant PD patients compared with akinetic-rigid PD patients and healthy control subjects. The reduction of raphe nuclei serotonin transporters correlated with the severity of postural tremor, rather than the severity of resting tremor or total tremor, in tremor-predominant PD patients, suggesting a pre-synaptic mechanism of postural tremors (Politis et al. 2010a). In addition, in a 123 I-β-CIT SPECT study, Caretti et al. (2008) investigated the availability of thalamic serotonin transporter in de novo PD patients with moderate to severe tremors and those without tremors. PD patients with moderate to severe tremors showed lower thalamic serotonin transporter levels than those without tremors.

Altered pre-synaptic and post-synaptic cholinergic functions
Acetylcholine closely interacts with the striatal dopaminergic system. Striatal dopamine release can be triggered not only by the ascending activity of substantia nigra dopamine neurons but also by the activity of striatal cholinergic interneurons (Threlfell et al. 2012;Wang et al. 2014). After acute dopamine depletion, striatal cholinergic interneurons exhibit higher excitability in vitro recordings (Fino et al. 2007) and stronger synchronized oscillatory activity in vivo recordings (Raz et al. 2001).
In rat models of PD, dopamine depletion also leads to an anatomical reorganization of cholinergic synaptic connections, shifting cholinergic innervations towards striatopallidal neurons than striatonigral neurons, possibly via modulating Kir2 channels of striatopallidal neurons (Salin et al. 2009;Shen et al. 2007). PET and SPECT imaging of vesicular acetylcholine transporter (VAChT, a transporter that loads acetylcholine into secretory organelles), acetylcholinesterase (AChE, an enzyme that catalyzes the breakdown of acetylcholine), muscarinic acetylcholine receptor (mAChR) and nicotinic acetylcholine receptor (nAChR) has shed light on the pathological change of pre-synaptic and post-synaptic cholinergic functions in PD. In an early PET study, Kuhl et al. (1996) investigated the integrity of cholinergic terminals in PD patients with and without dementia, using the VAChT marker 123 I-IBVM. Nondemented PD patients only showed a regional decrease of IBVM binding in the parietal and occipital lobes, whereas demented PD patients showed a global cortical decrease of IBVM binding, especially in the frontal cortex and posterior cingulate cortex, similar to patients with earlyonset Alzheimer's disease. This finding was confirmed by PET studies using 11 C-MP4A, which labels AChE in cholinergic axons (Bohnen et al. 2003;Klein et al. 2010). Moreover, Klein et al. (2010) suggested that the cholinergic deficit profile of PD patients with dementia is similar to that of patients with Lewy Body Dementia.
Previous PET and SPECT studies have shown a global reduction of nAChR in PD, especially in the brainstem, striatum, and frontal cortex (10-25% reduction), using 123 I or 18 F labeled A-85380 (Fujita et al. 2006;Kas et al. 2009;Meyer et al. 2009;Oishi et al. 2007). However, some studies only observed a weak correlation between A-85380 binding and global cognition due to the small sample size (e.g., 10-13 patients), patient heterogeneity (e.g., H-Y staging 1-5), and lack of domain-specific measures of cognitive functions. Recent PET studies focusing on mild to moderate PD provided a different picture. Isaias et al. (2014) found that nAChR density decreased in the caudate nucleus and orbitofrontal cortex, and middle temporal gyrus but increased in the putamen, insular cortex and supplementary motor area in PD patients. The nAChR density of the ipsilateral putamen (relative to the clinically most affected hemi-body) positively correlated with disease duration. Lorenz et al. (2014) found that the nAChR density of the right superior parietal lobule, left thalamus, and posterior subcortical regions correlated with language and learning ability.
Different from nAChR, mAChR is up-regulated in PD. In an early 11 C-NMPB PET study, Asahina et al. (1998) showed that the cortical level of mAChRs selectively increased in the frontal and temporal lobes but not in the parietal or occipital lobes in non-demented patients with PD compared with healthy control subjects. But regional NMPB binding did not correlate with disease duration, disease severity, global cognition, or executive function. In a recent (R,R) 123 I-QNB SPECT study, Colloby et al. (2006) further reported an elevation of occipital mAChR level in PD patients with dementia and patients with Lewy Body Dementia. But again, regional QNB binding did not correlate with any scores of neuropsychological or neuropsychiatric tests.

Acetylcholine and dementia
Robbins and colleagues put forward an intriguing dual syndrome hypothesis that (i) the posterior cortical cholinergic dysfunction is responsible for early deficits in visuospatial function and semantic fluency in PD patients who exhibit gait problems and a rapid cognitive decline to dementia; (ii) the frontostriatal dopaminergic dysfunction is responsible for early deficits in working memory and executive functions in non-demented PD patients with mild cognitive impairment and a tremor-dominant phenotype (Kehagia et al. 2013;Robbins and Cools, 2014). The acetylcholine hypothesis of visuospatial function is supported by pharmacological studies in elderly healthy adults. For example, Wezenberg et al. (2005) found that visuospatial function, but not visual working memory, was impaired by the mAChR antagonist abiperiden and improved by the AChE inhibitor rivastigmine. But the cholinergic basis of semantic fluency needs further investigation, given conflicting findings in the literature (Aarsland et al. 1994;Pompéia et al. 2002;Thienel et al. 2009).
Currently, AChE inhibitors such as rivastigmine are the primary and most effective drugs for treating PD dementia, Lewy Body Dementia, and Alzheimer's disease, especially for mild to moderate dementia (O'Brien et al. 2017). These drugs prolong the activity of acetylcholine in the synaptic cleft and thus indirectly enhance acetylcholine transmission. However, the exact mechanisms of how rivastigmine enhances cognitive functions are still unclear. A recent resting-state fMRI showed that rivastigmine restored the spontaneous brain activity in the left inferior frontal gyrus, and pre-supplementary motor area in PD patients with dementia or mild cognitive impairment. In particular, the rivastigmine-induced activity change in the left inferior frontal gyrus correlated with attention performance (Possin et al. 2013).

Acetylcholine and gait
Recent 11 C-PMP PET studies of Bohnen and colleagues focused on the gait-related loss of cortical cholinergic terminals (Bohnen et al. 2014(Bohnen et al. , 2013. They distinguished between hypo-cholinergic PD patients (with a PMP binding below the normal range) and normal cholinergic PD patients. Hypo-cholinergic PD patients showed slower gait speed and more freezers than normal cholinergic PD patients. Normal cholinergic PD patients were similar to healthy control subjects in gait. In another retrospective study, Bohnen and colleagues identified clinical markers of cholinergic denervation in PD patients (Müller et al. 2015). They recommended the combination of rapid eye movement sleep behavior disorder symptoms and fall history for predicting combined thalamic and cortical cholinergic deficits, and the combination of 8.5-m walk time and Montreal Cognitive Assessment score for predicting isolated cortical cholinergic deficits.

Open questions and future directions
Non-dopaminergic dysfunction in the central nervous system of PD patients has not received much attention as the loss of midbrain dopamine neurons until recently. Recent PD studies with in vivo brain imaging have confirmed the pathological alteration of noradrenergic, serotonergic, and cholinergic innervations in the midbrain and forebrain. Some studies suggest that the loss of neuronal function happens many years after the aggregation of α-synuclein. The progression of functional loss may be a consequence of, but not equal to, the spreading of Lewy body pathology. Further studies are needed to understand how exactly α-synuclein misfolding leads to the loss of neuronal function in PD brains.
Non-dopaminergic dysfunction has been linked to non-motor and motor symptoms that are frequently unresponsive to dopamine replacement therapy and striatal dopamine neuron transplantation. In the motor domain, noradrenaline and acetylcholine relate to gait control, and serotonin relates to resting and postural tremors. In the cognitive domains, recent brain imaging studies go beyond the dual syndrome hypothesis, suggesting that fronto-subcortical noradrenaline and serotonin deficiency are responsible for executive dysfunction in non-demented PD patients. But the relationship between posterior cortical cholinergic dysfunction and visuospatial function or semantic fluency needs further investigation. Evidence from recent brain imaging studies is against a simple serotonin theory of PD depression but implies a potential role of the locus coeruleus noradrenergic system. Our current understanding of the biochemistry of cognitive, affective, and motor symptoms in PD mainly relies on a correlation between the loss of specific neurons and the severity of functional impairment. There is a need for properly designed pharmacological imaging studies to advance mechanistic understanding of how each neuromodulator regulates a given cognitive, affection, or motor function.
We did not discuss the functional role of histamine in PD due to the lack of in vivo brain imaging study. An early post-mortem study showed higher levels of histamine in the substantia nigra, putamen, and globus pallidus in PD patients compared with healthy control subjects (Rinne et al. 2002). A recent study further linked the increase of pallidum histamine H3 receptors with psychotic symptoms in patients with Lewy Body Dementia (e.g., delusion, visual hallucination) (Lethbridge and Chazot 2016). Visual hallucination is another frequent non-motor symptom in PD. The question of whether PD visual hallucination has a histaminergic basis or, as some researchers suggested, a cholinergic basis, is open for future studies (Yarnall et al. 2011).
Non-dopaminergic systems interact closely with the dopaminergic system in the striatum. The activity of striatal dopamine terminals is not only modulated by cholinergic interneurons, as we discussed, but also by ascending noradrenergic and serotonergic projections from the midbrain (Haleem 2015;Szot et al. 2012). A monkey model with double dopaminergic and serotonergic lesions has recently been developed for PD (Beaudoin-Gobert et al. 2015), emphasizing the impact of the combined depletion of monoamines. Understanding the interaction between dopaminergic and non-dopaminergic transmissions is another challenge for future research.

Conclusions
Pathology of PD is not limited to the loss of substantia nigra dopamine neurons. In the pathological process of the disease, locus coeruleus noradrenaline neurons and raphe nuclei serotonin neurons are affected before midbrain dopamine neurons. It is not unexpected that some motor symptoms (e.g., gait and balance problems) and a broad range of non-motor symptoms (e.g., cognitive impairment and dementia, depression) are frequently unresponsive to dopaminergic drugs and striatal dopamine neuron transplantation. This review summarizes recent in vivo human imaging studies that show the parallel alteration of noradrenergic, serotonergic, and cholinergic systems in PD. In particular, the review highlights PET, SPECT, NM-MRI, and fMRI studies that characterize the functional role of each neuromodulater in cognition, mood, and motor, providing a bigger picture of neuromodulatory systems in PD.