Targeted Metabolomic Analysis in Parkinson's Disease Brain Frontal Cortex and Putamen With Relation To Cognitive Impairment

www.nature.
com/npjparkd
ARTICLE OPEN
Targeted metabolomic analysis in Parkinson’s disease brain

frontal cortex and putamen with relation to cognitive
impairment
1,2 ✉ 2
Karel Kalecký and Teodoro Bottiglieri
We performed liquid chromatography tandem mass spectrometry analysis with the targeted metabolomic kit Biocrates MxP Quant
500, in human brain cortex (Brodmann area 9) and putamen, to reveal metabolic changes characteristic of Parkinson’s disease (PD)
and PD-related cognitive decline. This case-control study involved 101 subjects (33 PD without dementia, 32 PD with dementia
(cortex only), 36 controls). We found changes associated with PD, cognitive status, levodopa levels, and disease progression. The
affected pathways include neurotransmitters, bile acids, homocysteine metabolism, amino acids, TCA cycle, polyamines, β-alanine
metabolism, fatty acids, acylcarnitines, ceramides, phosphatidylcholines, and several microbiome-derived metabolites. Previously
reported levodopa-related homocysteine accumulation in cortex still best explains the dementia status in PD, which can be
modified by dietary supplementation. Further investigation is needed to reveal the exact mechanisms behind this pathological
change.
1234567890():,;
npj Parkinson’s Disease (2023)9:84 ; https://doi.org/10.1038/s41531-023-00531-y
INTRODUCTION quantitative investigation to be performed across numerous

More than 6 million people worldwide live with Parkinson’s metabolic pathways simultaneously. To date, only few metabo-
disease (PD), the second most prevalent neurodegenerative lomic studies have been performed directly in PD brain tissue.
disorder, and the number is on the rise1. Progressive death of Some of them measured a single metabolite6–8, single pathway9,
dopaminergic neurons connecting substantia nigra and putamen, while others focused only on lipids10–13. The findings include
accompanied by accumulation of α-synuclein aggregates called decreased glutathione6, pantothenic acid7, increased urea8;
Lewy bodies, leads to gradual loss of motor control. The quality of increased sphingomyelins, ceramides, oxysterols, cholesterol, and
life can be further reduced by developing dementia, which occurs altered glycerophospholipids in cortex10. In addition, reports
up to 6 times more often among PD patients2. Pure PD dementia indicate increased diacylglycerols, decreased ceramides, and
is histopathologically different from Alzheimer’s disease dementia glycerophospholipids in cortex11; decreased sphingomyelins and
although both conditions can develop together, resulting in a phosphatidylinositol in putamen12; and decreased ceramides and
mixed pathology3. increased sulfatide in putamen13. None of these studies included
There is no known cure for PD. At best, the severity of motor more than 15 PD cases, only one corrected for multiple hypothesis
symptoms can be reduced by providing the brain with exogenous testing10, and only two had post-mortem collection intervals with
L-3,4-dihydroxyphenylalanine (DOPA), also referred to as levodopa differences between samples less than 1 day6,12. Larger high-
in the context of the medication, which passes the blood-brain
quality metabolomic studies are clearly needed.
barrier and is converted into dopamine. Levodopa is often sold as
We previously performed14 a metabolomic study in PD brain
mixture with N-amino-α-methyl-3-hydroxy-L-tyrosine monohy- tissue that was focused on a single metabolic pathway, specifically
drate (carbidopa), which inhibits amino acid decarboxylase
one-carbon metabolism. There, we identified homocysteine (Hcy)
thereby reducing peripheral metabolism of levodopa, so that
elevation in frontal cortex with acute levodopa presence as the
more of the drug can reach the brain. Unfortunately, the
treatment effectivity subsides over time, resulting in shorter most characteristic sign of dementia among PD subjects that
therapeutic windows of symptom reduction and more frequent could not be explained by medication dosage or disease
side effects such as dyskinesia. progression. Since the involvement of Hcy in dementia is well
PD is a complex disease. Multiple genetic and environmental established15 as well as the potential of levodopa to generate Hcy
factors contribute to the etiology4 although not all aspects have through its metabolism by catechol-O-methyltransferase (COMT),
been fully elucidated. Parkinson’s disease dementia (PD-D) seems we provided strong evidence for what has long been sus-
to be even more elusive5. Further understanding of the underlying pected16–19: the importance of the direct involvement of levodopa
mechanisms is needed to identify reliable biomarkers for early in dementia in susceptible individuals.
detection and intervention. In the current case-control study, we performed a broad
On a higher level, the quality of intracellular processes and explorative targeted metabolomic analysis in human frontal cortex
homeostasis is reflected in metabolism, which represents the (Brodmann area 9) and putamen, encompassing the main
functional part of cells and the interplay between tissues. Recent metabolic pathways as well as certain lipid classes, to better
methodological developments in metabolomics have allowed understand brain metabolic changes associated with PD at various
1
Institute of Biomedical Studies, Baylor University, Waco, TX 76712, USA. 2Center of Metabolomics, Institute of Metabolic Disease, Baylor Scott & White Research Institute, Dallas,
TX 75204, USA. ✉email: [email protected]
Published in partnership with the Parkinson’s Foundation

K. Kalecký and T. Bottiglieri
2
stages of cognitive impairment and the effect of levodopa There is a large increase in Hcy specifically in PD-D (Fig. 3b,
medication. Supplementary Fig. 2), which we previously reported14. This is the
most distinguishing alteration that we found related to dementia
status in PD, reaching area under the receiver operating
RESULTS characteristic curve (AUC) 89% (95% confidence interval (CI95)
Metabolic differences between controls and PD subjects in 75–100%) among L+ subjects (Supplementary Fig. 2). Two other
relation to cognitive impairment metabolites related to one-carbon metabolism, betaine and
Through a series of linear regression models with covariates sarcosine, follow the Hcy increase (Fig. 3b). In PD-D, levodopa is
adjusting for potential confounding factors (see Methods), we further associated with elevated phospholipase A2 (PLA2) activity
found multiple areas of metabolism different in PD, especially in and altered arginine metabolism, as evident by decreased
the groups with cognitive impairment (PD-CI; encompasses both indicators of homoarginine synthesis and symmetric dimethylated
mild cognitive impairment and dementia, which was confirmed as arginine (SDMA) methylation (Fig. 3b).
non-Alzheimer’s dementia – see Methods, section Study subjects). For several areas of metabolism, levodopa has seemingly
Statistically significant results are summarized in Fig. 1 for liquid opposite effects in putamen and in cortex of PD-D (Fig. 3c). In
chromatography (LC) part (small molecules and free fatty acids) fact, L+ changes observed in putamen, not present in the state of
and Fig. 2 for flow-injection analysis (FIA) part (complex lipids). All physiological DOPA levels (L−), well correspond to L- changes in
results are listed in detail in Supplementary File 1. Changes in PD-D, which disappear in L+, hence seeing the opposite direction
metabolic indicators depend on the context of effects of their for the levodopa effect (Figs. 1, 2, 3c). Therefore, what we see
constituent metabolites and are mentioned where relevant. Some rather appears to be a delayed effect (or side effect) of the
changes seem to be related to levodopa cycle and will be medication on metabolism in PD-D. This behavior, an example of
described in the next section. which is captured in Fig. 4, is significant (for both L+ PD putamen
The largest difference between PD and controls is in low and L- PD-D cortex) for decrease in spermidine and several lipid
concentrations of dopamine in PD putamen (Fig. 1), which reflects related indicators: decreased ratio of MUFAs compared to
the pathology definition. Another impacted neurotransmitter is polyunsaturated fatty acids (PUFAs), decreased ratio of MUFA
1234567890():,;
serotonin, visibly decreased in both brain regions in PD. Indicators phosphatidylcholines (PCs) compared to saturated fatty acid (SFA)
of phospholipase A2 activity are downregulated in PD cortex PCs, decreased ratio of acyl-alkyl PCs compared to diacyl PCs, and
(Fig. 2). We detected several bile acids (BAs) in brain tissue decreased ratios of simpler GCs to complex GCs. Certain indicators
including the metabolic indicator of glycine conjugation of related to sphingolipids show the same trend with significant
deoxycholic acid (DCA) to form glycodeoxycholic acid (GDCA) changes in L+ PD putamen and less significant opposite changes
that was significantly increased in PD cortex, with a similar but (FDR > 0.05, but p ≤ 0.05) in L+ PD-D cortex: decreased glycosyl-
non-significant trend in putamen. ceramides (GCs; ceramides with any glycosidic bond), decreased
Other changes from controls seem to be confined to PD-CI ratio of simpler GCs to ceramides, but increased ratio of
subjects: aspartate and α-aminoadipate are elevated in both brain trihexosylceramides and SMs to ceramides, and decreased ratio
regions, while betaine (trimethylglycine) is decreased (Fig. 1). of hydroxylated SMs to non-hydroxylated SMs. Certain triglycer-
Glutamate is elevated in PD-CI groups only in cortex, whereas the ides (TGs) and with lower significance BA indicators exhibit similar
polyamine putrescine shows a significant decrease in putamen patterns.
(Figs. 1, 2). Indicator of ornithine synthesis is higher in PD-D, less Additionally, there are metabolites dysregulated in a similar
significantly in cortex of other PD groups, compared to controls. fashion in L- PD-D (Fig. 1): increased indoxyl sulfoxide, and
Several metabolic indicators containing phenylalanine are dysre- indicators of synthesis of para-cresol sulfoxide and trimethylamine
gulated in PD-D with a similar but less significant trend in PD N-oxide. These trends are apparent in L+ PD-MCI cortex as well,
subjects with mild cognitive impairment (PD-MCI). Succinate is but do not reach statistical significance.
downregulated in PD-D in cortex and PD-MCI putamen. α- Interestingly, β-alanine metabolism is more dysregulated in L-
Aminobutyrate is increased only in PD-MCI cortex. and normalized in L+ (Figs. 1, 3c). The dysregulation consists of
The decreased indicator of glycine conjugation of primary BAs increasing β-alanine almost universally, but differs in other parts of
in PD-D (Fig. 1) is confounded by PD duration as explained further the pathway: anserine decreases in cortex of PD-CI only,
in the section Metabolic associations with disease progression 3-methylhistidine (π-methylhistidine) slightly fluctuates, especially
scores. in non-demented PD subjects (PD-ND), while 1-methylhistidine (τ-
methylhistidine) increases in PD-D.
Levodopa impact on metabolism and its distinctive effect in
PD-D Affected metabolic pathways based on KEGG database
The effect of levodopa-carbidopa medication (on top of the PD We also looked at the differential analysis results for each PD
subgroup effects in the linear regression sense) appears rather group in terms of metabolic set analysis using Kyoto Encyclopedia
complex. Statistically significant changes with acute levodopa of Genes and Genomes (KEGG) pathways20 for which we were
presence (L+), which we define as elevated DOPA levels (see able to map at least 4 metabolites. The results mostly overlap with
Methods), in combination with the dementia status, are depicted the findings for individual metabolites. The detected changes
in Fig. 3 (see Supplementary Fig. 1 for individual lipid species) and (Table 1) are confined mainly to PD-CI and are potentially affected
listed in detail in Supplementary File 1. However, to fully elucidate by the putative delayed levodopa medication effects.
the metabolic effects, we need to consider results for the effect of Besides β-alanine, histidine, cysteine and methionine, and
PD (Figs. 1 and 2) as well. D-amino acid metabolism, altered in PD-CI cortex and with a
The largest discovered levodopa-related change is high similar trend in putamen, we also see disrupted branched-chain
concentration of DOPA (Fig. 3a), which reflects our definition of amino acid biosynthesis in PD-CI cortex (Table 1). Glycine, serine
L+ subjects. Another change is in long-chain fatty acid (LCFA) and threonine metabolism is impacted in PD-MCI cortex, with a
acylcarnitines (ACs), which are decreased in putamen, showing a similar tendency in other cognitively impaired groups, alongside
similar but non-significant trend in cortex. In putamen, we also aminoacyl-tRNA biosynthesis. This suggests that the amino acid
observe increased lysophosphatidylcholines (LPCs) and mono- metabolism is affected more than discovered from individual
unsaturated fatty acid (MUFA) cholesteryl esters (CEs). metabolites.
npj Parkinson’s Disease (2023) 84 Published in partnership with the Parkinson’s Foundation
3
Fig. 1 Metabolic differences between subject groups in liquid chromatography experiment. Forest plot for differentially detected
metabolites and metabolic indicators in the LC part with regression coefficients for individual PD groups (CN Cognitively normal, MCI – with
mild cognitive impairment, D – with dementia of non-Alzheimer’s type) and brain areas. The “N” in the legend denotes the number of samples
effectively used to estimate each regression parameter. Values are normalized regression coefficients (depicted as the central points; the
shape reflects the significance) with 95% confidence intervals (horizontal range lines; lower opacity for non-significant coefficients). The
dashed vertical black line represents a zero effect, i.e. equivalent to controls (for a given brain region). Every analyte listed is statistically
significant (FDR ≤ 0.05) for one of the PD groups considered separately or all PD subjects (for a given brain region) considered together. The
regression models included covariates as detailed in the Methods.
Published in partnership with the Parkinson’s Foundation npj Parkinson’s Disease (2023) 84
4
Fig. 2 Metabolic differences between subject groups in flow-injection experiment. Forest plot for differentially detected metabolites and
metabolic indicators in the FIA part with regression coefficients for individual PD groups (CN – cognitively normal; MCI – with mild cognitive
impairment; D – with dementia of non-Alzheimer’s type) and brain areas. The “N” in the legend denotes the number of samples effectively
used to estimate each regression parameter. Values are normalized regression coefficients (depicted as the central points; the shape reflects
the significance) with 95% confidence intervals (horizontal range lines; lower opacity for non-significant coefficients). The dashed vertical
black line represents a zero effect, i.e., equivalent to controls (for a given brain region). Every analyte listed is statistically significant
(FDR ≤ 0.05) for one of the PD groups considered separately or all PD subjects (for a given brain region) considered together. The regression
models included covariates as detailed in the Methods.
Among lipids, glycerolipid and glycerophospholipid metabolism cognitive impairment, including both clinical and histopathologi-
are altered in PD-CI, while several PUFA pathways (arachidonic cal scores, among PD subjects. Covariates were included as
acid, linoleic acid, and α-linolenic acid metabolism) are mainly previously, only the acute levodopa status was omitted for its
affected in PD-D and in PD-MCI putamen but not PD-MCI cortex complicated relationship with PD cognitive subgroups, which
(Table 1). Sphingolipid metabolism is similar, additionally showing were not considered in this part and all PD subjects were analyzed
a significant difference in PD-CN cortex. together. The significant results are listed in Table 2 (see
Interestingly, we see pantothenate and coenzyme A biosynth- Supplementary File 2 for complete results) and most of them
esis affected in PD-CI (Table 1). relate to putamen. The less significant brain region follows the
same trends with only several exceptions.
Metabolic associations with disease progression scores In cortex, duration of PD since diagnosis is associated with
Next, we explored associations between analytes (metabolites and decreasing glycine conjugation of primary bile acids to form
metabolic indicators) and measures of disease progression and glycocholic and glycochenodeoxycholic acids (Table 2). A similar
5
Fig. 3 Effect of acute levodopa medication presence on metabolism in PD groups. Forest plot for differentially detected metabolites (LC
only) and metabolic indicators (both LC and FIA) with regression coefficients for the acute levodopa presence in PD in interaction with
dementia status (ND – no dementia; D – with dementia of non-Alzheimer’s type) and brain areas. These effects are additive to the PD group
effects (Figs. 1 and 2) for subjects with acute levodopa presence, as the regression parameters were a part of the same linear regression
models. The “N” in the legend denotes the number of samples effectively used to estimate each regression parameter. Values are normalized
regression coefficients (depicted as the central points; the shape reflects the significance) with 95% confidence intervals (horizontal range
lines; lower opacity for non-significant coefficients). The dashed vertical black line represents a zero effect, i.e. equivalent to PD subjects with
physiological DOPA levels (in a given group). Every analyte listed is statistically significant (FDR ≤ 0.05) for levodopa medication in putamen or
cortex in interaction with dementia status. Pathways are grouped by similarity in the levodopa effect: (a) in all samples or putamen only, (b) in
PD with dementia cortex or similarly in putamen, and (c) in PD with dementia cortex and differently in putamen (upon interpretation with
Figs. 1 and 2). The regression models included covariates as detailed in the Methods.
6
although weaker association with glycine conjugation of primary diagnosis duration: Diagnosis duration correlates with the glycine
BAs was found in connection with PD-D, and given that the PD-D conjugation of primary BAs within PD-D subjects alone (r = −0.42,
group also has longer PD duration compared to other groups, we p = 0.015), whereas PD-D shows only a non-significant difference
suspected a confounding effect. Indeed, further exploration compared to PD-MCI and PD-CN at the similar levels of diagnosis
revealed that this association for PD-D is confounded by the duration. This conclusion is also supported by the fact that the two
measured glycine-conjugated primary BAs (GCA and GCDCA) are
associated with the diagnosis duration but not with PD-D, and on
the other hand, other BA-related analytes that are significantly
different in PD-D are unrelated to glycine-conjugated primary BAs
and are not significantly associated with the disease duration.
Motor section score of the Unified Parkinson’s Disease Rating
Scale (UPDRS-M)21 is associated with decreasing α-aminobutyrate
synthesis in both brain regions (Table 2). Unified Staging System
for Lewy Body Disorders (USSLB)22 score is associated with a
marker of carnitine-acylcarnitine translocase deficiency in cortex.
The remaining results are significantly associated only in
putamen: USSLB score is associated with decreasing glutamine
(increasing glutaminase activity), betaine, lactate, as well as
glycine conjugation of chenodeoxycholic acid and sum of
neurotransmitters (Table 2). Global senile plaque density shows
significant associations with decreasing threonine, glutamine
(increasing glutaminase activity and glutaminolysis rate), seroto-
nin, and two lipid species – PC aa C28:1 and TG(22:6_32:1). Mini-
Fig. 4 Example of phase delay in levodopa effect in PD with Mental State Examination (MMSE) score23 is associated with
dementia. In this box plot (center line – median; box limits – upper decreasing α-aminobutyrate synthesis and indicator of short/
and lower quartiles; whiskers – 1.5× interquartile range), depicted branched-chain acyl-coenzyme A dehydrogenase deficiency, and
across subject groups (CN – cognitively normal; MCI – with mild
with increasing betaine and several indicators with phenylalanine.
cognitive impairment; D – with dementia of non-Alzheimer’s type) in
dependence on the acute levodopa medication presence (L+; red There was no statistically significant association with neurofi-
points) or physiological range (L−; black points), we show an brillary tangles (Table 2). We mentioned several analytes in
example of a metabolite (phosphatidylcholine PC aa C38:4; normal- association with multiple scores, so we investigated these
ized values) with a seemingly opposite levodopa effect in PD-D relationships further by testing correlation between the scores
subjects (decreasing red arrow). However, the values do not go in and by combining them into a mutual regression model. MMSE
the opposite direction (below the baseline), but rather the same and USSLB do not correlate well (r = −0.19, p = 0.34) and remain
change happens in the opposite levodopa state (increase in L- highly significant (both p = 0.002) in a combined regression model
instead of a change in L+), suggesting a delay in the levodopa-
for betaine. Therefore, betaine seems to be associated with both
induced effect in PD-D cortex, potentially starting to appear in PD-
MCI cortex. Similar patterns were observed across multiple scores independently. MMSE and UPDRS-M correlate very mildly
metabolic classes. Statistical significance of regression coefficients (r = 0.28, p = 0.20), but only UPDRS-M (p < 0.0001), and not MMSE
for acute levodopa presence (from the main regression analysis, (p = 0.11), remains significant in a combined regression model for
which includes covariates) is provided. α-aminobutyrate. Therefore, UPDRS-M is the main score associated
Table 1. Metabolic set analysis in PD cognitive groups using KEGG pathways.
KEGG pathway metabolic set statistics with statistically significant difference (FDR ≤ 0.05) for at least one of the PD groups as compared to controls. Computed
from the main regression models with covariates as detailed in the Methods. Yellow: FDR ≤ 0.05, light orange: p-value ≤ 0.05, green: average absolute effect
(jβj) – the darker the color, the higher the value.
CN Cognitively normal, CoA Coenzyme A, D Dementia (non-Alzheimer’s type), FDR False discovery rate, MCI Mild cognitive impairment, PD Parkinson’s disease.
7
Table 2. Metabolic associations with progression scores in PD subjects.
Statistically significant (FDR ≤ 0.05) associations with several progression scores in either brain region among PD subjects. UPDRS-M score, measured in off-
levodopa state, is known for 57% of PD subjects. MMSE score is known for 88% of PD subjects. Yellow: FDR ≤ 0.05, light orange: p-value ≤ 0.05, red/blue:
positive/negative effect (the darker the color, the higher the magnitude) with two-way standardization.
aa diacyl, β Effect size (regression coefficient), L95 Lower bound of 95% confidence interval, MMSE Mini-Mental State Examination, PC Phosphatidylcholine, PD
Parkinson’s disease, TG Triglyceride, U95 Upper bound of 95% confidence interval, UPDRS-M Motor section of the Unified Parkinson’s Disease Rating Scale,
USSLB Unified Staging System for Lewy Body Disorders.
with α-aminobutyrate, not MMSE. Accordingly, only the associa- derived metabolites are also differentially present in brain.
tion with UPDRS-M was significant in both brain regions. USSLB Levodopa-related homocysteine accumulation in brain best
and senile plaque density show a medium correlation (r = 0.40, explains the dementia status in PD.
p = 0.021) and both scores remain significant (p = 0.03, 0.007) in a Reduced serotonin synthesis in PD alongside dopamine reflects
combined regression model for glutamine. The size of the effects the degeneration of both neurotransmitter systems. Serotonergic
also remain similar (β = −0.40, −0.51) and we cannot tell whether impairment in PD is well-known but poorly understood24. We
the significance is due to the correlation or due to independent show that the decrease in serotonin is apparent regardless of
associations with glutamine. However, observing the trends in cognition status but we found a significant association with senile
cortex, glutamine and related indicators show more consistent plaque formation. This is in line with the evidence that stimulation
effects in association with senile plaque density rather than USSLB. of serotonin receptors leads to decrease in amyloid plaque
formation via activation of α-secretase25. The reason behind the
serotonergic impairment is unclear and it may be a sign of
DISCUSSION
broader neurodegeneration where dopaminergic neurons are the
To the best of our knowledge, this is the largest metabolomic
most but not the only impacted neurons. Some authors suggest
study in PD with human brain tissue to date and the largest
levodopa contribution, especially since it is processed in both
metabolomic study in human putamen in general. Our results
dopaminergic and serotonergic neurons, and several mechanisms
reveal metabolic changes associated with PD, with specific
of levodopa toxicity on the serotonergic system have been
cognitive subgroups, with levodopa, and with clinical and
pathological scores of disease progression. The affected metabolic proposed26. This process would happen as a result of long-time
pathways include neurotransmitters and bile acids in PD; levodopa exposure and would likely not be reflected in
glutamate metabolism, polyamines, and betaine metabolism in momentary DOPA levels that fluctuate as the medication is
PD with cognitive impairment; while differences in one-carbon absorbed and eliminated from the body. We see a mild, non-
metabolism, LPCs, cholesteryl esters, ACs, TCA cycle, and β-alanine significant trend of decreased serotonin with increased DOPA in
metabolism are associated with acute levodopa presence. Some cortex (r = −0.20, p = 0.11). Alternatively, we cannot exclude the
levodopa-related changes further exhibit an apparent temporal possibility that a portion of carbidopa is passing through the
delay in PD-D: spermidine, FAs, PCs, and GCs. Several microbiome- blood-brain barrier when its integrity is impaired, directly
8
elevated. Glutamate accumulation is neurotoxic and its involve-
ment in PD has long been suspected37. Glutamate can be
converted into 2-oxoglutarate and consumed in the tricarboxylic
acid (TCA) cycle via 2-oxoglutarate dehydrogenase (OGDH)
complex. However, our results suggest a bottleneck in the TCA
cycle as evident by the decreased succinate. OGDH complex
function depends on active forms of several B vitamins, including
CoA, and its dysfunction would compromise the cellular energy
metabolism and homeostasis. Glutamate and aspartate do not
pass the blood-brain barrier into the brain, but glutamate passes
outside through cystine/glutamate antiporter38, an important
system in the clearance of excessive glutamate and import of
cystine, a precursor of antioxidant glutathione. The cystine/
glutamate antiporter system is inhibited by α-aminoadipate39,
which we found elevated, thus potentially impairing the antiporter
function. A particularly sensitive area for its dysfunction is retina40,
Fig. 5 Glutamate metabolism and related pathways altered in PD and consistent with our hypothesis, retinal damage has been
with cognitive impairment. Selected metabolic reactions pertaining recently proposed as an early marker of PD41. Additionally,
to glutamate metabolism and its connection with α-aminoadipate changes in DNA methylation of the antiporter gene SLC7A11 have
and observed changes. Arrows: red – upregulated; blue – down-
regulated; black solid – metabolic reaction; black sparsely dashed – been associated with a risk of PD42 and lower glutathione levels
export from brain; black densely dashed – inhibition. Font: gold – have been reported in PD brain6. The reason for increased α-
measured metabolite; gray – metabolite not included in the assay; aminoadipate is not obvious. Interestingly, its degradation path-
purple – enzymes and cofactors; bright blue – transporters. 2-AA α- way employs 2-oxoadipate dehydrogenase complex (OADH),
Aminoadipate, 2-OA 2-Oxoadipate, 2-OG 2-Oxoglutarate, Asp which is very similar to OGDH complex including shared
Aspartate, Bn Active form of vitamin Bn, CoA Coenzyme A, DH subunits43 and dependence on B vitamins and CoA. We
Dehydrogenase complex, Glu Glutamate, Lys Lysine, NAA N- hypothesize that the observed changes are caused by dysregula-
Acetylaspartate, OA Oxaloacetate, Orn Ornithine, Suc Succinic acid, tion in these complexes, possibly due to low levels of vitamin B5
TCA Tricarboxylic acid, xCT Cystine/glutamate antiporter. and CoA. Accordingly, N-acetylaspartate is acetyl-CoA-dependent
product of aspartate/glutamate metabolism and has been found
inhibiting serotonin synthesis. Disrupted blood-brain barrier has downregulated in imaging studies44, implying a bottleneck,
been associated with PD27. whereas ornithine is a CoA-independent product of glutamate.
PLA2 plays an important role in regulation of phospholipid Interestingly, we found the indicator of ornithine synthesis
metabolism and inflammation. The observed reduction in increased similarly as glutamate, showing undisturbed propaga-
indicator of PLA2 activity in PD cortex is, however, likely a result tion of the glutamate change.
of changes in composition of PCs (with elevation of specific diacyl Ornithine is further metabolized into putrescine through the
PCs) rather than the absolute PLA2 activity and its production of action of the rate-limiting enzyme ornithine decarboxylase, which
LPCs and arachidonic acid, as they remain unaltered. was decreased in PD-CI putamen. A previous investigation of basal
The detection of bile acid changes in brain confirms the ganglia found no differences in putrescine in PD9 but the number
existence of gut-brain cross-talk. Primary BAs are produced in liver, of samples was small, and cohorts differed in post-mortem
secreted into small intestine when needed, and later reabsorbed. collection intervals. Putrescine has been detected and shown to
Meanwhile, gut bacteria can convert them into rather pro- be decreased in red blood cells in PD45 but elevated in
inflammatory secondary BAs. Conjugation with amino acids cerebrospinal fluid (CSF)46,47, although these results might be
readily occurs in liver as well as microbiota28, and changes in confounded by age. Another metabolomic study in serum found
the BA composition have been associated with various patholo- no differences in putrescine in PD48. There is also a report of
gies29, including AD in our previous research30. The involvement putrescine as one of the metabolites contributing to diagnosis of
of BAs in PD is also suspected31 and has been connected with dementia status in PD49. Since the change that we observed is not
microbial dysbiosis32. Our results show increased glycine conjuga- propagated to other polyamines, it might not have any important
tion of secondary bile acid GDCA across the cognitive groups, and biological effect. However, it could be a surrogate marker of a
progressive decline in glycine conjugation of primary BAs with deficiency in pyridoxal 5-phosphate (PLP), the active form of
respect to the disease duration (cortex) and USSLB (putamen). This vitamin B6, which acts as a cofactor for ornithine carboxylase. We
is consistent with reports of elevated GDCA in PD plasma33,34 and observed other signs of potentially low PLP in brain of subjects
gut bacteria changes related to reduced primary BA biosynth- with cognitive impairment in our study of one-carbon metabo-
esis35. Worth noting is the connection with coenzyme A (CoA) in lism14. Lower plasma B6 has been previously associated with PD as
the process of BA conjugation36, since our pathway analysis well50.
detected a disturbance in pantothenate and CoA biosynthesis, at Betaine is also related to one-carbon metabolism, as it facilitates
least in the cognitively impaired groups. Inadequate amount of re-methylation of the toxic sulfur amino acid Hcy, although its
CoA would slow down the rate of primary BA conjugation in liver. functions are wider and exhibit anti-inflammatory, anti-apoptotic,
Indeed, decreased pantothenate (vitamin B5), the precursor of and anti-diabetic properties51. We previously reported decreased
CoA, has been observed in PD brain7. concentration of betaine in cortex of cognitively impaired
We found no significant disturbances in PD related to subjects, both PD and AD14. Here, we confirm that the same
diacylglycerols, glycerophospholipids, and sphingolipids reported reduction exists in PD putamen and is independently associated
by previous studies in cortex10,11 although we found changes in with cognitive decline (MMSE) and spread of histopathological
glycerophospholipids and sphingolipids in association with acute changes (USSLB). Given the interesting properties of betaine, it is
levodopa presence as discussed further in sections devoted to plausible that its deficiency directly contributes to neurodegen-
levodopa. eration. Indeed, there are reports showing benefits of betaine
There are several interconnected areas of metabolism altered in supplementation on cognition52,53.
PD-CI groups, as depicted in Fig. 5. Glutamate and aspartate are There are several altered metabolic indicators related to
mutually convertible and both metabolites are significantly phenylalanine, with the largest difference in lower tyrosine/
9
phenylalanine ratio, mainly due to higher phenylalanine. This Hcy accumulation in brain cortex of human PD subjects as
indicates an impaired function of phenylalanine hydroxylase, a characteristic of their dementia status14.
crucial enzyme for dopamine synthesis. Its dysfunction in PD is a The current study found no better discriminator of dementia in
known phenomenon and is considered a secondary event PD than the levodopa-related Hcy accumulation, supporting the
reflecting the death of dopaminergic neurons54. Similarly, we importance of its direct link to the dementia in PD. Plasma Hcy can
found several metabolic indicators with phenylalanine associated be lowered by supplementation of certain B vitamins70 and the
with MMSE score, with cognitive decline always in the direction of same effect is also expected in brain (due to the presence of all
higher phenylalanine concentrations. vitamin B-dependent Hcy-related pathways), which is supported
Some changes associated with scores of disease progression by high brain Hcy levels in a mouse model for B12 deficiency71.
have already been mentioned. Another strong relation is between Several other changes were related to methylation (arginine
UPDRS-M motor symptoms progression and lower α-aminobuty- methylation, betaine, and sarcosine), which might be explained by
rate. This metabolite is a downstream product of catabolism of disproportion in S-adenosylhomocysteine (SAH) and
threonine and cystathionine and has been directly related to S-adenosylmethionine caused by levodopa methylation.
glutathione compensation against oxidative stress55. In PD, lower LPCs, together with arachidonic acid, are produced by PLA2, the
levels of α-aminobutyrate have been previously observed in activity of which is increased during inflammation72, and have a
cerebrospinal fluid56. Given the decreased levels of glutathione in potentially detrimental effect on mitochondria73. The increase of
PD brain6, it is possible that our finding of the association with α- LPCs in PD L+ putamen, along with a similar although non-
aminobutyrate reflects the protective effect of glutathione on significant trend in arachidonic acid, suggests a possible
deterioration of motor control in PD. levodopa-related inflammatory process. LPCs have been pre-
Senile plaque density is associated with decreasing glutamine. viously associated with upregulated cholesterol biosynthesis74,
This observation is consistent with the literature, where glutamine consistent with our finding of increased MUFA-CEs in L+
synthetase has been found downregulated in AD astrocytes putamen.
especially near the senile plaques and without the reduction of The observed decrease in LCFA-ACs in L+ putamen reflects a
astrocyte count57. Dysregulated glutamine metabolism promotes disturbance in FA β-oxidation, whether due to diminished
glutamate toxicity and neurodegeneration, which can occur in substrate availability (CoA, FAs) or downregulated function of FA
response to inflammatory cytokines58. The association of USSLB transport into mitochondria or peroxisomes. Similar LCFA-ACs
with decreasing lactate is also consistent with astrocytic impair- decrease has been observed in plasma of PD subjects as
ment. Lactate, in brain produced mainly by astrocytes, is an compared to controls75 although not directly correlated with
important energy source for neurons. Astrocytes mediate α- levodopa. Here, we showed clear levodopa-related downregula-
synuclein clearance, and this mechanism is impaired in a known tion in putamen, with a similar but non-significant trend in cortex.
genetic form of PD59. It has been shown that lactate helps activate Any energy metabolism disruption can negatively impact cell
autophagy and rescue cells in a PD model60. survival during cellular stress as suspected with levodopa.
There are several more associations with senile plaque. Surprising is the finding of apparent phase delay in the
Triglyceride TG(22:6_32:1) and phosphatidylcholine PC aa C28:1 levodopa cyclical effect in cortex of PD-D subjects, potentially
decrease with higher plaque load in PD putamen. One chain of starting to appear already in PD-MCI. Since our L+/L− classifica-
TG(22:6_32:1) is formed by ω-3 docosahexaenoic acid (DHA), tion is based directly on DOPA levels in the tissues, this effect
which is thought to impede amyloid production61, so inadequate cannot be caused by delayed drug absorption or transport into
levels of DHA trafficked into brain could accelerate plaque brain. The differential response in one-carbon metabolism with
formation. The meaning of the association with PC aa C28:1 is Hcy (and SAH) accumulation in PD-D might play a role, as there
unclear. Similarly for threonine although the pattern is a bit are multiple regulatory mechanisms inhibiting numerous methyl-
different: the PD subjects with lower plaque density in putamen transferases, including DNA methyltransferases affecting gene
have levels in the higher range of controls, whereas those with expression76, and increasing Hcy clearance. Therefore, we
higher plaque density have threonine in the lower range of hypothesize that the levodopa effect can show variable delay
controls. This relationship is interesting, raising a question whether depending on the magnitude of Hcy accumulation, strength of
threonine can have a protective effect on senile plaque formation the initiated regulatory response, and speed of Hcy clearance –
in PD or if it mirrors another underlying mechanism. parameters which are different between the PD cognitive groups.
Increasing marker of short/branched-chain acyl-CoA dehydro- This observed effect warrants further investigation. Temporal
genase deficiency with decreasing MMSE score suggests affected phenomena can be ideally confirmed in a longitudinal study.
mitochondrial β-oxidation via downregulation of 2-methylbutyryl- However, this might not be realizable with human brain tissue
CoA dehydrogenase. Its dysfunction promotes lipid oxidative (using a similarly invasive technique) and less powerful
damage and depletes glutathione levels62. Thus, it is possible that approaches might be necessary, including larger association
downregulation of this enzyme contributes to cognitive impair- studies or investigations with animal models.
ment, especially in the presence of concomitant insults as in PD. Nevertheless, we detected this pattern across multiple classes of
Fatty acid β-oxidation impairment is further indicated in cortex in metabolites. The changes in fatty acid are mainly due to
association with higher USSLB score, where the marker of decreased MUFAs. This would be consistent with reduced lipolysis,
carnitine-acylcarnitine translocase deficiency suggests a lower a consequence of Hcy metabolism generating cysteine77, and a
activity of this enzyme as the pathology spreads. possible cause of disturbed β-oxidation. Decreased ratio of acyl-
The effect of levodopa medication on metabolism is not well alkyl/diacyl PCs then points towards reduced function of
explored and there is a debate whether its side effects accelerate alkylglycerone phosphate synthase or fatty acid reductase. These
neurodegeneration63. There are reports of affected glucose and enzymes are located in peroxisomes, so the change might reflect
fatty acid (FA) metabolism64,65, lipid peroxidation66, as well as peroxisomal dysregulation. Hcy might be the cause via inter-
increased stress on proteolytic67 and lysosomal68 systems due to ference with peroxisome proliferator-activated receptor methyla-
DOPA incorporation into proteins, resulting in mitochondrial tion78. Several decreased ceramides and sphingomyelins were
dysfunction. Levodopa and carbidopa also sequester PLP, an previously found in PD putamen12,13 and we see their decrease in
important cofactor for many biological processes, and have been relation to levodopa. Altered ratios of ceramides, GCs, and SMs
shown to reduce plasma PLP in PD69. We have previously suggest impeded degradation of 3GCs to simpler GCs, which are
described levodopa-induced changes in methylation and τ- directed by lysosomal enzymes α-galactosidase and
protein phosphorylation in mice19 and showed levodopa-related β-glucocerebrosidase. Lysosomal clearance can also be reduced
10
Table 3. Sociodemographic and clinical characteristics of the cohort.
Groups P-value
Controls (N = 36) PD-CN (N = 14) PD-MCI (N = 19) PD-D (N = 32) All groups PD groups
PMCI, hours, mean (SD) 3.2 (1.0) 3.3 (0.9) 3.4 (1.2) 3.4 (0.9) 0.78 0.85
Freezer storage, years, mean (SD) 10 (4) 8 (4) 6 (4) 8 (4) 0.005 0.31
Sex
Male, No. (%) 22 (61) 8 (57) 14 (74) 24 (75) 0.49 0.52
Female, No. (%) 14 (39) 6 (43) 5 (26) 8 (25)
Ethnicity
Non-Hispanic, No. (%) 32 (89) 14 (100) 16 (84) 28 (88) 0.59 0.44
Not provided, No. (%) 4 (11) 0 (0) 3 (16) 4 (13)
Race All White 1 1
Age, years, mean (SD) 82 (10) 85 (6) 82 (6) 80 (5) 0.057 0.033
Education, years, mean (SD) 14 (3) 16 (3) 14 (3) 16 (3) 0.039 0.073
BMI, kg/m2, mean (SD) 25 (5) 22 (4) 25 (8) 27 (9) 0.042 0.024
APOE, ε4 allele, No. (%) 5 (14) 2 (14) 3 (16) 3 (9) 0.86 0.70
PD duration, years, mean (SD) – 11 (7) 12 (6) 18 (9) – 0.008
Dementia onset, years, mean (SD) – – – 75 (6) – –
Levodopa dosage, mg/day, mean (SD) – 450 (217) 500 (173) 509 (226) – 0.86
UPDRS-M score, mean (SD) 7 (5) 32 (15) 31 (15) 50 (16) <0.001 0.005
USSLB score, mean (SD) 0.0 (0.0) 3.0 (0.5) 3.7 (0.5) 3.4 (0.6) <0.001 0.002
MMSE score, mean (SD) 28 (1) 28 (1) 25 (3) 20 (6) <0.001 <0.001
Neurofibrillary tangle density, mean (SD) 2.9 (1.6) 5.3 (2.6) 5.2 (1.6) 4.8 (2.1) <0.001 0.68
Senile plaque density, mean (SD) 2.7 (3.9) 4.6 (4.7) 5.1 (6.0) 1.8 (3.3) 0.056 0.025
Hyperlipidemia, No. (%) 13 (36) 7 (50) 9 (47) 11 (34) 0.65 0.49
Diabetes mellitus, No. (%) 12 (33) 1 (7) 3 (16) 8 (25) 0.21 0.34
Renal insufficiency, No. (%) 5 (14) 3 (21) 2 (11) 5 (16) 0.87 0.68
Hypothyroidism, No. (%) 9 (25) 3 (21) 5 (26) 5 (16) 0.74 0.61
Levodopa dosage is unambiguously known only for 31% of PD subjects. MMSE score is known for 88% of PD subjects and 75% of controls. UPDRS-M score,
measured in off-levodopa state, is known for 57% of PD subjects and 67% of controls. P-values are listed for comparison between all groups and between PD
groups only.
APOE Apolipoprotein E, BMI Body-mass index, CN Cognitively normal, D Dementia (non-Alzheimer’s type), MCI Mild cognitive impairment, MMSE Mini-Mental
State Examination, PD Parkinson’s disease, PMCI Post-mortem collection interval, UPDRS-M Motor section of the Unified Parkinson’s Disease Rating Scale, USSLB
Unified Staging System for Lewy Body Disorders.
by homocysteine79, and moreover, lysosomal dysfunction further for their interaction with levodopa is not clear and it is conceivable
attenuates peroxisomal function80. Furthermore, decreased sper- that the medication stimulates growth of DOPA-metabolizing
midine can downregulate lysosomal and autophagosomal func- bacteria.
tion81 as well as mitochondrial respiration82. Spermidine is This study has multiple strengths: We analyzed changes directly
connected to the methionine cycle through PLP-dependent SAM in brain tissue, which can reveal more information about the
decarboxylase and it is unclear whether the levodopa-carbidopa neurodegeneration than surrogate measurements in biofluids. The
medication can directly interfere with this pathway through PLP samples had exceptionally low post-mortem collection intervals
sequestration or changes in SAM. Lower levels of spermidine have and were homogeneous across the groups. This is highly
been previously reported in CSF of PD subjects47. important for reliable comparison, as post-mortem degradation
The oscillation of β-alanine metabolism results in notably processes affect metabolism and energy balance. We were able to
lowered anserine (and to a similar degree but without statistical analyze PD subjects in relation to cognitive impairment, which
significance carnosine) levels in PD-CI cortex. These histidine revealed interesting changes. Similarly, we investigated the
dipeptides are important antioxidants and have been previously impact of acute PD medication presence, which turned out to
associated with cognitive status83 and PLP availability84. Interest- have a profound effect on brain metabolism. Furthermore, our
ingly, the changes in PD-D include muscle-related τ-methylhisti- analysis considered multiple confounding effects, including
dine and might be an indicator of faster muscle catabolism, while diseases notoriously impacting metabolism.
in PD-ND, diet-related π-methylhistidine85 is more affected. There are also several limitations. By design, this is an
Finally, the group of microbiome-related metabolites consisting association study, which prohibits us from validating causal
of para-cresol sulfate, indoxyl sulfate, and trimethylamine N-oxide relationships although we discuss plausible connections with the
fluctuates in PD-CI, especially dementia. These compounds are etiology of PD or related cognitive impairment. For PD-D subjects,
considered metabotoxins and have been previously found we had no putamen tissue, which would help us confirm the
elevated in PD biofluids86. We have reported their increase in consistency of our findings. Not all disease progression scores
AD subjects30, which strengthens the association of these were known for all subjects and even though we attempted to
compounds with cognitive impairment in general. The reason detect confounding through comparison of associations between
11
PD groups and the progression scores, a larger study would be Americans, either non-Hispanic or with ethnicity not provided. No
needed to fully disentangle their individual contributions. Despite other major pathologies of the central nervous system were
analyzing the levodopa impact, this does not capture any present.
potential effect of carbidopa for its substantially larger elimination Clinical profiles with assessment of progression scores were
halftime. No genotyping was performed to identify subjects with a updated during the last pre-mortem visit. Histopathological
known genetic cause of PD. Next, we did not have any information density scores of neurofibrillary tangles and senile plaque
regarding diet. PD patients are often advised to avoid protein (neuritic, cored, and diffuse) were evaluated using templates of
intake immediately with levodopa (due to absorption), which the Consortium to Establish a Registry for Alzheimer’s Disease
could affect several areas of metabolism. We found some (CERAD)90 and the scores were summed over five brain regions:
levodopa-related oscillation in π-methylhistidine, a suggested frontal, temporal, parietal, entorhinal, and hippocampal CA1
marker for meat consumption85, but it was in the opposite region.
direction and did not correlate well with other findings, which Basic sociodemographic and clinical characteristics are sum-
would remain significant after adjustment for π-methylhistidine marized in Table 3. Sociodemographic variables show either no
levels. Furthermore, all subjects were non-Hispanic (where differences between the groups or minor differences with a good
ethnicity was provided) White Americans and it is not certain overlap. Levodopa medication dosage is similar across the PD
whether the metabolic changes generalize outside this context. groups although this information is available only for a third of the
In conclusion, we performed a large metabolomic analysis in subjects. Scores of disease progression and histopathological
human brain. Most of the discovered changes are associated with examination show major differences as expected. Senile plaque
cognitive impairment, especially glutamate, aspartate, and α- seems to be elevated only in PD-ND groups. UPDRS-M score and
aminoadipate metabolism and one-carbon metabolism with PD duration are particularly higher in PD-D group, which is a
levodopa-related homocysteine accumulation in brain best concern for a possible confounding effect. As the UPDRS-M score
explaining the dementia status in PD (AUC = 89%, is known only for 57% of PD subjects, we could not include the
CI95 = 75–100%). These alterations may indicate insufficient levels score in the main differential regression for PD groups. Instead, as
of vitamins B5 (pantothenic acid), consistently with a previous described in the Statistical analysis section below, we analyzed
finding7, and B6 (pyridoxal 5’-phosphate). Therefore, it is plausible disease progression scores separately and then compared the
that normalization of B vitamins and homocysteine metabolism in results to detect possibly confounded associations.
brain would significantly reduce the risk of dementia in PD, which There are no differences in several disorders that impact
warrants further research. We have also found multiple levodopa- metabolism neither in the frequency of apolipoprotein E ε4 allele,
related associations, mainly in lipids, and many of the changes which is consistent with the histological confirmation that
exhibit an apparent temporal delay in PD-D. This interaction and cognitive impairment of the PD subjects is unrelated to AD
its nature should be further studied and validated since its (where the allele frequency is several times higher91). Importantly,
omission may mask the levodopa effects if confirmed. post-mortem collection intervals are very homogeneous between
the groups. Freezer storage since collection is somewhat higher
for the control samples, but we did not observe any signs of
METHODS related tissue degradation (e.g. choline levels of the longer stored
This is a targeted metabolomic case-control study in PD with post- control samples were indistinguishable from those with shorter
mortem human brain tissue using liquid chromatography and storage; two-tailed Welch’s t-test p-value = 0.72). Besides, we
mass spectrometry. controlled for any freezer storage effect in the regression
alongside other covariates.
Study subjects
Ethics statement
From the Banner Sun Health Research Institute brain bank (Sun
City, AZ, USA)87, we obtained post-mortem brain frontal cortex The authors purchased all tissue samples from the Banner Sun
samples (Brodmann area 9; 500 mg) of 36 cognitively normal Health Research Institute brain bank (Sun City, AZ, USA), which
controls and 65 PD subjects at various stages of cognitive manages tissue collection and deposition, records written
impairment: 32 with dementia (PD-D) and 33 without dementia informed consents from all subjects prior to their death, and
(PD-ND) subdivided into 19 with MCI (PD-MCI) and 14 cognitively pledges to perform all methods in accordance with relevant
normal (PD-CN). We also obtained post-mortem putamen samples guidelines and regulations under approval by the Banner Sun
(20 mg) of the identical 35 cognitively normal controls (1 sample Health Institutional Review Board87. The authors obtained prior
was destroyed during analysis) and 33 PD-ND subjects. The approval from the Banner Sun Health Research Institute brain
numbers were based on available tissue in the biobank with bank to use the tissue samples for research purposes and adhered
required characteristics and post-mortem collection interval range, to relevant guidelines and regulations.
combined with allocated funding, and exceed the size of most
studies with human brain tissue. The samples were collected Chromatography and mass spectrometry
between 2004 and 2018 from deceased donors, flash frozen with Targeted metabolomic analysis was based on triple quadrupole
post-mortem collection interval averaging 3 h, and continuously ultra-high-performance liquid chromatography tandem mass
stored at −80 °C. The diagnosis followed clinical records and spectrometry (UHPLC-MS/MS) using Shimadzu Nexera chromato-
histopathological examination: PD subjects had two of the three graphy platform (Shimadzu Corporation, Kyoto, Japan) coupled to
cardinal clinical signs of resting tremor, muscular rigidity, and Sciex QTrap 5500 mass spectrometer (AB Sciex LLC, Framingham,
bradykinesia, along with pigmented neuron loss and presence of Massachusetts, USA). We applied the Biocrates MxP Quant 500
Lewy bodies in substantia nigra, and were treated with levodopa- targeted kit (Biocrates Life Sciences AG, Innsbruck, Austria),
carbidopa medication. The status of dementia and MCI corre- potentially quantitating 106 small molecules and free fatty acids
sponds to clinical diagnosis using a global Clinical Dementia in chromatography mode and 524 complex lipids in positive flow-
Rating scale88, defining MCI as score 0.5 and dementia as score 1 injection mode (FIA-MS/MS), exploring a broad range of metabolic
or higher. To differentiate from Alzheimer’s type of dementia, the pathways. Annotations for the individual metabolites with
subjects scored no more than “low likelihood of AD” in NIA- identifiers to external databases are provided in Supplementary
Reagan classification89. Controls were without a history of File 3. However, Biocrates prefers to keep mass transitions of
cognitive impairment and parkinsonism. All subjects were White individual metabolites undisclosed as proprietary knowledge.
12
Additionally, 232 metabolic indicators were calculated from this new extension of median normalization is provided in Eqs. (1)
sums or ratios of relevant metabolites according to Biocrates to (3).
MetaboINDICATOR formulas92 (see Supplementary File 3). We refer
ðmÞ
to the whole set of metabolites and metabolic indicators as 8m; s; m 2 M; s 2 S : r ðmÞ
s ¼ x sðmÞ = median x t (1)
ðg Þ s
t2S
analytes. The indicators can be regarded as physiologically

relevant measures and are statistically analyzed separately from ðmÞ ðmÞ
8m; p; m 2 M; p 2 P : qpðmÞ ¼ median r t = median r t (2)
metabolites. The indicators denoted as “X synthesis” are t2Sp t2S
computed as a ratio of metabolite X and its main precursors in
an attempt to reflect the conversion ratio. Since there are multiple 8m; s; m 2 M; s 2 S : y ðsmÞ ¼ x ðsmÞ =qpðmÞ
s
(3)
explanations why such an indicator could be altered, the
interpretation needs to be done cautiously in context of the Where: M – set of all metabolites
individual metabolites. P – set of all plates
Brain cortex samples were extracted in plastic vials with 85% S – set of all samples (subject samples and QCs)
ethanol in phosphate-buffered saline at concentration 3 µl/mg, SðgÞ
p – all samples from plate p and group g
homogenized with sonicator, and centrifuged for 20 minutes. The gs – group of sample s (one of the subject groups or QC)
clear extract (2 × 15 µl for cortex, 15 µl for putamen) was ps – plate of sample s
transferred onto a kit plate with pre-injected internal standards qp – normalization quotient for plate p
ðmÞ
and dried down. In brief, the rest of the assay includes x s – original value of metabolite m for sample s
ðmÞ
derivatization with 5% phenylisothiocyanate in pyridine, ethanol r s – pre-normalized reference value for metabolite m and
and water (1:1:1), and subsequent extraction with 5 mM ammo- sample s
ðmÞ
nium acetate in methanol. Chromatography was done with 0.2% y s – normalized value of metabolite m for sample s
formic acid in acetonitrile (organic mobile phase) and 0.2% formic
acid in water (inorganic mobile phase). Flow-injection analysis was Limits of detection (LODs). LODs were calculated as 2× median
performed with methanol and Biocrates MxP Quant 500 additive signal in blanks. Metabolites with more than 50% values below
of undisclosed composition. All solvents used were of LC/UHPLC- LOD in all subject groups were filtered out. Values below LOD
MS grade, except for ethanol with the American Chemical Society were not adjusted, since they represent the best estimate of the
and United States Pharmacopeia grade. true values. However, completely zero values were interpolated as
Sample handling was done on dry ice to avoid multiple freeze- half of the minimal non-zero value for a given metabolite to avoid
thaw cycles. We randomized the samples across plates, with strict zeros, since zero values are more difficultly handled by
stratification, already prior to their processing to avoid any subsequent transformations as well as biologically unlikely.
accidental bias towards one of the subject groups. Plates included
blanks to calculate limits of detection, repeats of a kit quality Calculated analytes. Metabolic indicators were calculated accord-
control sample to calculate concentrations and monitor the ing to Biocrates MetaboINDICATOR formulas92. Ratios with zeros
coefficient of variation (median for analyzed compounds: <10% were treated as missing values and not included in the analysis.
for UHPLC, <20% for FIA), and kit calibrators for seven-point The metabolic indicators were also plate-normalized.
calibrations of certain compounds.
Data transformation. In R environment93,94, we applied Box-Cox
transformation with R package car95 to better approximate
Data preprocessing Gaussian distributions. Tukey’s fencing96 was used to adjust
Peak areas and concentrations. Mass spectrometry signal was remote outliers (k = 3) to protect against skewing the means by
acquired in Sciex Analyst v1.6.24 (AB Sciex LLC, Framingham, extreme values while not reducing the variance greatly compared
Massachusetts, USA) and chromatographic peaks were identified to outlier removal. Finally, the values were standardized with
and integrated in Biocrates MetIDQ Oxygen-DB110-3005 (Bio- respect to control samples to facilitate comparison of regression
crates Life Sciences AG, Innsbruck, Austria). We have reviewed the coefficients in the statistical analysis.
integration process using the same software by checking
integration boundaries for captured signal of every metabolite Missing values. The statistical analysis requires all regressors to
and internal standard across the samples to confirm that whole be non-missing. Therefore, several missing sociodemographic
peaks were captured and that shared peaks were integrated values were imputed: Missing body mass index (BMI) values of
consistently. Areas of metabolite peaks were divided by areas of 4 subjects and length of education of 4 subjects were imputed as
their respective internal standards (Supplementary File 3). Further a mean value conditional on the subject group and sex. The
processing was done in R v3.6.193 with RStudio v1.2.503394. For imputed values were distributed across the groups and sex as
most compounds, concentrations were estimated linearly from follows: BMI – 1× controls male (5%), 1× controls female (7%), and
expected concentrations in the quality control sample using their 2× PD-MCI male (14%); education – 2× controls male (9%), 1×
median. Seven-point quadratic calibration was applied where controls female (7%), and 1× PD-MCI male (7%). The imputation
possible. did not substantially affect the results compared to completely
removing the samples, but it achieves higher statistical power.
Plate normalization. Cortex samples were run in two plates and
to account for potential batch effects, plates were normalized (per
Statistical analysis
metabolite) by scaling through median normalization: For a given
metabolite, values of reference samples in each plate are scaled by Sociodemographic and clinical characteristics. Key characteristics
such a factor so that their median is equivalent to the median of of subjects and samples in each group were compared with two-
values of all reference samples before normalization. For this tailed Fisher’s exact test for binomial variables and analysis of
purpose, we were able to leverage information from quality variance (ANOVA) on a linear model constructed with the R
control samples as well as all human samples owing to a simple package nlme97 for continuous variables, not assuming equiva-
trick: A copy of values in each subject group and QCs were first lence of variance among the subject groups.
divided by the respective group median, upon which all of these
pre-normalized samples were used as the reference samples for Differential analysis. The analysis of differences in analytes
estimation of the normalization factors. The formal definition of between PD groups and controls and the effect of levodopa
13
medication was based on a series of multivariable linear Pathway analysis. We downloaded definitions of human meta-
regression models with R package nlme97, one for each analyte, bolic pathways from KEGG database20 as publicly available on
with its values as dependent variables and subject groups – either December 07, 2021 and matched them with the measured
separately by cognitive status, or all PD together – followed by the metabolites. Since certain measurements in the performed assay
indicator of acute levodopa presence in PD-ND and in PD-D may represent multiple isoforms undistinguishable by the mass
(cortex only) as independent variables, without the assumption of spectra and each isoform can have its own annotations and
equivalence of variance among the groups. The models further pathway memberships, we accounted for this by assigning the
included covariates for age, sex, education, BMI, diagnosis of measured metabolites into all pathways with any of the possible
several frequent disorders affecting metabolism: hyperlipidemia, isoforms of the metabolite. Several metabolites remained unas-
diabetes mellitus, renal insufficiency, and hypothyroidism, as well signed to any pathway, especially the ones related to microbial
as post-mortem collection interval and the total length of freezer activity. Therefore, we created a custom metabolite set with only
storage. The last two covariates were log-transformed to be able microbial metabolites (indoles, 5-aminovaleric acid, trimethyla-
to capture any time-related exponential decay. Due to standardi- mine N-oxide, para-cresol sulfate, and secondary bile acids). Only
zation, the regression coefficients have a unit of 1 standard metabolic pathways with 4 or more assigned metabolites were
deviation of the distribution of controls. considered. We followed the statistical approach of ChemRICH set
enrichment analysis99, which relies on application of one-sided
Acute levodopa presence. We define the acute levodopa medica- Kolmogorov-Smirnov test over the distribution of p-values of
tion presence (L+) as elevated DOPA levels in PD subjects, which metabolites assigned to the same set (pathway) using the uniform
happens due to levodopa medication, and absence (L−) as distribution as a reference. The advantage of this approach is that
physiological DOPA levels, which may also be partially contributed the test is done over p-values, which can be obtained from any
by the medication but within the physiological range. The comparative model, in our case the main regression model, so the
physiological threshold was estimated from normalized DOPA covariates are considered. This is in contrast with currently
concentrations as 95% quantile of the values of controls in the available pathway tools, which, besides having problems with
respective brain region. We have previously established that the pairing multiple isoforms to a single measurement, cannot include
effect of acute levodopa medication presence differs in demented covariates in the analysis, resulting in less effective analysis or
and in non-demented subjects in this cohort14, so we included this potentially false positive results. We also performed FDR control
interaction in the differential analysis as two regressors – Boolean via q-values98.
indicators of acute levodopa presence in PD-D subjects and
separately in PD-ND subjects. Given that all subjects were AUC analysis. We performed a simple univariate AUC analysis
chronically taking the medication, we expected that the acute using R package pROC100 to distinguish between PD subjects with
levodopa presence at the time of death is a result of a random and without dementia, with and without interaction with acute
process and not directly related to disease progression scores, levodopa presence. The best score is reported. Confidence
thus resulting in no differences in the scores between L+ and L− intervals are computed using the default DeLong method.
PD-D/ND subjects. We have verified this assumption using two-
tailed Welch’s t-test and found no significant differences except Reporting summary
for a potential difference in senile plaque score in the PD-D group
Further information on research design is available in the Nature
(L+ vs L− p = 0.04) although observing such a difference is
Research Reporting Summary linked to this article.
expected due to multiple hypotheses testing (FDR = 0.50) and is
consistent with the random process assumption. There were no
significant associations with senile plaque score in cortex, so no DATA AVAILABILITY
confounding effect is suspected. The authors declare that the data supporting the findings of this study are available
within the manuscript: Measured area ratios along with calculated concentrations
Progression analysis. Associations between scores of disease and metabolic indicators as well as sociodemographic and clinical information are
progression and cognitive decline among PD subjects followed provided in the Supplementary File 4.
a procedure similar to the differential analysis, with the
independent variable being a progression score instead of subject
groups and levodopa presence. The regression models were CODE AVAILABILITY
constructed so that the controls contributed to the estimation of Programming code and scripts used for data processing and statistical analysis are
the effect of covariates but not the progression score. Regression available upon request for the purpose of replication of the study results and can be
coefficients are two-way standardized: 1 unit corresponds to the further used or shared only after obtaining an approval from the corresponding
change in the value of the analyte by 1 SD of the values of controls author.
with the progression score being changed by 1 SD of values of PD
subjects. Received: 22 October 2022; Accepted: 22 May 2023;
False discovery rate (FDR) control. For each regression coefficient

of interest, two-tailed p-values across the models were controlled
for FDR using the q-value procedure with R package qvalue98. The
control was done separately for each combination of plate REFERENCES
analytical method (UHPLC/FIA) and analyte type (measured/ 1. GBD 2016 Parkinson’s Disease Collaborators. Global, regional, and national
calculated). Effects with FDR ≤ 0.05 were considered statistically burden of Parkinson’s disease, 1990–2016: A systematic analysis for the Global
Burden of Disease Study 2016. Lancet Neurol. 17, 939–953 (2018).
significant.
2. Aarsland, D. et al. Risk of dementia in Parkinson’s disease: A community-based,
prospective study. Neurology 56, 730–736 (2001).
Collinearity check. Assessment of collinearity among all regres- 3. Irwin, D. J., Lee, V. M.-Y. & Trojanowski, J. Q. Parkinson’s disease dementia:
sors was based on the magnitude of Pearson’s correlation Convergence of α-synuclein, tau and amyloid-β pathologies. Nat. Rev. Neurosci.
coefficients and adjusted generalized variable inflation factor 14, 626–636 (2013).
(GVIF) calculated with the R package car95. We found no evidence 4. Kalia, L. V. & Lang, A. E. Parkinson’s disease. Lancet 386, 896–912 (2015).
of significant collinearity (all Pearson’s |r | < 0.6 and adjusted GVIFs 5. Koros, C., Stefanis, L. & Scarmeas, N. Parkinsonism and dementia. J. Neurol. Sci.
< 1.5). 433, 120015 (2022).
14
6. Perry, T. L. & Yong, V. W. Idiopathic Parkinson’s disease, progressive supranuclear 34. Chen, S.-J. et al. Alteration of gut microbial metabolites in the systemic circu-
palsy and glutathione metabolism in the substantia nigra of patients. Neurosci. lation of patients with Parkinson’s disease. J. Parkinsons. Dis. 12, 1219–1230
Lett. 67, 269–274 (1986). (2022).
7. Scholefield, M. et al. Substantively lowered levels of pantothenic acid (vitamin 35. Bolliri, C. et al. Gut Microbiota in monozygotic twins discordant for Parkinson’s
B5) in several regions of the human brain in Parkinson’s disease dementia. disease. Ann. Neurol. 92, 631–636 (2022).
Metabolites 11, 569 (2021). 36. Shonsey, E. M. et al. Bile acid coenzyme A: Amino acid N-acyltransferase in the
8. Scholefield, M. et al. Severe and regionally widespread increases in tissue urea in amino acid conjugation of bile acids. Methods Enzymol. 400, 374–394 (2005).
the human brain represent a novel finding of pathogenic potential in Parkin- 37. Wang, J., Wang, F., Mai, D. & Qu, S. Molecular mechanisms of glutamate toxicity
son’s disease dementia. Front. Mol. Neurosci. 14, 711396 (2021). in Parkinson’s disease. Front. Neurosci. 14, 585584 (2020).
9. Vivó, M. et al. Polyamines in the basal ganglia of human brain. Influence of aging 38. Hawkins, R. A. The blood-brain barrier and glutamate. Am. J. Clin. Nutr. 90,
and degenerative movement disorders. Neurosci. Lett. 304, 107–111 (2001). 867S–874S (2009).
10. Cheng, D. et al. Lipid pathway alterations in Parkinson’s disease primary visual 39. Kato, S., Ishita, S., Sugawara, K. & Mawatari, K. Cystine/glutamate antiporter
cortex. PLoS One 6, e17299 (2011). expression in retinal Müller glial cells: implications for DL-alpha-aminoadipate
11. Wood, P. L., Tippireddy, S., Feriante, J. & Woltjer, R. L. Augmented frontal cortex toxicity. Neuroscience 57, 473–482 (1993).
diacylglycerol levels in Parkinson’s disease and Lewy Body Disease. PLoS One 13, 40. Hu, R. G., Lim, J., Donaldson, P. J. & Kalloniatis, M. Characterization of the cystine/
e0191815 (2018). glutamate transporter in the outer plexiform layer of the vertebrate retina. Eur. J.
12. Xicoy, H., Brouwers, J. F., Wieringa, B. & Martens, G. J. M. Explorative combined Neurosci. 28, 1491–1502 (2008).
lipid and transcriptomic profiling of substantia nigra and putamen in Parkin- 41. Ahn, J. et al. Retinal thinning associates with nigral dopaminergic loss in de
son’s disease. Cells 9, 1966 (2020). novo Parkinson disease. Neurology 91, e1003–e1012 (2018).
13. Beger, A. W., Dudzik, B., Woltjer, R. L. & Wood, P. L. Human brain lipidomics: Pilot 42. Vallerga, C. L. et al. Analysis of DNA methylation associates the cystine-
analysis of the basal ganglia sphingolipidome in Parkinson’s disease and Lewy glutamate antiporter SLC7A11 with risk of Parkinson’s disease. Nat. Commun.
body disease. Metabolites 12, 187 (2022). 11, 1238 (2020).
14. Kalecký, K., Ashcraft, P. & Bottiglieri, T. One-carbon metabolism in Alzheimer’s 43. Nemeria, N. S. et al. The mitochondrial 2-oxoadipate and 2-oxoglutarate dehy-
disease and Parkinson’s disease brain tissue. Nutrients 14, 599 (2022). drogenase complexes share their E2 and E3 components for their function and
15. Seshadri, S. et al. Plasma homocysteine as a risk factor for dementia and Alz- both generate reactive oxygen species. Free Radic. Biol. Med. 115, 136–145
heimer’s disease. N. Engl. J. Med. 346, 476–483 (2002). (2018).
16. Müller, T., Renger, K. & Kuhn, W. Levodopa-associated increase of homo- 44. Chaudhary, S. et al. Frontal lobe metabolic alterations characterizing Parkinson’s
cysteine levels and sural axonal neurodegeneration. Arch. Neurol. 61, 657–660 disease cognitive impairment. Neurol. Sci. 42, 1053–1064 (2021).
(2004). 45. Gomes-Trolin, C., Nygren, I., Aquilonius, S.-M. & Askmark, H. Increased red blood
17. O’Suilleabhain, P. E., Bottiglieri, T., Dewey, R. B. Jr, Sharma, S. & Diaz-Arrastia, R. cell polyamines in ALS and Parkinson’s disease. Exp. Neurol. 177, 515–520
Modest increase in plasma homocysteine follows levodopa initiation in Par- (2002).
kinson’s disease. Mov. Disord. 19, 1403–1408 (2004). 46. Plewa, S. et al. The metabolomic approach reveals the alteration in human
18. Shin, J. Y. et al. Elevated homocysteine by levodopa is detrimental to neuro- serum and cerebrospinal fluid composition in Parkinson’s disease patients.
genesis in parkinsonian model. PLoS One 7, e50496 (2012). Pharm. (Basel) 14, 935 (2021).
19. Bottiglieri, T. et al. Acute administration of L-DOPA induces changes in 47. Paik, M.-J. et al. Polyamine patterns in the cerebrospinal fluid of patients with
methylation metabolites, reduced protein phosphatase 2 A methylation, and Parkinson’s disease and multiple system atrophy. Clin. Chim. Acta 411,
hyperphosphorylation of Tau protein in mouse brain. J. Neurosci. 32, 9173–9181 1532–1535 (2010).
(2012). 48. Klatt, S. et al. A six-metabolite panel as potential blood-based biomarkers for
20. Kanehisa, M., Furumichi, M., Tanabe, M., Sato, Y. & Morishima, K. KEGG: New Parkinson’s disease. NPJ Parkinsons Dis. 7, 94 (2021).
perspectives on genomes, pathways, diseases and drugs. Nucleic Acids Res 45, 49. Han, W., Sapkota, S., Camicioli, R., Dixon, R. A. & Li, L. Profiling novel metabolic
D353–D361 (2017). biomarkers for Parkinson’s disease using in-depth metabolomic analysis. Mov.
21. Movement Disorder Society Task Force on Rating Scales for Parkinson’s Disease. Disord. 32, 1720–1728 (2017).
The Unified Parkinson’s Disease Rating Scale (UPDRS): Status and recommen- 50. Murakami, K. et al. Dietary intake of folate, vitamin B6, vitamin B12 and ribo-
dations. Mov. Disord. 18, 738–750 (2003). flavin and risk of Parkinson’s disease: a case-control study in Japan. Br. J. Nutr.
22. Adler, C. H. et al. Unified Staging System for Lewy Body disorders: Clin- 104, 757–764 (2010).
icopathologic correlations and comparison to Braak staging. J. Neuropathol. Exp. 51. Zhao, G. et al. Betaine in inflammation: Mechanistic aspects and applications.
Neurol. 78, 891–899 (2019). Front. Immunol. 9, 1070 (2018).
23. Folstein, M. F., Folstein, S. E. & McHugh, P. R. “Mini-mental state”. A practical 52. Chai, G.-S. et al. Betaine attenuates Alzheimer-like pathological changes and
method for grading the cognitive state of patients for the clinician. J. Psychiatr. memory deficits induced by homocysteine. J. Neurochem. 124, 388–396 (2013).
Res. 12, 189–198 (1975). 53. Sun, J., Wen, S., Zhou, J. & Ding, S. Association between malnutrition and
24. Frouni, I., Kwan, C., Belliveau, S. & Huot, P. Cognition and serotonin in Parkinson’s hyperhomocysteine in Alzheimer’s disease patients and diet intervention of
disease. Prog. Brain Res. 269, 373–403 (2022). betaine. J. Clin. Lab. Anal. 31, e22090 (2017).
25. Fisher, J. R., Wallace, C. E., Tripoli, D. L., Sheline, Y. I. & Cirrito, J. R. Redundant Gs- 54. Nagatsu, T., Nakashima, A., Ichinose, H. & Kobayashi, K. Human tyrosine
coupled serotonin receptors regulate amyloid-β metabolism in vivo. Mol. Neu- hydroxylase in Parkinson’s disease and in related disorders. J. Neural Transm.
rodegener. 11, 45 (2016). (Vienna) 126, 397–409 (2019).
26. Stansley, B. J. & Yamamoto, B. K. L-dopa and brain serotonin system dysfunction. 55. Irino, Y. et al. 2-Aminobutyric acid modulates glutathione homeostasis in the
Toxics 3, 75–88 (2015). myocardium. Sci. Rep. 6, 36749 (2016).
27. Al-Bachari, S., Naish, J. H., Parker, G. J. M., Emsley, H. C. A. & Parkes, L. M. Blood- 56. Molina, J. A. et al. Decreased cerebrospinal fluid levels of neutral and basic
brain barrier leakage is increased in Parkinson’s disease. Front. Physiol. 11, amino acids in patients with Parkinson’s disease. J. Neurol. Sci. 150, 123–127
593026 (2020). (1997).
28. Lucas, L. N. et al. Dominant bacterial phyla from the human gut show wide- 57. Robinson, S. R. Neuronal expression of glutamine synthetase in Alzheimer’s
spread ability to transform and conjugate bile acids. mSystems 6, e0080521 disease indicates a profound impairment of metabolic interactions with astro-
(2021). cytes. Neurochem. Int. 36, 471–482 (2000).
29. Guzior, D. V. & Quinn, R. A. Review: microbial transformations of human bile 58. Zou, J. et al. Glutamine synthetase down-regulation reduces astrocyte protec-
acids. Microbiome 9, 140 (2021). tion against glutamate excitotoxicity to neurons. Neurochem. Int. 56, 577–584
30. Kalecký, K., German, D. C., Montillo, A. A. & Bottiglieri, T. Targeted metabolomic (2010).
analysis in Alzheimer’s disease plasma and brain tissue in non-Hispanic whites. J. 59. Streubel-Gallasch, L. et al. Parkinson’s disease-associated LRRK2 interferes with
Alzheimers Dis. 86, 1875–1895 (2022). astrocyte-mediated alpha-synuclein clearance. Mol. Neurobiol. 58, 3119–3140
31. Moustafa, S. A. et al. Gut brain axis: An insight into microbiota role in Parkinson’s (2021).
disease. Metab. Brain Dis. 36, 1545–1557 (2021). 60. Fedotova, E. I., Dolgacheva, L. P., Abramov, A. Y. & Berezhnov, A. V. Lactate and
32. Li, P. et al. Gut Microbiota dysbiosis is associated with elevated bile acids in pyruvate activate autophagy and mitophagy that protect cells in toxic model of
Parkinson’s disease. Metabolites 11, 29 (2021). Parkinson’s disease. Mol. Neurobiol. 59, 177–190 (2022).
33. Shao, Y. et al. Comprehensive metabolic profiling of Parkinson’s disease by 61. Grimm, M. O. W. et al. Docosahexaenoic acid reduces amyloid beta production
liquid chromatography-mass spectrometry. Mol. Neurodegener. 16, 4 (2021). via multiple pleiotropic mechanisms. J. Biol. Chem. 286, 14028–14039 (2011).
15
62. Knebel, L. A. et al. 2-Methylbutyrylglycine induces lipid oxidative damage and 88. Hughes, C. P., Berg, L., Danziger, W. L., Coben, L. A. & Martin, R. L. A new clinical
decreases the antioxidant defenses in rat brain. Brain Res 1478, 74–82 (2012). scale for the staging of dementia. Br. J. Psychiatry 140, 566–572 (1982).
63. Olanow, C. W. Levodopa: effect on cell death and the natural history of Par- 89. Consensus recommendations for the postmortem diagnosis of Alzheimer’s
kinson’s disease: Levodopa and the Natural History of PD. Mov. Disord. 30, 37–44 disease. The National Institute on Aging, and Reagan Institute Working Group
(2015). on Diagnostic Criteria for the Neuropathological Assessment of Alzheimer’s
64. Sirtori, C. R., Bolme, P. & Azarnoff, D. L. Metabolic responses to acute and chronic Disease. Neurobiol. Aging 18, S1–S2 (1997).
L-dopa administration in patients with parkinsonism. N. Engl. J. Med. 287, 90. Mirra, S. S. et al. The Consortium to Establish a Registry for Alzheimer’s Disease
729–733 (1972). (CERAD). Part II. Standardization of the neuropathologic assessment of Alzhei-
65. Adams, F. et al. Influences of levodopa on adipose tissue and skeletal muscle mer’s disease. Neurology 41, 479–486 (1991).
metabolism in patients with idiopathic Parkinson’s disease. Eur. J. Clin. Phar- 91. Crean, S. et al. Apolipoprotein E ε4 prevalence in Alzheimer’s disease patients
macol. 64, 863–870 (2008). varies across global populations: A systematic literature review and meta-
66. Tanaka, M., Yoshida, T., Okamoto, K. & Hirai, S. Dopamine and DOPA cause analysis. Dement. Geriatr. Cogn. Disord. 31, 20–30 (2011).
release of iron from ferritin and lipid peroxidation of liposomes. Neuroreport 10, 92. Limonciel, A. et al. MetaboINDICATOR™: Translate Metabolomics & Lipidomics
1883–1887 (1999). into Knowledge. MetaboNews 10, 3–5 (2020). http://www.metabonews.ca/
67. Rodgers, K. J., Hume, P. M., Dunlop, R. A. & Dean, R. T. Biosynthesis and turnover Feb2020/MetaboNews_Feb2020.pdf.
of DOPA-containing proteins by human cells. Free Radic. Biol. Med. 37, 93. Ripley, B. D. The R project in statistical computing. MSOR connect 1, 23–25
1756–1764 (2004). (2001).
68. Giannopoulos, S., Samardzic, K., Raymond, B. B. A., Djordjevic, S. P. & Rodgers, K. 94. RStudio Team. RStudio: Integrated Development for R. (RStudio, Inc.), (2019).
J. L-DOPA causes mitochondrial dysfunction in vitro: A novel mechanism of http://www.rstudio.com
L-DOPA toxicity uncovered. Int. J. Biochem. Cell Biol. 117, 105624 (2019). 95. Fox, J. & Weisberg, S. An R companion to applied regression. (SAGE Publications),
69. Rojo-Sebastián, A., González-Robles, C. & García de Yébenes, J. Vitamin B6 (2018).
deficiency in patients with Parkinson disease treated with levodopa/carbidopa. 96. Tukey, J. W. Exploratory Data Analysis. (Addison-Wesley), (1977).
Clin. Neuropharmacol. 43, 151–157 (2020). 97. Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D. & R Core Team. nlme: Linear and
70. Olaso-Gonzalez, G. et al. Impact of supplementation with vitamins B6, B12, and/ Nonlinear Mixed Effects Models [R Package Version 3.1-152]. The Comprehensive
or folic acid on the reduction of homocysteine levels in patients with mild R Archive Network https://CRAN.R-project.org/package=nlme (2021).
cognitive impairment: A systematic review. IUBMB Life 74, 74–84 (2022). 98. Storey, J. D., Bass, A. J., Dabney, A. & Robinson, D. Qvalue: Q-value estimation for
71. Lai, S.-C. et al. The transcobalamin receptor knockout mouse: a model for false discovery rate control [R package qvalue version 2.18.0]. GitHub http://
vitamin B12 deficiency in the central nervous system. FASEB J. 27, 2468–2475 github.com/jdstorey/qvalue (2019).
(2013). 99. Barupal, D. K. & Fiehn, O. Chemical Similarity Enrichment Analysis (ChemRICH) as
72. Murakami, M., Nakatani, Y., Atsumi, G.-I., Inoue, K. & Kudo, I. Regulatory functions alternative to biochemical pathway mapping for metabolomic datasets. Sci. Rep.
of phospholipase A2. Crit. Rev. Immunol. 37, 127–195 (2017). 7, 14567 (2017).
73. Kalous, M., Rauchová, H. & Drahota, Z. The effect of lysophosphatidylcholine on 100. Robin, X. et al. pROC: an open-source package for R and S+ to analyze and
the activity of various mitochondrial enzymes. Biochim. Biophys. Acta. 1098, compare ROC curves. BMC Bioinforma. 12, 77 (2011).
167–171 (1992).
74. Cha, M. H., Lee, S. M. & Jung, J. Lysophosphatidylcholine induces expression of
genes involved in cholesterol biosynthesis in THP-1 derived macrophages. ACKNOWLEDGEMENTS
Steroids 139, 28–34 (2018). This study was made possible, in part, through funding received from the Aging Mind
75. Saiki, S. et al. Decreased long-chain acylcarnitines from insufficient β-oxidation
Foundation Dallas (T.B.), by the Barbara Wallace and Kelly King Charitable Foundation
as potential early diagnostic markers for Parkinson’s disease. Sci. Rep. 7, 7328
Trust (T.B.), and Baylor Scott & White Foundation, Dallas, Texas (T.B.). We are grateful
(2017).
to the Banner Sun Health Research Institute Brain and Body Donation Program of Sun
76. Gou, Y. et al. Homocysteine restrains hippocampal neurogenesis in focal
ischemic rat brain by inhibiting DNA methylation. Neurochem. Int. 147, 105065 City, Arizona, for the sale of human brain tissue. The Brain and Body Donation
(2021). Program is supported by the National Institute of Neurological Disorders and Stroke
77. Elshorbagy, A. K., Kozich, V., Smith, A. D. & Refsum, H. Cysteine and obesity: (U24 NS072026 National Brain and Tissue Resource for Parkinson’s Disease and
consistency of the evidence across epidemiologic, animal and cellular studies. Related Disorders), the National Institute on Aging (P30 AG19610 Arizona Alzheimer’s
Curr. Opin. Clin. Nutr. Metab. Care 15, 49–57 (2012). Disease Core Center), the Arizona Department of Health Services (contract 211002,
78. Yideng, J. et al. Homocysteine-mediated PPARalpha,gamma DNA methylation Arizona Alzheimer’s Research Center), the Arizona Biomedical Research Commission
and its potential pathogenic mechanism in monocytes. DNA Cell Biol. 27, (contracts 4001, 0011, 05-901 and 1001 to the Arizona Parkinson’s Disease
143–150 (2008). Consortium), and the Michael J. Fox Foundation for Parkinson’s Research.
79. Khayati, K. et al. The amino acid metabolite homocysteine activates mTORC1 to
inhibit autophagy and form abnormal proteins in human neurons and mice.
FASEB J. 31, 598–609 (2017). AUTHOR CONTRIBUTIONS
80. Tan, H. W. S. et al. Lysosomal inhibition attenuates peroxisomal gene tran-
1. Research Project: A. Conception, B. Organization, C. Execution; 2. Statistical Analysis:
scription via suppression of PPARA and PPARGC1A levels. Autophagy 15,
A. Design, B. Execution, C. Review and Critique; 3. Manuscript Preparation: A. Writing
1455–1459 (2019).
of the First Draft, B. Editing, C. Review and Critique. Karel Kalecký: 1BC, 2ABC, 3ABC.
81. Zhang, H. et al. Polyamines control eIF5A hypusination, TFEB translation, and
autophagy to reverse B cell senescence. Mol. Cell 76, 110–125.e9 (2019). Teodoro Bottiglieri: 1AB, 2 C, 3BC.
82. Puleston, D. J. et al. Polyamines and eIF5A hypusination modulate mitochondrial
respiration and macrophage activation. Cell Metab. 30, 352–363.e8 (2019).
83. Masuoka, N., Lei, C., Li, H. & Hisatsune, T. Influence of imidazole-dipeptides on COMPETING INTERESTS
cognitive status and preservation in elders: A narrative review. Nutrients 13, 397 The authors declare no competing interests.
(2021).
84. Suidasari, S., Hasegawa, T., Yanaka, N. & Kato, N. Dietary supplemental vitamin
B6 increases carnosine and anserine concentrations in the heart of rats. ADDITIONAL INFORMATION
Springerplus 4, 280 (2015). Supplementary information The online version contains supplementary material
85. Kochlik, B., Gerbracht, C., Grune, T. & Weber, D. The influence of dietary habits available at https://doi.org/10.1038/s41531-023-00531-y.
and meat consumption on plasma 3-methylhistidine-A potential marker for
muscle protein turnover. Mol. Nutr. Food Res. 62, e1701062 (2018). Correspondence and requests for materials should be addressed to Karel Kalecký.
86. Sankowski, B. et al. Higher cerebrospinal fluid to plasma ratio of p-cresol sulfate
and indoxyl sulfate in patients with Parkinson’s disease. Clin. Chim. Acta. 501, Reprints and permission information is available at http://www.nature.com/
165–173 (2020). reprints
87. Beach, T. G. et al. Arizona Study of Aging and Neurodegenerative Disorders and
brain and Body Donation Program: Arizona brain and body donation program. Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims
Neuropathology 35, 354–389 (2015). in published maps and institutional affiliations.
16
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative
Commons license, and indicate if changes were made. The images or other third party
material in this article are included in the article’s Creative Commons license, unless
indicated otherwise in a credit line to the material. If material is not included in the
article’s Creative Commons license and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly
from the copyright holder. To view a copy of this license, visit http://
creativecommons.org/licenses/by/4.0/.
© The Author(s) 2023

Targeted Metabolomic Analysis in Parkinson's Disease Brain Frontal Cortex and Putamen With Relation To Cognitive Impairment

Uploaded by

Copyright:

Available Formats

Targeted Metabolomic Analysis in Parkinson's Disease Brain Frontal Cortex and Putamen With Relation To Cognitive Impairment

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Targeted Metabolomic Analysis in Parkinson's Disease Brain Frontal Cortex and Putamen With Relation To Cognitive Impairment

Uploaded by

Copyright:

Available Formats

www.nature.

Targeted metabolomic analysis in Parkinson’s disease brain

npj Parkinson’s Disease (2023)9:84 ; https://doi.org/10.1038/s41531-023-00531-y

INTRODUCTION quantitative investigation to be performed across numerous

Published in partnership with the Parkinson’s Foundation

Table 1. Metabolic set analysis in PD cognitive groups using KEGG pathways.

False discovery rate (FDR) control. For each regression coefﬁcient

© The Author(s) 2023

You might also like