1. Introduction
Parkinson’s disease (PD) is the second most common neurodegenerative disease, after Alzheimer’s disease. PD is rare in individuals less than 50 years old, but its prevalence increases with age, reaching up to 4% in older-age groups [
1]. Clinically, PD presents with motor symptoms, e.g., rigidity, bradykinesia, tremor and postural instability [
2], and non-motor symptoms, e.g., depression, hyposmia, constipation and sleep disorders [
3]. The neuropathology of PD is characterized by the formation of Lewy bodies (abnormal intracellular aggregates containing metals, lipids and proteins like α-synuclein), a progressive loss of dopaminergic neurons that project from the substantia nigra (SN) to the dorsal part of the striatum (which consists of the caudate nucleus and putamen), and microgliosis [
4].
Current knowledge of the molecular mechanisms underlying these pathological hallmarks has been derived from studies on 27 monogenic familial forms of PD [
5], which represent only 5–10% of the cases, and from toxin-induced PD cellular and animal models (e.g., use of 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine, rotenone or 6-hydroxydopamine) [
6,
7,
8,
9]. The most salient mechanisms described up to now include mitochondrial dysfunction and oxidative stress, unfolded protein response, protein aggregation, lysosomal dysfunction and neuroinflammation [
10]. Unfortunately, the existing data has not resulted in the development of disease-modifying treatments for PD.
Recent studies have pointed towards a key role for lipids in PD [
11,
12,
13]. Based on their chemical and biochemical properties lipids can be classified into eight different classes, namely fatty acyls, glycerolipids, glycerophospholipids, sphingolipids, sterols, prenols, saccharolipids and polyketides [
14]. Lipids are mainly known for their role in energy storage [
15], but they are also involved in cellular signaling and transport [
16,
17]. Furthermore, lipids are the main constituents of cellular membranes [
18], and form part of membrane rafts and anchors [
19,
20]. As such, an abnormal lipid composition has been linked to the molecular mechanisms underlying PD, i.e., protein aggregation [
21], mitophagy [
22] and immune processes [
23].
Unbiased post mortem lipid profiling of brain tissue samples from PD patients and control individuals [
24] could be a rich source of hypothesis-generating findings and key to further understanding of PD etiology. Of note, the lipidome is less susceptible to postmortem changes than the water-soluble metabolome and therefore the lipid composition of brain sections may offer a reliable representation of PD-related alterations in lipid homeostasis in vivo. Furthermore, genome-wide mRNA expression profiling, also known as transcriptomics, is one of the most-used -omics techniques and may be employed to shed light on how specific lipid changes are coupled to transcriptional regulation. Studies analyzing the PD brain transcriptome have mostly involved the use of microarrays [
25]. Microarrays are more cost-effective than RNA sequencing (RNA-seq) [
26], but the latter methodology is superior at detecting splicing events, and novel and/or low-abundance transcripts [
27].
Here we perform lipid as well as RNA-seq analyses to determine the lipid and transcriptome profiles, respectively, of SN and putamen samples from the same PD patients and controls. Additionally, we compare our genome-wide mRNA expression findings to previous PD/control SN and putamen transcriptomic results to draw more robust conclusions about differential mRNA expression profiles in these PD brain regions.
4. Discussion
The data presented here represent the first lipid and transcriptome analyses performed on PD and control SN as well as putamen samples from the same individuals. We conducted a transcriptome analysis by RNA-seq rather than microarrays in order to identify not only annotated transcripts, but also currently unannotated sequences for future re-evaluation (50/354 SN DEGs and 29/261 putamen DEGs represent not-annotated transcripts) [
26]. Additionally, we analyzed our data together with publicly available transcriptome datasets to draw more robust conclusions about the relevance of gene functions and pathways for PD.
Lipidomics likely represents the most informative -omics technology complementary to RNA-seq. In this context, it is indeed important to realize that lipids are structural and bioactive molecules essential for multiple brain processes such as myelination, synaptic signal transduction and acting as second-messenger precursors and, relative to water-soluble metabolites, generally have a much slower turnover and are thus less prone to post-mortem degradation. Previously, neuropathological lipidomic studies have been conducted on primary visual cortex [
48], primary motor cortex [
49], hippocampus [
50] and cerebellum [
51] from PD and/or Lewy body disease individuals. Our exploratory lipid profiling involved the analysis of seven classes of lipids and showed several lipid species modulated in the SN and putamen of PD patients compared to age-matched controls. Two other studies have been performed on the lipid profile of the SN [
52,
53] and one on the putamen [
51] of PD patients, but they have reported only differences in lipid classes rather than lipid species and focused on sphingolipid and gangliosides, respectively, hindering a comparative analysis. The differences we found in the putamen include PI 34:2 and two saturated SM species, namely SM d18:1;14:0 and SM d18:1;16:0. Unfortunately, the limited current knowledge of putamen lipids does not allow a functional interpretation of the putamen lipid changes we detected. Lipids BMP 42:8 and PI 42:10, both with increased abundance in the SN of PD patients compared to controls, are thought to have arachidonic acid (20:4) as one of the two side chains, while the three decreased lipid species, PC 36:3, PE A36:2 and PS 36:3, are likely to have linoleic acid (18:1) as one of their two side chains. Although such specific changes in lipid composition are difficult to directly link to distinct enzymatic activities in specific metabolic conversion steps, it is tempting to speculate on their wider biological implications. The presence of the five differentially expressed SN lipid species could point towards a neuroinflammatory component in disease etiology, since increased arachidonic acid has been demonstrated in acute neuroinflammation [
54], while linoleic acid is one of the sources of arachidonic acid [
55]. A neuroinflammatory involvement is also in line with the findings of our transcriptomic analysis revealing DEGs associated with the transport of prostaglandins, which regulate neuroinflammatory pathways. Lipid BMP 42:8 belongs to a lipid class highly enriched in lysosomal membranes [
56] and is key for the proper functioning of lysosomal hydrolases [
57]. Interestingly, BMP accumulation also occurs in macrophages from patients with Gaucher disease [
58], a rare genetic disorder characterized by deposition of the lipid glucocerebroside in macrophages and that highly increases the life-time risk to develop PD [
59]. Nevertheless, we realize that it is difficult to draw functional conclusions concerning individual lipids since the number of samples employed in this proof-of-concept study is relatively small and only the roles of lipid classes have been characterized.
Even though our sample size was small, it is important to note that we found gender-related differences in the abundance of lipids both in PD patients and controls. This observation is consistent with the gender-related differences that have been found in the SN lipid profiles of PD patients [
53] and the “female pregnancy” pathway detected in our transcriptional analysis. Current data indicates that gender differences exist not only in PD incidence and prevalence [
60], but also in its clinical manifestation [
61,
62,
63], with estrogens as a possible mediator for these differences [
64], although the underlying molecular mechanisms are unknown. Since estrogens are steroid hormones and thus lipids modulating lipid metabolism in the brain [
65], and gender dimorphism has been observed in the fatty acid profiles of control mouse brains [
66] as well as in our data, brain lipidome modulation by sex hormones may be one of the underlying protective mechanisms in PD. Clearly, the possible link between sex hormones, brain lipids and PD deserves further attention.
In our exploratory transcriptomic study, the top upregulated DEG in the SN was
ORC7C1, which encodes an odorant receptor. Mesencephalic dopaminergic neurons express olfactory receptors and they respond to odorant-like molecules in mice [
67]. Several odorant receptors have been found to be downregulated in the frontal cortex of PD patients [
68], but the difference in the directionality of expression (up- or down-regulation) could be brain region related. The fact that the top upregulated DEG as well as the 7th upregulated DEG (i.e.,
OR7A5) represent odorant receptors highlights the importance of a poorly studied group of receptors that could modulate the cell behavior and fate via binding of small molecules and, thus, could play a crucial role in PD. The relationship between the upregulated
MTRNR2L8 and the mitochondrial
MT-RNR2 is at present unclear [
69] and its role in PD has not been analyzed yet. The protocadherin encoded by
PCDH20, which is thought to be involved in the establishment and function of specific cell–cell connections in the brain, has been found to be also upregulated in a meta-analysis of transcriptomic studies from SN of PD patients [
70]. Nevertheless, the role of this protein has thus far not been studied in the context of PD. The transcription factor KLF5 binds to GC box promoter elements and activates their transcription, mediating multiple cellular processes, such as proliferation, migration and differentiation, among others [
71], but its significance for PD is currently unknown.
The top-downregulated DEG in the SN is
TPH2, which encodes the rate-limiting enzyme in the synthesis of serotonin. Interestingly, serotoninergic neurons are progressively lost in PD, which is implicated both in motor and non-motor manifestations of PD [
72]. Moreover, polymorphisms in
TPH2 are associated with addictive behaviors [
73] and depression [
74] in PD patients. Additionally, Tph2 KO mice, which have serotonin deficiency, present with systemic oxidative stress and lipidomic abnormalities [
75], and swallowing dysfunction [
76]. The alpha-D1 adrenergic receptor encoded by
ADRA1D is also downregulated in the hippocampus of Alzheimer’s disease and dementia with Lewy bodies patients [
77], and noradrenergic impairment is present in PD patients as well [
78]. Both findings highlight the importance of neurotransmitters other than dopamine in the pathology of PD. It is therefore imperative to do further studies into their role to obtain a better understanding of the complexity of disease. The observed downregulated expression of
IL1B is in contrast with earlier findings showing increased
IL1B expression in striatum [
79] and cerebrospinal fluid [
80] of PD patients. However, the complex effects of IL1B seem to depend on the time, place and level of
IL1B expression, e.g., infusion of IL1B in the striatum of rats 5 days before 6-OHDA injection had a protective effect [
81]. Finally,
DOK7 encodes a protein involved in neuromuscular synaptogenesis, but its role in PD pathogenesis remains to be established.
PD is characterized by the loss of dopaminergic neurons from the SN [
4], which reaches a reduction of at least 70% by the time the disease is diagnosed. Thus, our top finding of decreased expression of genes associated with neurotransmitter levels, and dopamine synthesis and transport in postmortem PD SN samples is likely linked to the decreased number of dopaminergic neurons and fibers in the SN and putamen, respectively. One of these was the PD-associated
RET gene [
82], encoding a GDNF receptor required for the preventive and compensatory mechanisms linked to dopaminergic system degeneration that is triggered by the neurotrophic factor [
83,
84]. Remarkably, RET and some of its interaction partners (GFRA1, DOK4 and DOK6) are downregulated in multiple SN transcriptomic studies, highlighting the importance of the process of dopaminergic neurodegeneration in PD brains. The two most-enriched upregulated pathways were “cellular response to heat” and “protein folding”, which have an overlap of 80% among their DEGs. Since misfolded proteins are a hallmark of PD [
85], upregulation of genes associated with these pathways, such as the molecular chaperone HSPA1A, may represent a compensatory cellular mechanism to handle the accumulation of misfolded proteins. Furthermore, we confirmed other previously reported PD mechanisms, such as a dysfunction of protein folding. Interestingly, other SN DEGs comprise novel PD-related pathways that may provide a better understanding of the disease pathology, including “positive regulation of acute inflammatory response” (GO:0002675), “female pregnancy” (GO:0007565) and “G-protein-coupled glutamate receptor signaling pathway” (GO:0007216).
The overlap of our RNA-seq data with results from earlier microarray studies on postmortem PD and control SN confirms the importance of the degeneration of dopaminergic neurons in this brain region, since 10 out of the 12 DEGs that were found in at least half of the studies are associated with dopaminergic neurons or PD (ALDH1A1, DDC, DLK1, DRD2, GAP43, KCNJ6, SLC18A2, SLC6A3, SV2C and TH). The other two genes are AGTR1, the angiotensin II receptor subtype AT1, which has been shown to be downregulated in 5 out of the 8 SN studies, and the neuroendocrine proprotein convertase 1, PCSK1, which was downregulated in 4 out of the 8 SN studies. AGTR1 and PCSK1 do not have a known role in dopaminergic neurons or PD and are therefore interesting candidates for future functional studies. The reason for the restricted overlap between the results of our study and those of the only other PD/control SN RNS-seq analysis is at present unclear.
The top upregulated protein-encoding transcript in the putamen is MYOT, which codes for a component of a complex of actin cross-linking proteins. Transcript levels of MYOT have been found to be upregulated and downregulated in the SN and blood of PD patients, respectively [
86]. Furthermore, PD patients present a higher prevalence of MYOT autoantibodies than controls [
87]. Thus, MYOT may play distinct roles in the pathology of PD, but the nature of its involvement has not been elucidated. Neither have the function of the transcription factor ZNF646 or its link to PD been widely studied. The upregulated gene
HP1BP3 mediates chromatin condensation and modulates cognitive aging [
88], but not in known association with PD.
The top downregulated transcript in the putamen is SLC30A3, which encodes a protein involved in the accumulation of zinc in synaptic vesicles, rather than in the cytosol. Decreased levels of the protein have been observed in the (pre)frontal cortex of DLB and PDD patients [
89,
90] and is thus associated with dementia. Moreover, PD patients present changes in the cellular distribution of zinc in the SN [
91] and treatment of midbrain primary cultures with MPP+ leads to zinc ion accumulation [
92]. These findings highlight the relevance of bivalent metal ions in the pathology of PD and encourage further studies in this direction. The second most downregulated transcript was DUP2, a phosphatase involved in the negative regulation of ERK1 and ERK2. Members of the ERK family of proteins are known to control cellular proliferation and differentiation, but association of this function with PD has not yet been found. Similarly,
GNG2 encodes a gamma subunit of heterotrimeric G proteins, which are involved in cellular responses to external signals, but the significance of this function for PD still needs to be clarified.
PPL codes for a component of desmosomes and is thus involved in filament binding. Although
PPL has not been associated with PD, genetic variants of another Plakin family member,
MACF1, lead to decreased expression and confer risk for PD [
93]. Finally,
GRM2 encodes a metabotropic glutamate receptor and, interestingly, glutamate-induced excitotoxicity has been suggested to result in the loss of dopaminergic cells in PD [
94]. Moreover, activation of both metabotropic glutamate receptors 2 and 3, which have inhibitory functions, gives rise to striatal protection in PD rodent models [
95]. Hence, this finding can be considered further support for the involvement of multiple neurotransmitter systems in PD.
Analysis of data from various transcriptomic studies on putamen showed no overlapping genes. While SN is considered to be one of the primary sites of degeneration in PD, the neurodegeneration in the dorsal striatum may be mainly a consequence of the depletion of dopaminergic innervation [
96]. Thus, the lack of overlap among putamen DEGs could point to more heterogeneous pathology in the putamen than in the SN. We found an enrichment of downregulated putamen genes that play a role in the control of neurotransmitter levels and transport, which may well be associated with the loss of dopaminergic innervation [
96]. Furthermore, we observed DEGs associated with prostaglandin transport, which may argue for a neuroinflammatory response in this brain region [
97]. Indeed, neuroinflammatory events may play a role in nearly all known neurodegenerative diseases [
98].
Interestingly, we found 33 genes that were differentially expressed in both the PD SN and putamen, hinting at mechanisms shared by the two brain regions. The use of pathway analysis software nor the findings from extensive literature searches led to the identification of a specific pathway within the set of 33 genes. Of note is the downregulation of β-synuclein, the paralog of α-synuclein, that inhibits the assembly [
99], aggregation [
100] and toxicity [
101,
102] of α-synuclein. Since α-synuclein aggregation is one of the neuropathological hallmarks of PD [
103], β-synuclein upregulation may represent a potential target to ameliorate the disease. Hence, overexpression of β-synuclein in cellular and animal PD models would be central to better understand its role in the disease. Other potentially interesting examples of the set of common SN and putamen DEGs include
DDIT4, which mediates apoptosis in cellular PD-models [
104], and
P2RX7, the antagonist of which is neuroprotective in cellular and animal models for PD [
105]. Additionally, SNPs in the common genes encoding
EDN1 and
MYO5B have been associated with multiple systems atrophy (which presents with parkinsonism [
106]) and loss of the sense of smell (an early non-motor PD-symptom [
107]), respectively. Therefore, functional studies on the genes that are differentially expressed in both brain regions, e.g., by manipulating their expression in cellular and animal PD models, may provide clues for further understanding of the pathobiological complexity of PD.
The parallel analysis of the lipid and transcriptomic profiles of the same samples allowed us to integrate the two datasets. In this connection, one has to realize that the characterization of lipid species is limited, while the proteins involved in the metabolism of lipid classes have been well studied. Yet, targeted gene set enrichment analysis revealed that gene sets associated with the lipid classes differentially expressed in the SN and putamen were deregulated in the respective brain region. Thus, the lipidome and the transcriptome appear to be tightly connected, and transcriptomic changes leading to differences in the lipid profile may be one of the molecular mechanisms underlying neurodegeneration in PD. Still, additional studies involving multiple -omics approaches are necessary to allow a better characterization of the role of lipids in PD. Such studies will be of particular interest since several familial PD genes, such as
LRRK2 [
108],
PINK1 [
109],
DJ1 [
110],
ATP13A2 [
111],
PLA2G6 [
112] and
ATXN2 [
113], are known to modulate various lipid metabolism pathways.
This study has several limitations. First, relative to control samples, the SN samples from PD patients are characterized by an extensive loss of dopaminergic neurons and gliosis, which hinders making a distinction between lipid/mRNA changes caused by the disease-generated differences in the composition of cellular populations or by the molecular mechanisms leading to neurodegeneration (i.e., the “consequence” or the “cause”). Second, our lipid analysis was focused on phospholipids and sphingolipids, and did not include other lipid classes that could be relevant for PD, such as sterols and glycolipids [
24]. Third, we pooled samples for RNA-seq analysis, which hampered a gender-dependent analysis as performed with the lipid profiling, although the hierarchical clustering showed a clear separation by gender. Moreover, the pooling of samples might have masked minor differences in mRNA expression levels due to a small sample size and thus less statistical power.