Skip to main content
Download PDF

Metabolomics of blood reveals age-dependent pathways in Parkinson’s Disease

Cell & Bioscience volume 12, Article number: 102 (2022) Cite this article

Abstract

Background

Parkinson’s Disease (PD) is the second most frequent degenerative disorder, the risk of which increases with age. A preclinical PD diagnostic test does not exist. We identify PD blood metabolites and metabolic pathways significantly correlated with age to develop personalized age-dependent PD blood biomarkers.

Results

We found 33 metabolites producing a receiver operating characteristic (ROC) area under the curve (AUC) value of 97%. PCA revealed that they belong to three pathways with distinct age-dependent behavior: glycine, threonine and serine metabolism correlates with age only in PD patients; unsaturated fatty acids biosynthesis correlates with age only in a healthy control group; and, finally, tryptophan metabolism characterizes PD but does not correlate with age.

Conclusions

The targeted analysis of the blood metabolome proposed in this paper allowed to find specific age-related metabolites and metabolic pathways. The model offers a promising set of blood biomarkers for a personalized age-dependent approach to the early PD diagnosis.

Background

Parkinson’s disease (PD) is the second most frequent neurodegenerative disorder. The Global Burden of Disease Study estimates the prevalence of PD to be approximately 6.2 million people worldwide, with a prediction of nearly 13 million by 2040 [1]. Although the connection between the accumulation of α-synuclein and the PD pathology has been extensively studied [2], the multiple mechanisms contributing to its pathogenesis, including protein misfolding and aggregation, mitochondrial injury, oxidative stress and inflammation, are not fully understood yet. As a result, a PD diagnosis is only possible after the insurgence of mobility functions impairment, such as the well-known tremor, following a consistent loss of dopaminergic neurons in the substantia nigra and in the striatum. However, at sub-clinical stage, PD is accompanied by a non-specific symptomatology including depression, sleep disturbance, or loss of olfactive sense, which most often remain unaddressed, due to the absence of a sub-clinical diagnostic PD test [3, 4].

The understanding of the relationship between PD and age has the potential to play a significant role in the personalized early PD diagnosis [5]. Sub-clinical PD manifestation occurs early, up to 20 years before the first clinical symptoms, with a risk increasing with age. This suggests that a series of age-related disfunctions may contribute to the development of PD. Proteins accumulation, genetic factors, autophagy, mitophagy, and lower protection against oxidative stress have been proposed as possible age-related PD factors [6,7,8,9].

Against this background, the blood metabolome contains a rich source of information. In fact, the metabolic pathways affected by PD originate metabolites, which are released in the blood stream. The concentration of a series of amino acids, fatty acids, acylcarnitine, lipids, purines, organic acids, and sugars revealed PD-related disfunctions of the metabolism of branched chain amino acids, tryptophan, lipid, energy, and purine, and of oxidative stress/redox homeostasis metabolic pathway [10]. Blood metabolome reflects in-vivo physiological states influenced by genetics, epigenetics and lifestyle and is therefore intimately connected with the biological age. It has been found that antioxidants, nitrogen and muscle- or kidney-related elements indicate characteristic age-related accumulation and deficiencies of metabolites [11]. Although PD is an age-related neurodegenerative disorder [12], an evidence of age-related PD significant metabolite disorder has not been found until now. Besides blood, metabolites contained in saliva, cerebrospinal fluid (CSF) and sebum have been investigated. Although a series of metabolic pathways such as carnitine shuttle, valine, leucine and isoleucine degradation, fatty acid biosynthesis, sphingolipid, arachidonic acid, primary bile acids, fatty acids, ether lipids and vitamin E metabolism have been identified as possible metabolic pathways deeply related with PD, no significant correlation between the involved metabolites, PD diagnosis and age was observed [10, 13,14,15,16].

The main aim of this paper was to study a possible relationship between PD and age by identifying blood metabolites, which express a significant age-dependent pattern in PD. This would allow to develop personalized age-dependent biomarkers. We used a supervised machine learning approach to find novel blood PD biomarkers by comparing the blood metabolome of PD and healthy subjects. We further applied a statistical analysis based on unsupervised machine learning and Bonferroni correction to discover possible age correlations of the selected biomarkers. We finally applied an enrichment pathway analysis to the selected biomarkers to identify the metabolic pathways related to age, either in PD or in healthy patients.

Methods

Sample participants

The subjects included in this study were consecutively enrolled at the Centre for Research and Training in Medicine of Aging of the University of Molise. We recruited 39 patients affected by PD and 39 healthy controls (HC). The subjects followed the same Mediterranean alimentation, without any specific dietary requirement. The demographic and clinical characteristics of the two groups are described in Table 1. PD patients were included in the study if the following conditions were verified:

  1. 1.

    A “clinically established” diagnosis of PD according to the criteria published by the Movement Disorder Society (MDS) [17]

  2. 2.

    Mini Mental State Examination (MMSE) score higher than or equal to 24;

  3. 3.

    Clinical Dementia Rating (CDR) scale score lower than or equal to 6;

  4. 4.

    Treatment with L-DOPA for at least 3 months.

Table 1 Description of the participants included in the statistical analysis
Full size table

We assessed the degree of PD severity with the MDS-revised Unified Parkinson's Disease Rating Scale (MDS-UPDRS) and the Hoehn and Yahr scale [18]. The study was conducted in accordance with the ethical principles stated in the Declaration of Helsinki, and with approved national and international guidelines for human research. The Institutional Review Board (IRB) of the University of Molise reviewed and approved the study (IRB Prot. n. 17/2020). A written informed consent was obtained from each participant.

Sample collection

The blood sampling was performed between 8:00 and 8:30 a.m. after an overnight fasting of at least 8–10 h. Antecubital venous blood was collected in vacutainer tubes for plasma preparation (Becton & Dickinson, Milan, Italy) and immediately centrifuged to obtain plasma samples, which were stored at -80 °C until shipment to the analytical laboratory of BIOCRATES Life Sciences AG. The metabolomic approach based on mass spectrometry was used to obtain a quantitative determination of 630 endogenous metabolites, grouped in different biochemical classes for each plasma sample (MxP® Quant 500 kit). All pre-analytical and analytical procedures related to this project were performed, documented, and reviewed according to BIOCRATES Life Sciences AG’s ISO 9001: 2015 certified quality management guidelines and standards. We used the MxP® Quant 500 kit (Biocrates) for the quantification of several endogenous metabolites of various biochemical classes. Lipids and hexoses were measured by flow injection tandem mass spectrometry (FIA-MS/MS) analysis using a QTRAP® 5500 instrument (AB Sciex, Darmstadt, Germany) with an electrospray ionization (ESI) source. Other metabolites were measured by liquid chromatography coupled to tandem mass spectrometry (LC–MS/MS) also using the QTRAP® 5500 Instrument. A summary of the 630 metabolites measured for this study is reported in the Additional file 1: Table S1. We removed dyhydroxy-phenilalanine and tyrosine from the list of metabolites as they were clearly related to the specific therapy of PD patients.

Statistical analysis

We pre-processed the metabolomic dataset, by reducing each variable to zero average and unitary variance. We compared two approaches for the assessment of the variation of the metabolome between the two groups of subjects. Partial least squares-discriminant analysis (PLS-DA) is a regression technique for modeling the relationship between the metabolomic data block and the respective labeled categories (PD = + 1; HC = − 1) by maximizing their covariance [19]. The PLS algorithm reduces the input metabolomic data to npls independent components and a regression coefficient matrix is used to predict the labels [20]. We compared the PLS-DA model with the orthogonal projections to latent structures (OPLS) approach [21], which provides a higher level of sophistication. It separates in fact the metabolomic dataset in two different parts. The first one consists of nopls predictive components, which are correlated with the expected categories. The second one consists of a nort-dimensional orthogonal non-predictive block, which is generated by the intrinsic variability of the samples and is not correlated with the expected categories. We estimated the number of optimal components by performing permutation tests for regression metrics and two-tailed permutation tests for each metabolite relative to its loading. We calculated the P-value of the PLS-DA and OPLS models by using a resampling with replacement (bootstrapping) validation technique. It is known that the higher is the number of components used to model the metabolomic dataset, the higher will be the probability of overfitting. We adopted a K-fold cross-validation method by comparing the goodness of fit R2Y and the predictability Q2 parameters after a random variation of the labels [22]. If the R2Y parameter after permutations is systematically higher than in the case of the correct data, then the model exhibits an illogic overfitting ability to predict any random permutation of the labels. Finally, we selected the best predictive model and the number of components avoiding overfitting and we calculated for each metabolomic parameter the variable influence on projection (VIP). However, the use of VIP values in the context of the selection of significant parameters needs a dedicated validation. There are in fact two problems. The first one is related to the intrinsic statistical fluctuations of the VIP value itself. Although the R2Y and Q2 parameters are calculated above by random label permutation and bootstrapping, that is not enough to prevent random fluctuations of the VIPs due to the highly variable nature of the data. There is therefore a certain probability that repeating the same test with an independent dataset may provide significantly different VIP values. This is particularly true in OPLS models, which, as mentioned above, aim at separating the systematic variation contained in the data into two parts—a predictive part that is correlated to the labels and an orthogonal part that is uncorrelated to the labels. A second problem, related to the first one, is the number of subjects in the study, which reflects to the statistical significance of the findings. To address this issue, we followed a four-steps procedure:

  1. (1)

    Model validation we randomly selected 80% of the entire dataset to compose a training sample. As mentioned above, we performed a permutation test for regression metrics, and a two-tailed permutation test for each variable to its loading (L). VIP values are selected with P < 0.05 and |L|> 0.04. We performed the receiver operating characteristics (ROC) analysis based on the selected parameters and we estimated the 95% confidence interval of the area under the curve (AUC) by using a resample with replacing (bootstrapping) approach.

  2. (2)

    VIP values validation we calculated the VIP values of the selected parameters repeating the model training 1000 times. We used a resample with replacing (bootstrapping) approach by randomly sampling 80% of the entire dataset at each evaluation. We selected only these parameters which exhibit a VIP > 1.0 in more than 95% of the trials. This test allows to remove parameters which are mostly affected by statistical fluctuations.

  3. (3)

    VIP values reduction following [23], we further strengthened the selection of the significant parameters by applying a features reduction strategy based on the principal components analysis (PCA). We analyzed the group of parameters selected at step (2) and we identified the number Npca of independent principal components fi accounting for 80% of the total variance. We extracted the metabolites with the most important contribution to each fi, identified as the ones with PCA coefficient \(\left|{c}_{j}\right|>0.9\mathrm{max}\left|{c}_{i}\right|\).

  4. (4)

    Metabolites selection We finally selected as significant only the metabolites which have both a validated VIP in step (2) and a significant PCA coefficient in step (3).

Age dependence

We calculated the correlation between the principal components fi and the age of the patients by using a Spearman correlation statistical method. We selected only those significant principal components exhibiting a correlation coefficient |r|> 0.5 (P < 0.05/Npca). We applied a Bonferroni correction taking into account the multiple comparisons between independent statistical tests [24,25,26,27,28,29]. We finally verified that the metabolites selected in (4) also exhibit an age correlation in correspondence to the principal component in which they have a significant contribution. Data and software have been included in RADIOLYTX (www.radiolytx.com) and in the HPE Ezmeral platform (www.hpe.com/us/en/software.html).

Pathway analysis

We performed an enrichment analysis with the software MetaboAnalyst (Version 4.0). Significant features were mapped into the Kyoto Enciclopedia of Genes and Genomes (KEGG). Feature hits on known metabolite networks were tested against a null distribution produced from permutations to yield significance values of metabolites enriched within any given network [30]. We performed first the enrichment analysis on the entire set of selected and validated metabolites with VIP > 1.0. At a second step, we restricted the analysis to these principal components fi exhibiting a significant age dependence. We extracted the metabolites with the most important contribution to each fi, identified as the ones with PCA coefficient \(\left|{c}_{j}\right|>0.9\mathrm{max}\left|{c}_{i}\right|\). We applied the enrichment analysis to each group of significant metabolites characterizing each component fi and we identified possible age-dependent metabolic pathways for PD.

Results

Analysis of patient metadata

An overview of the clinical and demographic data of the patients is reported in Table 1. Two tailed t-test did not reveal any significant difference in age and BMI between PD patients and HC (P = 0.054 and P = 0.194, respectively). The Mini Mental State Examination (MMSE) score was found slightly significantly lower in PD group (P = 0.047). The ratio between men and women was 2:1 in the PD group, confirming the male prevalence of the disease reported in previous studies [31]. We selected a HC group with the same composition to avoid bias. Other indicators, such as smoke, scholarity, alcohol consumption, hypertension, diabetes, and dyslipidemia are not found significantly related to PD.

Data driven prediction of PD

We reported the scores plot of the first two components of a PLS-DA model with npls = 3 in Fig. 1A. The PD patients and HC subjects appeared well separated. The model had an acceptable R2Y = 0.76, but a very poor Q2 = 0.06. A resampling with validation replacement technique (n = 500) confirmed that the model described the metabolomic data with a P-value of 0.02 (Fig. 1B). The reason of such a poor predictability was due to a sizable overfitting. As shown in Fig. 1C, a random permutation of the label vectors did not affect the goodness of the fit and the R2Y value. We verified that lowering the number of components did not improve the performances of the PLS-DA technique.

Fig. 1
figure 1

Data driven modeling for the discrimination between PD and HC subjects. The PLS model exhibits a good separation between the two categories of subjects (A) and its goodness is apparently confirmed by a resampling with replacement validation technique (B). However, we observe a non-negligible overfitting (C). With respect to PLS, the OPLS model exhibits a better separation between the two classes (D), confirmed with a P-value of 9 × 10–4 (E) and absence of overfitting (F)

Full size image

The OPLS model with 3 orthogonal and 1 predictive component exhibited a better performance than PLS-DA, as qualitatively visible in the scores plot (Fig. 1D). Such an excellent disentanglement between PD and HC subjects was confirmed by an R2Y = 0.83 and a Q2 = 0.57. The model was validated with p-value of 9 × 10–4 (Fig. 1E). Finally, the goodness of fit and the R2Y values were significantly affected by random permutations of the labels, therefore confirming the absence of overfitting (Fig. 1F). In particular, the smaller was the similarity between the randomly permuted and the original labels, the smaller was the R2Y. We verified that increasing the number of orthogonal or predictive components would increase the probability of overfitting. We therefore concluded that OPLS with 3 orthogonal and 1 predictive component was the best suited model for the metabolomic data considered in this study.

Selection of significant metabolites

The four steps of the parameter selection are shown in Table 2 and in the Additional file 1: Table S2. At a first step, we used the variable importance in projection (VIP) parameter to select those metabolomic parameters playing a significant role in the discrimination between PD and HC group. A total of 320 metabolites exhibited VIP > 1. We performed a receiver operating characteristics (ROC) analysis of this model restricted to these significant parameters (Fig. 2A). The 95% confidence interval of the area under the curve (AUC) was (0.972, 1.0), confirming the excellent discrimination power of the OPLS predictive model.

Table 2 Selection of the significant metabolites
Full size table
Fig. 2
figure 2

The OPLS model with 3 orthogonal and 1 predictive components can distinguish between PD and HC with an AUC in the range (0.97, 1.0) at a 95% confidence level (A). The principal component analysis applied to the significant metabolic parameters with VIP > 1 identifies 5 components expressing up to 80% of the total variance. In particular, f0,1,4 does not exhibit any age dependence (B), f2 exhibits a age correlation only for PD patients (C) and f3 exhibits a age correlation only for HC subjects (D). Sphingomyelin (E), Choline (F) and Hexadecanoyl-carnitine (G) are among the main contributions to the three principal components

Full size image

However, we found that not all the 320 parameters have a significant contribution to the model. A large fraction of these parameters exhibits VIP > 1 only because of statistical fluctuations. As mentioned above, we validated the significance of the parameters using a resample with replacing (bootstrapping) approach by randomly sampling 80% of the entire dataset at each evaluation and repeating the model training 1000 times. Only 76 metabolic parameters, highlighted in Table 2 and Additional file 1: Table S2, exhibited a VIP > 1.0 in more than 95% of the cases.

We performed the final validation of the selected parameters based on an unsupervised Principal Components Analysis (PCA) technique. We found that 80% of the total variance of the selected parameters could be described with five principal components. The significant parameters with PCA coefficient \(\left|{c}_{j}\right|>0.9\mathrm{max}\left|{c}_{i}\right|\) are reported in Table 2 and in the Additional file 1: Table S2. The cross-validated parameters, exhibiting a validated VIP > 1 and a significant contribution to a PCA component, are 33 and are highlighted in Table 2 and in the Additional file 1: Table S2. They belong to the groups of acylcarnitines, amino acids, ceramides, cholesteryl esters, dihexosyl ceramides, fatty acids, hexosyl ceramides, lysophosphatidyl- and phosphatidylcholines, sphyngomielins, triglycerides, and vitamins.

Age dependence of metabolites

The principal components identified in the study define three distinct groups in relation to age. The first group is represented by the principal components f0, f1, and f4. Figure 2B shows the scatter plot of the value of f1 versus the age for PD and HC: no significant correlation was observed (r = 0.08, P = 0.96; r = 0.02, P = 0.98) for f1. A similar situation was encountered for f0 and f4. The second group, more interestingly, as shown in Fig. 2C, was represented by the component f2 accounting for 21% of the total variance. It exhibited a clear correlation with the age of PD patients (r = − 0.52; P = 0.007) but not of HC (r = 0.33; P = 0.09). The third group, finally, as shown in Fig. 2D, was represented by the component f3 and accounted for the 11% of the total variance. It exhibited a clear correlation with the age of HC subjects (r = 0.64; P = 5.6 × 10–4) but not of PD patients (r = 0.11; P = 0.56). The main parameters contributing to the linear expansion of the principal components are reported in Table 2. They were identified as the ones with coefficients \(\left|{c}_{j}\right|>0.9\mathrm{max}\left|{c}_{i}\right|\). For instance, sphingomyelin C16:0 (Fig. 2E), choline (Fig. 2F), and hexadecanoyl-carnitine (Fig. 2G) were significant examples of the three categories of metabolites.

Pathway analysis and age in PD

We performed pathway enrichment analysis to identify the effect of age in the metabolic pathways associated to PD. At a first step we performed the analysis to the entire set of identified significant metabolites, thus neglecting any age dependence. To provide an impression of the variability of the validated parameters used at this first step of the enrichment analysis, we report the box plot of their concentration in Fig. 3A. We notice that, with respect to HC subjects, PD patients have an increased level of ceramides (d18:1/24:1) and (d18:1/26:1), triacylglycerides (18:1/32:3) and (18:2/32:1), phosphatidyl-choline aa C34:2, and putrescine. sphingomyelin C22:3 followed an opposite trend, being reduced in PD patients. As reported in Fig. 3B, the enrichment analysis reveals that the biosynthesis of unsaturated fatty acids (P = 0.004), the tryptophan metabolism (P = 0.01) and the glycine, serine and threonine metabolism (P = 0.05) are the main metabolic pathways, which express the significant selected metabolites. To verify the age dependence of these three metabolic pathways, we selected the metabolic parameters which have a significant impact into each principal component. As visible in Table 2, the parameters can be separated into three completely disjoint groups corresponding to f1, f2 and f3, which address age correlation in no subjects, only in PD and only in HC, respectively. We performed a pathway analysis to the three groups distinctly and we found that metabolites not exhibiting any age correlation are associated with tryptophan metabolism. Similarly, metabolites exhibiting age correlation only in HC subjects are associated with the biosynthesis of unsaturated fatty acids. Finally, metabolites exhibiting age correlation only in PD subjects are associated with the glycine, serine and threonine metabolism.

Fig. 3
figure 3

Pathway analysis. The box plots of 8 significant metabolomic parameters validated in OPLS reveal a difference of the concentration between PD and HC subjects (A). An enrichment analysis verifies that the selected parameters can be associated with the biosynthesis of unsaturated fatty acids, the tryptophan metabolism and the glycine, serine and threonine metabolism (B). We verified that the metabolites composing the first and the third pathway are correlated with age only in HG and PD subjects, respectively

Full size image

Discussion

Significant metabolites as PD biomarkers

As summarized in Table 2, most of the significant metabolomic parameters presented different concentration levels in PD and HC subjects, following the expectations of the current knowledge regarding the metabolic changes in PD. For instance, a decreased level of acyl-carnitines has been found in possible association to primary decrement of mitochondrial β-oxidation and has been identified as a possible biomarker for PD [32]. Similarly, increased levels of trimethylamine N-oxide have been observed in PD patients. This amine oxide can cross the blood–brain barrier promoting cellular α-synuclein aggregations, neuroinflammation, mitochondrial dysfunction, and neuronal senescence [33]. As for amino acids, results are more controversial. For instance, our results confirm that no difference of glutamate concentration has to be expected in PD compared to HC [34], although an higher glutamate concentration has been indeed reported in PD patients [35, 36]. More interestingly, the importance of threonine as a possible PD biomarker is confirmed also in other studies [37]. Similarly, high levels of putrescine and ceramides in the blood of PD patients have also been observed [38, 39]. Polyamines, such as putrescine, cadaverine, spermidine and spermine, are involved in many vital processes, including cell proliferation and differentiation, gene transcription and translation, modulation of ion channels and receptors, and can promote the aggregation and fibrillation of α-synuclein [38]. Ceramides play an essential role not only in shaping cell membranes, but also in regulating cellular processes of vital importance in PD. This leads to the observation, supported and confirmed also in other studies [39], of elevated Ceramide levels in PD patients.

In general, the finding that cholesteryl esters, sphingomielins, fatty acids, dihexosylceramides, hexosylceramides, ceramides, phosphatidylcholines, lysophosphatidylcholines, diglycerides and triglycerides have a high impact in the discrimination between PD and HC indicates a general alteration of lipid metabolism in PD [40, 41], supporting the hypothesis of PD as a lipidopathy [42].

Finally, the increase of the tryglicerides levels observed in this study disagrees with the literature, which shows a reduction or no differences in PD patients compared to HC [40]. However, our result is in line with a post-mortem lipidomic study showing an increase of tryglicerides in the CSF of PD patients [43]. The ethnical and demographical characteristics of the subjects might be a possible explanation for the observed discrepancy [41].

Glycine, threonine, and serine metabolism correlates with age only in PD patients

A striking result of our study is that metabolites related to the glycine, threonine and serine metabolism exhibit a correlation with age in PD patients, but not in healthy subjects. It is well known that, in the production process of ATP from ADP, the creatine generated from threonine provides the necessary phosphate groups. As for PD, the relationship between glycine, serine and threonine metabolism has been intensively proved at the onset of α-synuclein aggregation, when glycine, serine and threonine metabolism appear to be down-regulated [44]. However, a direct observation of the age dependency of this metabolic pathway in PD has never been reported. The fact that age-dependent mechanisms may occur in PD is corroborated by the recently observed age-dependent neuromelanin production in PD [45]. Although there is no direct association between neuromelanin, threonine, glycine, and serotonin, this finding shows that age dependent mechanisms must be expected in PD. The fact that glycine, serine and threonine metabolism have a remarkable age-dependence only in PD patients suggests that this metabolic pathway might be associated with the neurodegenerative process typical of PD. Choline is the metabolic product of glycine, serine and threonine metabolism and exhibits a striking age dependence only in PD subjects, as mentioned above. Abnormal choline transport and metabolism have been implicated in several neurodegenerative disorders. It is an essential nutrient for all cells because it plays a critical role in the synthesis of the membrane phospholipid, as well as in the synthesis of the neurotransmitter acetylcholine. Its deficiency affects the expression of genes involved in cell proliferation, differentiation, and apoptosis, and it has been associated with liver dysfunction and cancer [46]. Indirect evidence of the association of glycine metabolism to age is supported by a series of independent results. By comparing young and old human fibroblasts it was found that epigenetic downregulation of the glycine-C-acetyltransferase (GCAT) and serine hydroxymethyltransferase 2 (SHMT2) genes involved in mitochondrial glycine synthesis correspond to the aging-related loss of cellular respiration. Interestingly, the phenotype of aged cells is restored back to young cells by adding glycine to the culture media [47, 48]. The role of serine synthesis in age-related diseases has been intensively studied. It has been shown that serine can directly affect lifespan through metabolic regulation [49].

Our results also open new possible exploration pathways in the relationship between PD and other chronic neurodegenerative diseases, such as Alzheimer’s Disease (AD). Glycine, serine, and threonine metabolism is in fact one of the six metabolic pathways that distinguish HC from AD patients [50]. Furthermore, metabolomic analysis has recently shown that the cognitive impairment due to post-traumatic brain injuries is associated with aberrations in glycine, serine and threonine metabolism [51]. When it comes to age dependency, preclinical studies may provide indirect evidence. A recent metabolomics analysis was performed on triple transgenic AD (3xTg-AD) 2- and 6-month-old mice. Relevant metabolites were identified following a statistical analysis like the one presented in this paper. Glycine, serine and threonine metabolism appeared as a significant metabolic pathway in the 6-month-old mice but not in the 2-month-old mice [52]. An imbalance between excitability (aspartate and glutamine) and inhibition (GABA and Glycine) may be a signature of AD. While this finding may suggest a dependence of glycine, serine and threonine metabolism on the progression of the disease, it is not in direct relationship with age.

More suggestive implications of our results in the intimate connections between AD and PD may be found with a deeper understanding of the spleen to brain connection [53]. The splenic nerve connects to the vagus nerve, which is connected to the brain stem. In a metabolomic study on AD mice models, it has been observed that impaired glycine, serine and threonine pathways were correlated with the increase in spleen size of AD mice at 6 months of age and in control mice at an age of 24 months. More interestingly, the age dependence of glycine, serine and threonine pathways does not exhibit any difference between healthy controls and AD after 24 months, therefore suggesting the absence of any age dependency at a later stage of the disease [54]. Following this observation, although our results enhance for the first time to our knowledge a clear age-dependence of the glycine, serine and threonine metabolic pathway in PD, it will be interesting to understand the strength of this effect at later stages of the disease.

Biosynthesis of fatty acids correlates with age only in healthy subjects

As mentioned above, the metabolites identified in this study support the strict relationship between PD and lipids. Alterations in the biosynthesis of fatty acids are not only indicative of a mitochondrial disfunction, but also of possible processes of mitophagy and apoptosis implied by the development of PD [55]. It has been observed how the integration of fatty acids omega-3 have a neuroprotective action in a model of hemiparkinsonism [56]. The effect of age on the biosynthesis of unsaturated fatty acids in the HC group is clearly connected with the emerging role of lipid metabolism [57]. For instance, age correlation of lipids and fatty acids has been found in the analysis of plasma samples from a cohort of 269 individuals [58]. The observation of a different age correlation between HC and PD agrees with other findings. By way of example, aldosterone, pantothenic acid, and Nacetyl-l-methionine were associated with age in HC; however, only 1 metabolite FFA 12:0 showed association with age in PD [59]. Therefore, the absence of correlation in PD patients may indicate an already compromised metabolism and needs further investigation.

Tryptophan metabolism characterizes PD but does not correlate with age

The role of tryptophan in PD is well-documented. Particularly kynurenine, a key intermediate in the breakdown of tryptophan and formation of nicotinamide adenine dinucleotide (NAD +) via the kynurenine pathway (KP), is involved in a variety of physiopathological processes and diseases—including cancer, autoimmune diseases, inflammatory diseases, neurologic diseases and psychiatric disorders [60]. KP metabolites, such as quilinolinic acid, cause neurotoxicity and consequently neuronal apoptosis and neurodegeneration, while others, such as the kynurenic acid, act as neuroprotectant. Furthermore, excess levels of quinolinic acid lead to the formation of metabolite assemblies that causes ∝-synuclein aggregation, with consequent neuronal toxicity and PD [61]. Metabolites associated with tryptophan metabolism also modulate inflammation, regulate energy homeostasis and control mental health [62]. PD patients show lower kynurenic acid and higher quinolinic acid levels compared to HC, especially in advanced stages of the disease [63]. When it comes to other neurodegenerative diseases, tryptophan metabolism has been found in significant relationship with both AD and Mild Cognitive Impairment (MCI) progression [64]. Contrary to previous studies exhibiting a robust association of tryptophan metabolism with aging [65], we did not observe any significant age dependency.

Conclusion

In conclusion, the targeted analysis of the blood metabolome proposed in this paper allowed to find specific age-related metabolites and metabolic pathways. The predictive OPLS model developed in this paper has an excellent discrimination power between PD and HC and offers a promising set of blood biomarkers for a personalized age-dependent approach to the early PD diagnosis.

Availability of data and materials

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. In addition, the datasets and the analysis software will be made available in the digital healthcare platform prepared by the GATEKEEPER consortium.

Abbreviations

ADP:

Adenosine diphosphate

ATP:

Adenosine triphosphate

AUC:

Area under the curve

CDR:

Clinical dementia rating

CSF:

Cerebrospinal fluid

HC:

Healthy control

IRB:

Institutional Review Board

KEGG:

Kyoto Enciclopedia of Genes and Genomes

KP:

Kynurenine pathway

MDS:

Movements Disorder Society

MDS-UPDRS:

MDS-revised Unified Parkinson’s Disease Rating Scale

MMSE:

Mini Mental State Examination

NAD:

Nicotinamide adenine dinucleotide

OPLS:

Ordered projection to latent structures

PCA:

Principal component analysis

PD:

Parkinson’s Disease

PLS-DA:

Partial least squares-discriminant analysis

ROC:

Receiver operating characteristics

TCA:

Tricarboxylic acid cycle

VIP:

Variable influence on projection

References

  1. Ray Dorsey E. Global, regional, and national burden of Parkinson’s disease, 1990–2016: a systematic analysis for the Global Burden of Disease Study 2016. Lancet Neurol. 2018;17:939–53.

    Article  Google Scholar 

  2. Spillantini MG, Goedert M. Neurodegeneration and the ordered assembly of a-synuclein. Cell Tissue Res. 2018;373:137–48.

    Article  CAS  PubMed  Google Scholar 

  3. Bloem B, Okun M, Klein C. Parkinson’s disease. Lancet. 2021;397:2284–303.

    Article  CAS  PubMed  Google Scholar 

  4. Shao Y, Le W. Recent advances and perspectives of metabolomics-based investigations in Parkinson’s disease. Mol Neurodegener. 2019;14:3.

    Article  PubMed  PubMed Central  Google Scholar 

  5. Abdullah R, Basak I, Patil K, Alves G, Larsen J, Moller S. Parkinson’s disease and age: the obvious but largely unexplored link. Exp Gerontol. 2015;63:33–8.

    Article  CAS  Google Scholar 

  6. Davies K, Bohic S, Carmona A, Ortega R, Cottam V, Hare D, Finberg J, Reyes D, Halliday G, Mercer J. Double Copper pathology in vulnerable brain regions in Parkinson’s disease. Neurobiol Aging. 2014;35:858–66.

    Article  CAS  PubMed  Google Scholar 

  7. Esteves A, Swerdlow R, Cardoso S. LRRK2, a puzzling protein: insights into Parkinson’s disease pathogenesis. Exp Neurol. 2014;261:206–16.

    Article  CAS  PubMed  Google Scholar 

  8. Santini E, Heiman M, Greengard P, Valjent E, Fisone G. Inhibition of mTOR signaling in Parkinson’s disease prevents L-DOPA-induced dyskinesia. Sci Signal. 2009;2:ra3.

    Article  Google Scholar 

  9. Nyström T, Yang J, Molin M. Peroxiredoxins, gerontogenes linking aging to genome instability and cancer. Genes Dev. 2012;26:2001–8.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  10. Shao Y, Le W. Recent advances and perspectives of metabolomics-based investigations in Parkinson’s disease. Mol Neurodegeneration. 2019;14:3.

    Article  Google Scholar 

  11. Kondoh H, Kameda M, Yanagida M. Whole blood metabolomics in aging research. Int J Mol Sci. 2021;22:175.

    Article  CAS  Google Scholar 

  12. Hallett P, Engelender S, Isacson O. Lipid and immune abnormalities causing age-dependent neurodegeneration and Parkinson’s disease. J Neuroinflammation. 2019;16:153.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  13. Sinclair E. Metabolomics of sebum reveals lipid dysregulation in Parkinson’s disease. Nat Comm. 2021;12:1592.

    Article  CAS  Google Scholar 

  14. Mattsson N. CSF biomarkers in neurodegenerative diseases. Clin Chem Lab Med. 2011;49:345–52.

    Article  CAS  PubMed  Google Scholar 

  15. Engelborghs S, Marescau B, De Deyn P. Amino acids and biogenic amines in cerebrospinal fluid of patients with Parkinson’s disease. Neurochem Res. 2003;28:1145–50.

    Article  CAS  PubMed  Google Scholar 

  16. Hong Z. DJ-1 and α-synuclein in human cerebrospinal fluid as biomarkers of Parkinson’s disease. Brain. 2010;133:713–26.

    Article  PubMed  PubMed Central  Google Scholar 

  17. Postuma R, Berg D, Stern M. MDS clinical diagnostic criteria for Parkinson’s disease. Mov Disord. 2015;30(12):1591–601.

    Article  PubMed  Google Scholar 

  18. Goetz C, Tilley B, Shaftman S. Movement Disorder Society-sponsored revision of the Unified Parkinson’s Disease Rating Scale (MDS-UPDRS): scale presentation and clinimetric testing results. Mov Disord. 2008;23(15):2129–70.

    Article  PubMed  Google Scholar 

  19. Galindo-Prieto B, Eriksson L, Trygg J. Variable influence on projection (VIP) for OPLS models and its applicability in multivariate time series analysis. Chemom Intell Lab Syst. 2015;146:297–304.

    Article  CAS  Google Scholar 

  20. Lee L, Liong C, Jemain A. Partial least squares-discriminant analysis (PLS-DA) for classification of high-dimensional (HD) data: a review of contemporary practice strategies and knowledge gaps. Analyst. 2018;143:3526–39.

    Article  CAS  PubMed  Google Scholar 

  21. Trygg J, Wold S. Orthogonal projections to latent structures (O-PLS). J Chemom. 2002;16:119–28.

    Article  CAS  Google Scholar 

  22. Triba M, Le Moyec L, Amathieu R, Goossens C, Bouchemal N, Nahon P, Rutledgee D, Savarina P. PLS/OPLS models in metabolomics: the impact of permutation of dataset rows on the K-fold cross-validation quality parameters. Mol BioSyst. 2015;11:13–9.

    Article  CAS  PubMed  Google Scholar 

  23. Huang L, Fuchins H, Kawahara N, Narukawa Y, Hada N, Kiuchi F. Application of a new method, orthogonal projection to latent structure (OPLS) combined with principal component analysis (PCA), to screening of prostaglandin E2 production inhibitory flavonoids in Scutellaria Root. J Nat Med. 2016;70:731–9.

    Article  CAS  PubMed  Google Scholar 

  24. Gao X, Starmer J, Martin E. A multiple testing correction method for genetic association studies using correlated single nucleotide polymorphisms. Genet Epidemiol. 2008;32:361–9.

    Article  PubMed  Google Scholar 

  25. Orlhac F, Soussan M, Maisonobe J-A, Garcia C, Vanderlinden B, Buvat I. Tumor texture analysis in 18f-fdg pet: relationships between texture parameters, histogram indices, standardized uptake values, metabolic volumes, and total lesion glycolysis. J Nucl Med. 2014;55:414–22.

    Article  CAS  PubMed  Google Scholar 

  26. Johnson R, Nelson G, Troyer J, Lautenberger J, Kessing B, Wrinkler C, OBrien S. Accounting for multiple comparisons in a genome-wide association study (gwas). BMC Genom. 2010;11:724.

    Article  Google Scholar 

  27. Kolossvary M, Karady J, Szilveszter B. Radiomic features are superior to conventional quantitative computed tomographic metrics to identify coronary plaques with napkin-ring sign. Circulation. 2017;12:10.

    Google Scholar 

  28. Perneger T. What’s wrong with Bonferroni adjustments. Biomed J. 1998;316:1236.

    CAS  Google Scholar 

  29. Bonferroni C. Teoria statistica delle classi e calcolo delle probabilita. ubblicazioni dell Istituto Superiore di Scienze Economiche e Commerciali di Firenze. 1936;8:3–62.

    Google Scholar 

  30. Li S. Predicting network activity from high throughput metabolomics. PLoS Comput Biol. 2013;9: e1003123.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  31. Moisan F. Parkinson disease male-to-female ratios increase with age: French nationwide study and meta-analysis. J Neurol Neurosurg Psychiatry. 2016;87:952–7.

    Article  PubMed  Google Scholar 

  32. Saiki S, Hatano T, Fujimaki M, Ishikawa K, Mori A, Oji Y, Okuzumi A, Fukuhara T, Koinuma T, Imamichi Y, Nagumo M, Furuya N, Nojiri S, Amo T, Yamashiro K, Hattori N. Decreased long-chain acylcarnitines from insufficient β-oxidation as potential early diagnostic markers for Parkinson’s disease. Sci Rep. 2017;7(1):7328.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  33. Szu-Ju C, Ching-Hua K, Han-Chun K, Chieh-Chang C, Wei-Kai W, Jyh-Ming L, Ming-Shiang W, Chin-Hsien L. The gut metabolite trimethylamine N-oxide is associated with Parkinson’s disease severity and progression. Mov Disord. 2020;35(11):2115.

    Article  CAS  Google Scholar 

  34. Mally J, Szalai G, Stone T. Changes in the concentration of amino acids in serum and cerebrospinal fluid of patients with Parkinson’s disease. J Neurol Sci. 1997;151(2):159–62.

    Article  CAS  PubMed  Google Scholar 

  35. Iwasaki Y, Ikeda K, Shiojima T, Kinoshita M. Increased plasma concentrations of aspartate, glutamate and glycine in Parkinson’s disease. Neurosci Lett. 1992;145(2):175–7.

    Article  CAS  PubMed  Google Scholar 

  36. Jiménez-Jiménez F, Molina J, Vargas C, Gómez P, Navarro J, Benito-Leon J. Neurotransmitter amino acids in cerebrospinal fluid of patients with Parkinson’s disease. J Neurol Sci. 1996;141(1):39–44.

    Article  PubMed  Google Scholar 

  37. Figura M, Kuśmierska K, Bucior E, Szlufik S, Koziorowski D, Jamrozik Z, Janik P. Serum amino acid profile in patients with Parkinson’s disease. PLoS ONE. 2018;13(1): e0191670.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  38. Makletsova M, Syatkin S, Poleshchuk V, Urazgildeeva G, Chigaleychik L, Sungrapova C, Illarioshkin S. Parkinson’s disease: their role in oxidative stress induction and protein aggregation. J Neurol Res. 2019;9:1.

    Article  Google Scholar 

  39. Plotegher N, Bubacco L, Greggio E, Civiero L. Ceramides in Parkinson’s disease: from recent evidence to new hypotheses. Front Neurosci. 2019;13:330.

    Article  PubMed  PubMed Central  Google Scholar 

  40. Li T, Tang W, Zhang L. Monte Carlo cross-validation analysis screens pathway cross-talk associated with Parkinson’s disease. Neurol Sci. 2016;37:1327–33.

    Article  PubMed  Google Scholar 

  41. Xicoy H, Wieringa B, Martens G. The role of lipids in Parkinson’s disease. Cells. 2019;8(1):27.

    Article  CAS  PubMed Central  Google Scholar 

  42. Fanning S, Selkoe D, Dettmer U. Parkinson’s disease: proteinopathy or lipidopathy? NPJ Parkinson’s Dis. 2020;3:1–9.

    Google Scholar 

  43. Fernández-Irigoyen J, Cartas-Cejudo P, Iruarrizaga-Lejarreta M, Santamaría E. Alteration in the cerebrospinal fluid lipidome in Parkinson’s disease: a post-mortem pilot study. Biomedicines. 2021;9(5):491.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  44. Graham SF, Rey NL, Yilmaz A, Kumar P, Madaj Z, Maddens M, Bahado-Singh RO, Becker K, Schulz E, Meyerdirk LK, Steiner J, Ma J, Brundin P. Biochemical profiling of the brain and blood metabolome in a mouse model of prodromal Parkinson’s disease reveals dis. J Proteome Res. 2018;17:2460–9.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  45. Carballo-Carbajal I, Laguna A, Romero-Giménez J. Brain tyrosinase overexpression implicates age-dependent neuromelanin production in Parkinson’s disease pathogenesis. Nature Commun. 2019;10:973.

    Article  CAS  Google Scholar 

  46. Michel V, Yuan Z, Ramsubir S, Bakovic M. Choline transport for phospholipid synthesis. Exp Biol Med. 2006;231(5):490–504.

    Article  CAS  Google Scholar 

  47. Hashizume O, Ohnishi S, Mito T, Shimizu A, Ishikawa K, Nakada K, Soda M, Mano H, Togayachi S, Miyoshi H, Okita K, Hayashi J. Epigenetic regulation of the nuclear-coded GCAT and SHMT2 genes confers human age-associated mitochondrial respiration defects. Sci Rep. 2015;5:10434.

    Article  PubMed  PubMed Central  Google Scholar 

  48. Canfield C-A, Bradshaw P. Amino acids in the regulation of aging and aging-related diseases. Transl Med Aging. 2019;3:70–89.

    Article  Google Scholar 

  49. Qu Q, Chen X, Sun S. Serine and Metabolism regulation: a novel mechanism in antitumor immunity and senescence. Aging. 2020;11:1640–53.

    Google Scholar 

  50. Paglia G. Unbiased metabolomic investigation of Alzheimer’s disease brain points to dysregulation of mitochondrial aspartate metabolism. J Proteome Res. 2016;15:608–18.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  51. Yi L, Shi S, Wazng Y, Huang W, Xia Z-A, Xing Z, Peng W, Wang Z. Serum metabolic profiling reveals altered metabolic pathways in patients with post-traumatic cognitive impairments. Sci Rep. 2016;6:21320.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  52. Zhao Y, Chen H, Iqbal J, Liu X, Zhang H, Xiao S, Jin N, Yao F, Shen L. Targeted metabolomics study of early pathological features in hippocamus of trile transgenic Alzheimer’s disease male mice. J Neurosci Res. 2021;99:927–46.

    Article  CAS  PubMed  Google Scholar 

  53. Carnevale D. A cholinergic-sympathetic pathway primes immunity in hypertension and mediates brain-to-spleen communication. Nat Commun. 2016;7:13035.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  54. Hunsberger H, Greenwood B, Tolstikov V, Narain N, Kiebish M, Denny CA. Divergence in the metabolome between natural aging and Alzheimer’s disease. Sci Rep. 2020;10:12171.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  55. Tyurina Y, Polimova A, Maciel E, Tyurin V, Kapralova V, Winnica D, Vikulina A, Domingues M, McCoy J, Sanders L. LC/MS analysis of cardiolipins in substantia nigra and plasma of rotenone-treated rats: implication for mitochondrial dysfunctio. Free Radic Res. 2015;49:681–91.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  56. Barros A, Crispim R, Cavalcanti J, Souza R, Lemos J, Cristino G, Bezerra M, Pinheiro T, de Vasconcelos S, Macedo D, Viana G, Aguiar G. Impact of the chronic omega-3 fatty acids supplementation in hemiparkinsonis. Basic Clin Pharmacol Toxicol. 2017;120:523–31.

    Article  CAS  PubMed  Google Scholar 

  57. Johnson A, Stolzing A. The role of lipid metabolism in aging, lifespan regulation, and age-related disease. Aging Cell. 2019;18(6): e13048.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  58. Lawton K, Berger A, Mitchell M, Milgram K, Evans A, Guo L. Analysis of the adult human plasma metabolome. Pharmacogenomics. 2008;9:383–97.

    Article  CAS  PubMed  Google Scholar 

  59. Suao Y, Li T, Liu Z, Wang X, Xu X, Li S, Xu G, Le W. Comprehensive metabolic profiling of Parkinson’s disease by liquid chromatography-mass spectrometry. Mol Neurodegener. 2021;16:4.

    Article  CAS  Google Scholar 

  60. Torok N, Tanaka M, Vecsei L. Searching for peripheral biomarkers in neurodegenerative diseases: the tryptophan-kynurenine metabolic pathway. Int J Mol Sci. 2020;21:9338.

    Article  PubMed Central  CAS  Google Scholar 

  61. Venkatesan D, Iver M, Naravanasamv A, Siva K, Vellingiri B. Kynurenine pathway in Parkinson’s disease—an update. eNeurological Sci. 2020;21:100270.

    Article  Google Scholar 

  62. Cervenka I, Agudelo L, Ruas J, Kynurenines J. Tryptophan’s metabolites in exercise, inflammation, and mental health. Science. 2017;357:eaaf9794.

    Article  PubMed  CAS  Google Scholar 

  63. Chang K, Cheng M, Tang H, Huang C, Wu Y, Chen C. Alternations of metabolic profile and kynurenine metabolism in the plasma of Parkinson’s disease. Mol Neurobiol. 2018;55:6319–28.

    Article  CAS  PubMed  Google Scholar 

  64. Xie K, Qin Q, Log Z, Yang Y, Peng C, Xi C, Li L, Wu Z, Daria V, Zhao Y, Wang F, Wang M. High-throughput metabolomics for discovering potential biomarkers and identifying metabolic mechanisms in aging and Alzheimer’s disease. Front Cell Dev Biol. 2021;9:335.

    Google Scholar 

  65. Sorgdrager-Freek JH, Naudé Petrus JW, Kema Ido P, Nollen EA, Deyn Peter PD. Tryptophan metabolism in inflammaging: from biomarker to therapeutic target. Front Immunol. 2019;10:2565.

    Article  CAS  Google Scholar 

Download references

Acknowledgements

We thank Claudio Caimi and the Hewlett Packard Enterprise (HPE) team for the support in the operation of the Ezmeral platform.

Funding

The authors acknowledge the support of the Gatekeeper 1st open call, Grant number 857223 and of the MAECI Great Relevance 2019 contributions Italy-China (Grant No. PGR00846).

Author information

Authors and Affiliations

  1. Huazhong University of Science and Technology, Wuhan, China

    Nicola D’Ascenzo, Emanuele Antonecchia & Qingguo Xie

  2. Department of Medical Physics and Engineering, Istituto Neurologico Mediterraneo NEUROMED I.R.C.C.S., Pozzilli, Italy

    Nicola D’Ascenzo, Emanuele Antonecchia, Victor Bender & Qingguo Xie

  3. Centre for Research and Training in Medicine of Aging, Department of Medicine and Health Sciences “V. Tiberio”, University of Molise, Campobasso, Italy

    Antonella Angiolillo & Alfonso Di Costanzo

  4. NIM Competence Center for Digital Healthcare GmbH, Berlin, Germany

    Marco Camerlenghi

  5. University of Science and Technology of China, Hefei, China

    Qingguo Xie

Authors
  1. Nicola D’Ascenzo
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Emanuele Antonecchia
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Antonella Angiolillo
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Victor Bender
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Marco Camerlenghi
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Qingguo Xie
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Alfonso Di Costanzo
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

ND and AD designed the research and wrote the manuscript; ND designed the statistical analysis concept and developed the statistical analysis software; ND, EA, VB and QX performed the data analysis; MC was responsible of IT issues and platform integration; AA and AD were responsible of the clinical interpretation of the results. All authors read and approved the final manuscript.

Corresponding authors

Correspondence to Nicola D’Ascenzo, Antonella Angiolillo or Qingguo Xie.

Ethics declarations

Ethics approval and consent to participate

The Institutional Review Board (IRB) of the University of Molise reviewed and approved the study (IRB Prot. n. 17/2020).

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Additional file 1: Table S1.

List of the analyzed metabolites

Rights and permissions

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 licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence 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 licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and Permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

D’Ascenzo, N., Antonecchia, E., Angiolillo, A. et al. Metabolomics of blood reveals age-dependent pathways in Parkinson’s Disease. Cell Biosci 12, 102 (2022). https://doi.org/10.1186/s13578-022-00831-5

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s13578-022-00831-5

Keywords

  • Parkinson’s Disease
  • Metabolomics
  • Age
  • Biomarkers

Cell & Bioscience

ISSN: 2045-3701

Contact us