The pathology roadmap in Parkinson disease

Abstract

An under-appreciated clue about pathogenesis in Parkinson disease (PD) is the distribution of pathology in the early and middle stages of the disease. This pathological ‘roadmap’ shows that in addition to dopaminergic neurons in the substantia nigra pars compacta (SNc), a significant number of other central and peripheral neuronal populations exhibit Lewy pathology, phenotypic dysregulation or frank degeneration in PD patients. This spatially distributed, at-risk population of neurons shares a number of features, including autonomously generated activity, broad action potentials, low intrinsic calcium buffering capacity and long, poorly myelinated, highly branched axons. Many, and perhaps all, of these traits add to the metabolic burden in these neurons, suggesting that mitochondrial deficits could drive pathogenesis in PD—in agreement with a large segment of the literature. What is less clear is how this neuronal phenotype might shape the susceptibility to proteostatic dysfunction or to the spread of α-synuclein fibrils deposited in the extracellular space. The review explores the literature on these issues and their translational implications.

Introduction

The loss of mesencephalic substantia nigra (SN) dopaminergic neurons in PD is responsible for its core motor symptoms.¹ However, a variety of other neurons exhibit signs of pathology in post-mortem analysis of PD patients. For example, intracellular protein aggregates in the dorsal motor nucleus of the vagus (DMV), a region within the medulla oblongata, are a consistent feature of the pathology in the brains of PD patients.^2,3These aggregates are known as Lewy bodies and Lewy neurites or Lewy pathology. Alpha-synuclein is a major component of Lewy pathology, enabling pathologists to use immunocytochemical approaches to map Lewy pathology in postmortem samples from PD patients.⁴ Although they have caveats, these studies have given us a “roadmap” of the PD trajectory that should inform our theories about pathogenesis. This review attempts to summarize this literature and to determine if there is a connection to mitochondrial dysfunction, which has long been thought to be a causative factor in PD.⁹⁹

Vulnerable Neuronal Populations in PD

Both central and peripheral nervous systems are affected in PD. The peripheral nervous system can be divided into three parts: sensory, motor and autonomic. Generally, peripheral sensory and motor neurons do not display Lewy pathology or signs of degeneration in PD patients. In contrast, there are clear signs of pathology in the autonomic nervous system. The autonomic nervous system has three major divisions: sympathetic, parasympathetic and enteric. All three have been reported to exhibit Lewy pathology and dysfunction in PD patients. For example, orthostatic hypotension occurs in a significant fraction of late stage PD patients. This symptom has been attributed to a sympathetic denervation of the heart or vasculature.^5-9 In some PD and dementia with Lewy bodies patients, Lewy pathology is present in the peripheral vagal nerve and ganglia.^10,11 These axons originate in the DMV. There are reports of strong Lewy pathology in the DMV preganglionic parasympathetic neurons of nearly all PD patients studied (see below). A subset of parasympathetic neurons in the intrinsic cardiac ganglia also appear to be at risk in PD.⁸ The parasympathetic inferior salivatory nucleus (IXth cranial nerve), which innervates the parotid gland, also features Lewy pathology in PD,^11,12 but it is unclear whether there is neuronal loss. In PD patients, Lewy pathology has been seen in both in the submandibular gland and the superior cervical ganglion.¹¹ Djaldetti and colleagues have reported a marked denervation of all autonomic neurites in skin,¹³ work supported by analyses of skin biopsies from PD patients.^5,14

Lewy pathology and dopaminergic neuron loss also has been found in the enteric nervous system of many PD patients, particularly the lower gastrointestinal tract, which might be responsible for decreased gastric motility and constipation.^15,16 The inevitability of enteric nervous system pathology in PD has been challenged, however.^17-19 These studies show that enteric nervous system Lewy pathology is a frequent, but not necessary concomitant of PD. Conversely, although gastrointestinal dysfunction is frequently found in PD patients,⁵ it is also common among aged individuals without any sign of PD.²⁰ Making matters worse, enteric nervous system neurons are very heterogeneous and attempts to identify the phenotype of those exhibiting Lewy pathology have not reached a consensus.^21,22

In summary, Lewy pathology is found in several types of peripheral neuron in PD. The only neurons that are well established to be lost in PD are noradrenergic neurons innervating the heart and skin. Lower intestine enteric nervous system neurons commonly display Lewy pathology in PD patients and this might be responsible for constipation that commonly accompanies PD.

In the central nervous system, the vast majority of the neurons lost or displaying signs of pathology in early and mid-state PD patients are found in the brainstem. In the brainstem, Lewy pathology and cell loss has been reported in the region of the DMV, the medullary reticular formation, the raphe nuclei, the locus coeruleus, the pedunculopontine nuclei, the substantia nigra pars compacta (SNc) and, to a lesser extent, the ventral tegmental area and retrorubral area. The evidence for the involvement of these nuclei will be discussed in turn.

James Parkinson theorized that damage to the medulla caused PD. Axons entering the vagus nerve from the DMV can have strong Lewy pathology²³; Lewy pathology within the DMV has been reported in most,¹² but not all PD patients.^24,25 Additional neurons in this region have some variable degree of Lewy pathology.¹² Neuronal counts demonstrate that DMV neurons are lost in PD,²⁶ but there are differences in the literature regarding which neurons are lost, probably due to differences in stages of the disease. Classical studies report neuromelanin containing neurons in the DMV region were lost in PD brains, including the original report,²⁷ which states that “The pigmented cells of the dorsal vagal nucleus also degenerated, often with vacuolation, by contrast with the non-pigmented cells in this nucleus which remained healthy.” Eadie did not confirm this and reported that cholinergic motoneurons were lost, consistent with a later study.²⁶ Braak and collaborators suggest that the preganglionic parasympathetic projection neurons are first to degenerate, and neuromelanin neurons might be lost later.²⁸ These might not be distinct neuronal populations, however. New work suggests that some cholinergic DMV neurons also express tyrosine hydroxylase and aromatic acid decarboxylase, the enzymes that produce dopamine; while these neurons might synthesize catecholamines, they lack detectable vesicular monoamine transporter, making it unlikely they release it.²⁹

The loss of dopaminergic neurons in the SNc is the best-documented sequela of PD. The loss of neuromelanin in the SNc of PD patients was reported early in the 20th century³⁰ and confirmed by many subsequent studies.^27,31-33 The loss of neuromelanin is seen in all PD patients,³⁴ but not all patients with parkinsonism.³⁵ Parkinsonism is a bradykinetic syndrome that can include rigidity, tremor and postural instability. PD is the most common cause of parkinsonism, but progressive supranuclear palsy and multiple system atrophy also fall into this category. Lewy pathology is nearly always found in the neuromelanin-positive neurons in the SNc of PD patients, particularly in the posterolateral regions.¹² Some Lewy pathology has been reported in the neighboring ventral tegmental area and retrorubral field neurons as well,⁵³ but neuronal loss in these two regions is variable.^36,37 Profound neuronal loss in the SNc appears to be specific to parkinsonian etiologies¹¹⁹ and not Alzheimer disease.³⁸ The most heavily neuromelanin pigmented neurons that are lost in the SNc of PD patients appear to be those with lower levels of vesicular monoamine transporter type 2 and with decreased vesicular accumulation of dopamine as well as low levels of calbindin.^39-41

Locus coeruleus neurons have long been known to be lost in PD.^27,33 Lewy pathology is observed in neuromelanin pigmented locus coeruleus neurons and is absent from non-neuromelanin cells in the complex,³⁴ confirming selective damage to noradrenergic neurons. Virtually all PD patients appear to have a substantial loss of locus coeruleus neurons, with a mean neuronal loss of 83% in later stages of the disease.³⁸ There is also an average 68% loss of these neurons in Alzheimer disease, but the extent of locus coeruleus loss is more variable in Alzheimer disease patients than PD patients.³⁸

Neurons in the raphe nuclei are responsible for serotoninergic innervation of the CNS. In incidental Lewy body disease and PD, Lewy pathology is most apparent in medium-sized neurons of the caudal raphe nucleus.¹² The ventrorostral raphe region around the medial lemniscus also exhibits Lewy pathology in PD.³⁴ Raphe nuclei neurons not only display Lewy pathology, but are lost in PD, with a reduction of over half these neurons in the median raphe and a somewhat smaller fractional loss reported in the raphe obscurus.^34,42

The pedunculopontine nucleus is rostral to the locus coeruleus and includes a mixture of cholinergic, glutamatergic and GABAergic neurons with an array of targets in the mesencephalon and diencephalon.⁴³ The pedunculopontine nucleus was reported to exhibit Lewy pathology by Braak’s group^4,28. Lewy pathology and neuronal loss also were reported.^44,45 Although prominent in PD, pedunculopontine nucleus loss is not specific to PD and is also seen in progressive supranuclear palsy and Alzheimer disease.^42,44,45

There is also substantial pathology in regions rostral to the brainstem relatively early in PD. One of these is the olfactory system. In the olfactory bulb, mitral cells exhibit Lewy pathology, but this is also seen in Alzheimer disease.^46,47 The anterior olfactory nucleus also shows Lewy pathology in incidental Lewy body disease.¹² The existence and targeting of neuronal loss in the olfactory system is controversial.^48,49 This issue is complicated by the neurogenesis in this region.⁵⁰

There are several other regions that have been reported to have either neuronal loss or Lewy pathology in PD, many in single studies. These include the lateral hypothalamus⁵¹ and the intralaminar nuclei of the thalamus.⁵² The specificity of these changes is unclear. In the case of the thalamic nuclei, the loss is also seen in progressive supranuclear palsy.⁵² In apparent later stages of PD, Lewy pathology is also scattered throughout the cerebral cortex, amygdala and hippocampus.^53,54 Retinal dopaminergic also has been reported to decline, but Lewy pathology was not identified.⁵⁵

The nucleus basalis of Meynert is a prominent site of Lewy pathology and cell loss in PD patients. The specificity of the pathology is less clear as the nucleus basalis of Meynert also figures prominently in Alzheimer disease.⁵⁶ This issue was addressed by a comparative study that found a loss of about half of the nucleus basalis of Meynert neurons measured by cresyl violet label in 11 PD patients, which was significantly greater than age-matched control subjects.⁵⁷ Not all of these subjects displayed Alzheimer disease-like symptoms, suggesting that there was a symptomatic threshold. There was no correlation between PD symptom severity and nucleus basalis of Meynert loss in this study. The most extensive comparative study of PD and Alzheimer disease brains in this region found similar neuronal losses, also as labeled by cresyl violet, on average (~40%), but there was a great deal of variability.³⁸ Thus, Lewy pathology and neuronal loss in the nucleus basalis of Meynert appears to be a common feature of PD and Alzheimer disease.

Determinants of vulnerability

The obvious question is what links this seemingly diverse set of neurons. There are several possible phenotypic traits that have been proposed to underlie vulnerability.

_• A common reactive neurotransmitter. SNc, locus coeruleus, raphe nuclei, enteric dopaminergic neurons and sympathetic postganglionic neurons each synthesize a monoamine neurotransmitter and high levels of cytosolic monoamines are reactive and hypothesized to underlie selective neuronal death under several conditions.^58-62 Although pre-ganglionic DMV neurons are nominally cholinergic, some of these neurons appear to possess and/or secrete monoamines. SNc and locus coeruleus are the most similar in this regard and both have neuromelanin deposition in humans. These two nuclei exhibit the greatest loss in PD. Nevertheless, some highly pigmented A2 catecholaminergic, neuromelanin-positive neurons in the caudal medulla do not appear to be lost in PD,³² and so neuromelanin synthesis per se, which may be a neuroprotective response,^63,64 is not sufficient for neuronal death. In the periphery, the loss of NE neurons is variable, with some sympathetic neurons appearing to be vulnerable and others not. Moreover, their loss is not specific to PD, as is also the case for locus coeruleus neurons.

However, there are two arguments against the reactive monoamine transmitter hypothesis of PD. One is that there clearly is pathology and loss of neurons that do not use monoamines. Vulnerable neurons in the DMV, pedunculopontine nucleus, LH, nucleus basalis of Meynert and enteric nervous system do not use a monoamine transmitter. It is true that neuronal loss in at least some of these regions is known not to be specific to PD. But the same is true of neuronal loss in locus coeruleus. It has been recently suggested that loss of cholinergic neurons is a consequence of the preceding loss of monoamine neurons,¹²⁰ but this conjecture awaits compelling support. The other key argument is that use of L-DOPA—the precursor for dopamine—to treat PD does not accelerate the progression of the disease as one would expect if dopamine or noradrenaline was the toxic agent in the disease¹_.

_• A long, highly branched axon with multiple release sites. SNc, locus coeruleus, raphe nuclei, pedunculopontine nucleus, DMV and nucleus basalis of Meynert neurons all have unusually long highly branched axons that are unmyelinated or thinly myelinated.^7,65 This feature is particularly well documented for SNc dopaminergic neurons. Single SNc axons terminating in the striatum are highly branched and possess as many as several hundred thousand synaptic release sites.⁶⁶ This is an order of magnitude greater than most neurons that have been carefully studied. Interestingly, these terminals do not appear to have an elevated mitochondrial oxidant stress.⁶⁷ However, maintaining a massive terminal field is very likely to create a metabolic and proteostatic burden on the cell body. Mitochondrial trafficking could prove particularly problematic; in fact, mitochondrial density in the somatodendritic region of SNc dopaminergic neurons is low⁶⁸ possibly reflecting the need to traffic mitochondria to axons. It is worth noting that α-synuclein is a presynaptic regulator of synaptic vesicle exocytosis^69-71; the proteostatic burden it creates could scale with the number of synaptic release sites and may contribute to mishandling of presynaptic mitochondria.⁷² It is unclear whether ventral tegmental area and retrorubral field dopaminergic neurons have as extensive an axonal field as do SNc dopaminergic neurons. Matsuda et al. did not report profound differences in dorsal and ventral striatal terminal fields, which should correspond to SNc and ventral tegmental area; as a consequence, the differences in vulnerability between these regions would have to be explained by other factors. Neurons of the locus coeruleus also have very long and complex projections. Based upon the distance traveled and terminal field, the axons of DMV neurons are also long and highly branched, and many enteric nervous system neurons are also highly branched.

• A common physiological phenotype. An extended discussion of this hypothesis has recently been published.⁷³ PD is a disease of neurons, not of the liver, kidney or heart. An implication of this fact is that one or more of the features distinguishing neurons from these other cell types must contribute in a seminal way to pathogenesis. A cardinal feature of neurons that separates them from nearly all other cell types is excitability. Neurons use steep electrochemical gradients across their plasma membrane to perform computations on incoming chemical signals from other neurons and to pass the outcome of this computation to other cells. Each step in this process expends energy. Action potentials (or spikes) and synaptic transmission dissipate the ionic gradients for sodium, potassium, calcium and chloride that are maintained by adenosine triphosphate dependent pumps and exchangers. Although all neurons share this basic set of properties, the parameters of spikes and synaptic transmission vary dramatically. The physiological phenotype of neurons ranges from what might be called a “wallflower” or quiescent phenotype to a “chatter box” phenotype that never stops spiking. SNc, locus coeruleus, raphe nuclei, nucleus basalis of Meynert, pedunculopontine nucleus and DMV neurons all fall into the chatterbox phenotype. That is, all of them spike continuously in vivo during the waking state.^74-81 SNc, locus coeruleus, DMV and pedunculopontine nucleus neurons are autonomous pacemakers (they spike on their own in the absence of synaptic input).

One particularly expensive ion that enters neurons during spiking is calcium. It is metabolically expensive because it must be pumped out of the cell against a much steeper (~2000 fold) electrochemical gradient than sodium, potassium or chloride ions. Most neurons keep this burden to a minimum by restricting calcium entry to the brief period during spikes, keeping these spikes short in duration and by expressing specialized calcium binding proteins that effectively buffer calcium—“grabbing” it after it enters and keeping it in place for plasma membrane pumps and transporters. This is not true of vulnerable neurons; they seem to do all the wrong things. For example, typically vulnerable neurons are continually spiking and have broad, slow spikes; this is certainly true of SNc, locus coeruleus, raphe nuclei, pedunculopontine nucleus and nucleus basalis of Meynert neurons (see references above). Although less well studied, many of the neurons in the autonomic nervous system, particularly those in the enteric nervous system, also are spontaneously active and have broad spikes.^82-84 The expression of calcium binding proteins is low in those vulnerable neurons that have been studied carefully. SNc, locus coeruleus and DMV neurons express relatively little of these calcium binding proteins^77,85 (raphe nuclei, pedunculopontine nucleus and nucleus basalis of Meynert have not been rigorously characterized in this regard to our knowledge). In contrast, most other, relatively PD-resistant autonomous pacemakers in the brain express a high level of calcium binding proteins (e.g., ventral tegmental area neurons, Purkinje neurons, globus pallidus neurons, striatal cholinergic interneurons).^86,87 The expression of known calcium binding proteins in the autonomic nervous system and enteric nervous system varies from cell type to cell type.^88-90 In addition, SNc and locus coeruleus neurons allow significant amounts of calcium to enter during the period between spikes.^91-93

The pacemaking phenotype characteristic of vulnerable neurons opens the door to another potential source of stress. Pacemaking neurons reside at relatively depolarized membrane potentials where NMDA receptors are relieved of their magnesium block, creating another point of sodium and calcium entry during excitatory synaptic transmission; excitotoxicity is one of the earliest theories of pathogenesis in PD.^94,95

In addition to creating a metabolic stress, activity-dependent elevation in cytosolic calcium levels increases cytosolic dopamine in SNc neurons, possibly due to an effect on synthesis.⁶⁰ Enhanced dopamine levels in the cytosol can lead to α-synuclein-dependent neuronal death.⁶⁰

In most neurons, the metabolic burden associated with activity and synaptic transmission is thought to significantly diminish the respiratory reserve of mitochondria.⁹⁶ In SNc dopaminergic neurons, and others of its kind, this reserve should be even smaller. In fact, there is a measurable increase in the oxidation of mitochondrial thiol proteins in SNc and DMV neurons that are simply pacemaking.^97,98 Unpublished work by our group (DJS) has revealed that locus coeruleus neurons have a similar mitochondrial oxidant stress. Mitochondrial dysfunction is widely viewed as a pivotal step in PD pathogenesis.⁹⁹ A sustained mitochondrial oxidant stress should in principle lead to the accumulation of mitochondrial DNA (mtDNA) mutations and impaired complex I function seen in the SNc with aging and PD.^100-102 Genetic mutations affecting mitochondria and environmental toxins that compromise mitochondrial respiration could synergize with this cell type specific stress, hastening bioenergetic failure and degeneration.

An interaction between neuronal phenotype and proteostatic dysfunction

The data summarized thus far is consistent with the hypothesis that vulnerable neurons have a distinctive physiology, leading to increased mitochondrial oxidant stress and susceptibility to insults that compromise mitochondrial function. Does this model provide an explanation of Lewy pathology and proteostatic dysfunction in these neurons? The answer is clearly no. However, it is self-evident that bioenergetic deficits could impair proteostatic function simply by diminishing the availability of adenosine triphosphate. Elevated cytosolic calcium levels seen in many vulnerable neurons could also play a direct role. For example, increasing calcium concentration, even transiently, increases the aggregation of α-synuclein.¹⁰³ Lysosomes and autophagic vacuoles that deliver intracellular components to lysosomes, key elements in the catabolic machinery, are dependent upon calcium signaling for their regulation¹⁰⁴ and are potential sites of ROS generation.⁶⁴ A number of genes linked to PD have effects on lysosomal function as well.¹⁰⁵ These observations raise the possibility that proteostatic challenges that increase lysosomal activity could exacerbate basal oxidant stress in vulnerable neurons, promoting degeneration.

Another hypothesis of pathogenesis that has received a great deal of attention lately is the so-called prion hypothesis.¹⁰⁶ In this model, α-synuclein fibrils in the extracellular space are taken up by neurons and these fibrils seed Lewy pathology; some subset of these fibrils are then released, spreading the pathology. Thus, α-synuclein fibrils behave in a prion-like way. The most compelling support for this hypothesis comes from: 1) the observation that dopaminergic neurons grafted into the brains of PD patients rapidly developed Lewy pathology¹⁰⁷ and 2) the demonstration that inoculation of brains with α-synuclein fibrils leads to a spreading Lewy-like pathology.^108,118 On the face of it, this model is very difficult to reconcile with the pattern of PD pathology. In the brain of a PD patient, Lewy pathology does not follow a nearest neighbor rule. For example in the caudal brainstem, neurons in the nucleus tractus solitarius never show Lewy pathology or degeneration in PD, but are near neighbors of DMV neurons and project axons to them. Synaptic connectivity per se also does not predict the pattern of Lewy pathology in PD. The spread of pathology reported by Lee’s group following injection of α-synuclein fibrils into the brain appears to be more like a wave than spread based upon the strength of synaptic connectivity.¹⁰⁸ Thus, if α-synuclein fibrils spread the pathology in PD, then there has to be a modifier of susceptibility that accounts for its non-uniform distribution. Could susceptibility to α-synuclein fibril “infection” be determined by physiological phenotype? One of the key features of these neurons is that they have robust elevations in cytosolic calcium concentration. Calcium regulates exocytosis and exosome-mediated protein release.^109-111 These are the presumptive mechanisms by which α-synuclein fibrils will be delivered and taken up. Neuronal exosomal release is increased by depolarization¹¹² and exosome-mediated release of α-synuclein in particular is calcium-dependent.¹¹³ Endocytosis, at least at nerve terminals, is also increased by calcium.¹¹⁴ The vast axonal arbor of vulnerable neurons could also serve as a potent conduit for the propagation of pathology. If activity-dependent elevation in cytosolic calcium is required for the spread of α-synuclein pathology, then the pattern of pathology becomes understandable. It also would mean that vulnerable neurons are assaulted from two directions: metabolic and proteostatic.

Therapeutic implications

The translational question is how to devise a therapeutic strategy to stop or slow the progression of PD. Although there are not clear strategies for altering the risk factors associated with transmitter choice or axonal arbor, there is a way in which the consequences of the chatterbox phenotype on mitochondrial stress could be diminished. One of the ion channels contributing to the basal metabolic stress in SNc, DMV and locus coeruleus neurons is the L-type calcium channel with a Cav1.3 pore-forming subunit. As mentioned above, antagonizing these channels diminishes mitochondrial oxidant stress in these neurons and lowers potentially damaging levels of cytosolic dopamine. Diminishing intracellular calcium levels could also lower the risk of α-synuclein aggregation and of taking up and passing on α-synuclein fibrils. There are FDA approved antagonists (dihydropyridines) of these channels that have an excellent safety record in humans. Moreover, there is epidemiological evidence that sustained use of dihydropyridines reduces the observed risk of PD.^115-117

Summary

The pathological “roadmap” created by Lewy pathology and neuronal loss shows that PD is far from just a disease of dopaminergic neurons in the SNc. The spatially distributed, at-risk population of neurons share a number of features, including autonomously generated activity, broad action potentials, low intrinsic calcium buffering capacity and long, poorly myelinated, highly branched axons. These features might render them more vulnerable to infection by prion-like α-synuclein fibrils and more likely to propagate it. Although this constellation of features is likely to drive pathogenesis through a number of parallel pathways, mechanisms driven by the opening of L-type calcium channels appear to be the most therapeutically accessible.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.