Cross-sectional analysis of plasma and CSF metabolomic markers in Huntington’s disease for participants of varying functional disability: a pilot study

Huntington’s Disease (HD) is a progressive, fatal neurodegenerative condition. While generally considered for its devastating neurological phenotype, disturbances in other organ systems and metabolic pathways outside the brain have attracted attention for possible relevance to HD pathology, potential as therapeutic targets, or use as biomarkers of progression. In addition, it is not established how metabolic changes in the HD brain correlate to progression across the full spectrum of early to late-stage disease. In this pilot study, we sought to explore the metabolic profile across manifest HD from early to advanced clinical staging through metabolomic analysis by mass spectrometry in plasma and cerebrospinal fluid (CSF). With disease progression, we observed nominally significant increases in plasma arginine, citrulline, and glycine, with decreases in total and d-serine, cholesterol esters, diacylglycerides, triacylglycerides, phosphatidylcholines, phosphatidylethanolamines, and sphingomyelins. In CSF, worsening disease was associated with nominally significant increases in NAD+, arginine, saturated long chain free fatty acids, diacylglycerides, triacylglycerides, and sphingomyelins. Notably, diacylglycerides and triacylglyceride species associated with clinical progression were different between plasma and CSF, suggesting different metabolic preferences for these compartments. Increasing NAD+ levels strongly correlating with disease progression was an unexpected finding. Our data suggest that defects in the urea cycle, glycine, and serine metabolism may be underrecognized in the progression HD pathology, and merit further study for possible therapeutic relevance.


Results
Demographic information for participants is listed in Table 1. Ages of participants ranged from 24 to 60 years, with a mean age of 47.5 years. Nine of the 12 participants (75%) were female gender. Time since clinical diagnosis ranged from 1 to 12 years. CAG repeat lengths, when available (75% of participants) ranged from 41 to 50. Baseline TFC ranged from 3 to 13; three participants were HD Stage I (scores 12,13,13), four were Stage II (7,8,8,9), four Stage III (4,5,6,6), and one Stage IV (3). Concomitant medications are listed in Table 1.
UHDRS data for individual participants are listed in Table 2. Correlations between UHDRS domains and metabolites in plasma and CSF are depicted visually in Figs Table 1). Citrulline, which is involved in the nitric oxide and urea cycles with arginine, also negatively correlated with the Independence Scale score (p = 0.0400), Stroop Color (p = 0.0368), Stroop Interference (p = 0.0290), and Verbal Fluency (p = 0.0568). Plasma glycine negatively correlated with TFC (p = 0.034), Independence Scale (IND) (p = 0.038), and nearly all cognitive measures while positively correlating to TMS (p = 0.039), suggesting that rising glycine levels correlate with broad clinical progression ( Fig. 1; Supplemental Table 1). Total serine levels positively correlated with Stroop Color (p = 0.021) and Stroop Interference (p = 0.037), while d-serine (DSR) levels positively correlated with Stroop Word (p = 0.047) and trended towards positive correlation for the Independence Scale (p = 0.088) and SDMT (p = 0.081). Concentrations of DSR in plasma varied substantially (range 1.45 µM to 3.03 µM) with an average of 1.94 µM in plasma (Supplemental Table 3). Hexose demonstrated a positive trend with TFC (p = 0.073), IND (p = 0.0797), and a negative trend with TMS (p = 0.0676) suggesting lower levels with disease progression, though a contrasting positive correlation with Behavior scores (p = 0.037) was observed.
Circulating kynurenine levels in plasma showed positive correlation with Behavior scores (p = 0.0009), as did tryptophan (p = 0.001). While not nominally significant, increasing levels of kynurenine also negatively correlated with TMS (p = 0.0868). All other metabolites in the kynurenine metabolome showed no correlations with clinical outcomes, and the plasma kynurenine/tryptophan ratio was not associated with any measure of disease severity except a trending positive correlation with Behavior Scores (p = 0.06).
Circulating levels of several lipid classes decreased with clinical progression. These were cholesterol esters (CE), which positively correlated with Stroop Word (p = 0.0082) and Verbal (p = 0.0478); diacylglycerols (DAG), which negatively correlated with TMS (p = 0.0231) and positively with Independence (p = 0.0307) and Behavior (p = 0.0116); triacylglycerols (TAG), which negatively correlated with TMS (p = 0.0313) and positively with IND (p = 0.0357) and Behavior (p = 0.009); phosphatidylcholines (PC), which positively correlated with TFC  Table 1). No individual or total ceramide was found to correlate with any of the clinical outcomes. Among the CEs, only two species (CE 16:0 and CE 20:1) had a significant positive correlation with all cognitive components of the UHDRS (Fig. 1 www.nature.com/scientificreports/ Glycine levels across all 12 participants were near or below the limit of quantitation; however, those with more severe TFC scores had relatively higher circulating levels (Supplementary Table 4). No CSF kynurenine metabolites had a significant correlation with clinical outcome measures. Circulating kynurenic acid levels trended negatively with Stroop Word and SDMT outcomes (p = 0.0994 and 0.0774, respectively). Similar trends were observed for the kynurenic acid/kynurenine ratio with SDMT (p = 0.0981). The kynurenine/tryptophan ratio in the CSF was not associated with change in any UHDRS domain.

Discussion
To our knowledge, this pilot study is the first report in HD correlating metabolomic changes in plasma and CSF with clinical progression across both early (HD1 and 2) and more advanced stages (HD3 and 4) of the disease. Metabolomic analyses of both plasma and CSF were performed across a spectrum of disease severities to generate a broad exploratory profile for dynamic biological changes in HD. www.nature.com/scientificreports/ Here we report circulating levels of NAD + in the CSF of HD participants for the first time. NAD + is an important cofactor for several biochemical pathways, including glycolysis and three major classes of enzymes: SIRTuins, poly(ADP-ribose) polymerases (PARPs), and CD38/157 ecto enzymes 33,34 . NAD + cellular levels regulate several pathways involved in mHtt toxicity 35 , including SIRT1 modulation of PGC-1α, SIRT-3 and PARP-1. SIRTuins and PARPs have been recently targeted for HD therapy 36,37 . Recent reports suggest neuroprotective effects for NAD + in HD models, and suggest administration of nicotinamide riboside (NR), an NAD + precursor, as a potential therapy 36 . Contrary to recent studies suggesting decreased levels of NAD + would be expected with progressive neurodegeneration 38 , higher circulating CSF levels of NAD + correlated to worsening clinical status in our study. The reason for this is not clear. Considerations may include decreased levels or function of CD38, a highly expressed glycoprotein in neurons and astrocytes that generates cyclic ADP-Ribose (cADPR) from NAD + molecules 39 , or a proportional decrease in the intracellular concentration of NAD + . The latter case would be consistent with previous reports that NAD + is preferentially released from intracellular stores in conditions of cell stress or inflammation 40 .Decreased intracellular levels of NAD + are consistent with glutamate-induced excitotoxicity 41 , a process implicated in HD pathophysiology 42,43 . In the current study, levels of CSF glutamate were not determined, and increasing levels of plasma glutamate correlated only with declining behavioral scores. However, multiple pathways that indirectly modulate glutamatergic neurotransmission have also been proposed for involvement in neurodegenerative diseases, including signaling through the N-Methyl-D-aspartate (NMDA) receptor 44,45 . In this context, d-serine, a co-agonist for the NMDA receptor at the glycine site, has been investigated in neurodegeneration 43 . d-Serine plays a significant regulatory role in glutamate signaling, dendritic development, synaptic plasticity, long-term potentiation (LTP) and depression (LTD), and neuronal migration 44 . In this study, total plasma serine concentrations (l-serine and d-serine) positively correlated with two cognitive outcomes, while d-serine plasma concentrations correlated to one (Fig. 1), suggesting that both decreasing total serine and d-serine levels may be associated with clinical progression (Fig. 3). The total serine finding is consistent with previous studies showing decreased circulating plasma serine levels in HD relative to controls 45 . l-Serine is synthesized endogenously primarily from glucose via the glycolytic pathway (Fig. 3), while d-serine is synthesized from l-serine via serine racemase 44 . In our study, fasting glucose levels decreased with clinical progression, consistent with known impairment of glycolysis in HD and potentially influencing the low levels of total serine observed 46 . In CSF, no correlation was found between circulating levels of total serine or d-serine and clinical progression. We observed substantial inter-participant variation in d-serine CSF levels, which ranged from 0.61 to2.89 µM with a mean of 1.34 µM (Supplementary Table 4), consistent with previous reported values in healthy controls 47 .
Circulating plasma levels of glycine increased with clinical progression, consistent with previous studies 21,48,49 . No correlations were found in the CSF, largely due to levels at or below the limit of quantitation. However, circulating CSF glycine concentrations were higher for participants in later stages, similar to the clinical correlations in plasma. Elevated glycine with worsening clinical disease may be consistent with progressive excitotoxicity, a www.nature.com/scientificreports/ mechanism with numerous lines of evidence in HD 18 . The dichotomy between increasing glycine and decreasing total and d-serine levels in our study suggests either decreased metabolism of glycine or an increase in another glycine synthetic pathway. Glycine is mainly derived through de novo synthesis from the glycolytic pathway via serine, but may also derive from choline, threonine and glyoxylate (Fig. 3). Of these, only threonine was measured, which showed no association with progression. Choline, which is derived from PC in cell membranes, may increase via membrane breakdown from neurodegeneration causing release of membrane phospholipids 24,50 . In our study, an association between decreasing circulating plasma levels of PCs and clinical progression was observed. Future studies should determine circulating levels of choline and glyoxylate in HD subjects to further clarify these relationships. Upregulated plasma glycine may also indicate compensatory production of creatine and heme (Fig. 3), molecules derived from glycine that are involved in muscle/neuronal metabolism and proteins of the electron transport system, respectively 25 . Plasma creatinine showed no correlation to progression, whereas CSF creatinine significantly increased with worsening disease. It is not clear if this represents a preferential shift in CNS creatinine production and/or clearance 51 as a response to failing energetic mechanisms. Of note, creatinine supplementation in the long-term CREST-E clincial trial did not influence functional progression 52 . Nitric Oxide (NO) involvement in excitotoxicity has been well documented [53][54][55] . Increased formation of NO via nitric oxide synthase (NOS) depends upon an adequate supply of arginine (Fig. 3). Previous studies of increased activity for arginine synthase and arginine lyase suggests a role for the citrulline-NO cycle enzymes in excitotoxicity 55 . l-arginine is converted to L-citrulline and NO via NOS and the cofactor nicotinamide adenine dinucleotide phosphate (NADPH). Neuronal NOS mRNA is known to be reduced in HD striatum, particularly the dorsal caudate, and proportional to progression of disease 56 ; this intraneuronal reduction, if present, may account for the accumulation of arginine with disease progression. The present data suggest that in both plasma and CSF higher circulating concentrations of arginine correspond to disease progression. These findings are consistent with previous reports 22,57 . Notably, the administration of l-arginine to HD transgenic mice accelerated motor phenotype and weight loss, while diets absent in l-arginine slowed weight loss 58 . Plasma citrulline also negatively correlated with cognitive outcomes in our data. Higher circulating concentrations of citrulline have been reported in HD patients relative to controls 23 . This may reflect lesser activity of arginosuccinate synthase or lyase, a consequence demonstrated in murine models of mutant huntingtin suppressing C/EBP alpha, a critical regulator of urea cycle enzymes resulting in elevated citrulline and hyperammonemia 23 . Protein restriction in these animals normalized urea cycle function, lessened mHtt aggregation, and improved phenotype. Skene and colleagues have recently reported significantly elevated arginine and citrulline in transgenic HD sheep 59 . Taken together, additional investigation of the urea cycle in HD is warranted.
While no relation was observed in plasma between fatty acids and clinical outcomes, increasing circulating CSF levels of several saturated long chain fatty acids (LCFAs) correlated with disease progression, including stearic acid, margaric acid, palmitic acid, myristic acid and lauric acid. De novo fatty acid synthesis is closely associated with inflammatory cell subsets 60 and hypothesized to drive inflammation and disease activity for neurological conditions including Huntington's Disease 60 . Of note, fatty acid dysregulation in HD has been previously identified, with suggestion that brain fatty acids might be regulated differently and independently of circulating fatty acid concentrations 24 . Other reports indicate no difference in levels from controls or premanifest individuals compared to symptomatic persons 61 . Our results are consistent with recent data showing striatal astrocytes in the HdhQ (150/150) mouse model switch from glycolysis to fatty acid oxidation in the setting of decreasing glucose levels, a process involving increasing fatty acid concentrations that temporarily stabilize energetics but at the eventual expense of producing oxidative damage 62 . It may be that our divergent fatty acid results between plasma and CSF reflect an emerging preference in the brain, with progression, for high-energy substrates at the site of greatest vulnerability. Furthermore, our data show circulating plasma levels are not representative of circulating CSF levels, and therefore do not appear useful as prognostic lipid biomarkers.
Fatty acids are precursors for DAGs, TAGs, and phospholipid synthesis regulated by sterol regulatory element response protein 2 (SREBP-2), an enzyme known to be downregulated in murine HD models and manifest human postmortem brain 24 . Conversely, lipoprotein lipase (LPL)-mediated hydrolysis of triacylglycerol is an important source of diacylglycerols, free fatty acids and phospholipids in the brain 25 and is thought to influence responses to oxidative stress in several neurodegenerative disorders 63 . Therefore, the concomitant increase of circulating saturated LCFAs in CSF with corresponding DAGs, TAGs, phospholipids and sphingolipids is perhaps not surprising. In our study, increasing circulating CSF levels of DAG Whether DAG accumulation reflects a reaction to energetic abnormalities with disease progression (perhaps derived from TAG) or more specific second-messenger signaling activity is unclear; DAG is a highly active second messenger across numerous signaling cascades (the IP3-DAG system) 30 . Notably, none of the increasing DAGs or TAGs with clinical correlations in the CSF were found to have significant correlations in plasma; moreover, in contrast to the CSF, decreasing circulating plasma levels of the eight DAGs and 92 TAGs correlated with clinical progression. Decreasing levels of TAGs in plasma are consistent with decreasing circulating plasma glucose levels, as TAG synthesis occurs via the glycerol-3-phosphate pathway in endoplasmic reticulum and mitochondria 64 . Taken together, our data suggest substantially different metabolism for TAG and DAG species between peripheral tissues and brain in HD, an observation for which therapeutic implications are not yet clear.
In www.nature.com/scientificreports/ CSF, with PC (16:0/18:1) being the most abundant 65 . Of all PCs and DAGs assayed, PC (16:0/16:0) and DAG (16:0/16:0) associated most strongly with progression, correlating with 9/10 and 8/10 measured outcomes, respectively. The concomitant increase of circulating CSF levels of palmitic acid, DAG(16:0/16:0) and PC (16:0/16:0) would suggest an increase in SM (16:0) based on the sphingomyelin cycle, in which sphingomyelin synthase transfers a phosphorylcholine group from a PC to ceramide to generate DAG and SM 66 . Indeed, we observed SM (16:0) increasing with clinical progression, whereas increasing sphingomyelins 14:0, 18:0, and 18:1 also appeared to have a strong relationship with clinical decline. Increasing levels may simply reflect progressive neurodegeneration and membrane breakdown. However, a defective de novo biosynthetic pathway for sphingolipids has been demonstrated in multiple HD pre-clinical models. For example, decreases in dihydrosphingosine, dihydrosphingosine-1-phosphate and dihydroceramide 18:0 were observed in R6/2 mice 67 . Of note, no concomitant change in circulating ceramide levels was observed with progression in our data. This may reflect other peripheral sources of ceramide, as it is known to readily cross the blood-brain barrier, or possibly a preferential compensatory preservation of CSF levels 68 . Ceramide is degraded to sphingosine and phosphorylated to form sphingosine-1-phosphate (S1P), a molecule which may have relevance in HD 67,69 . There is suggestion that sphingosine-1-phosphate (S1P) species may be especially important in HD compensatory mechanisms, supporting glial survival, neurite outgrowth, blood-brain barrier integrity, and neurogenesis 67 . Several limitations to these pilot data should be acknowledged. Our findings must be interpreted with caution in light of the small sample size (n = 12) and unadjusted statistical analysis for multiple comparisons. Highlighted results obtained are trends or carry nominal significance, which are limited by the small sample and potential variability in studying both early and advanced stages in HD. An additional limitation is the absence of an age-matched healthy control group. These factors limit our ability to more specifically assess associations with disease progression. CSF was not quality checked for blood contamination, making contamination with blood metabolites possible, though on inspection of fluid and analysis of metabolite concentrations this risk was felt to be minimal. Behavioral and anti-chorea medications were allowed in the study, as washout of these treatments and the related risks for instability were not felt to be justified. As such, it is possible that clinical results were influenced by this decision. Of note, chorea is only one component of the Total Motor Score, which also assesses voluntary motor capability, gait, and other involuntary elements like bradykinesia, rigidity and dystonia. It is important to recognize CSF metabolite levels reflect the extracellular environment of the CNS and may not entirely or accurately reflect intracellular changes.

Conclusion
Overall, these data are advantageous in providing a simultaneous look at both plasma and CSF metabolites, and allow for a novel cross-sectional analysis across a broad spectrum of the disease. A number of changes consistent with previously reported metabolites were observed in our data and help generate hypotheses for future work. In particular, urea cycle defects appear to deserve further study for potential therapeutic implications, as do d-serine, glycine, and NAD + . Larger longitudinal studies are planned to clarify the biological and potentially therapeutic relevance of these findings.

Methods
Ethical approval for conduct of this study was obtained from the Institutional Review Board of Cooper University Hospital at Rowan University on July 18, 2017. All participants were over the age of 18, and all provided written informed consent to participate in the study. This pilot study was done in accordance with the Declaration of Helsinki and International Conference on Harmonization Good Clinical Practice guidelines. Twelve consenting participants with either genetically confirmed HD or an unequivocal phenotype (Diagnostic Confidence Level IV, UHDRS) and a genetically confirmed first-degree relative were enrolled from September 30, 2017 to July 18, 2018. Participants were stratified according to Total Functional Capacity Score 70 . This is a 13-point scale that assesses functional performance in 5 domains: capacity for work, finances, domestic chores, self-care, and care level required. Scores range from 13 (normal) to 0 (total incapacitation) and define stages of the disease: HD1 (TFC 11-13, early mild), HD2 (TFC 7-10, mild), HD3 (TFC 4-6, moderate, and HD4 (0-3, severe). All study visits were conducted in the morning at Cooper University Hospital (Camden, NJ). Participants were advised to arrive fasted the morning of study. Study visits consisted of vital signs, review of concomitant medications and medical conditions, performance of the Unified Huntington's Disease Rating Scale (UHDRS) by a certified examiner (AM), blood draw, and fluoroscopically guided lumbar puncture. The Unified Huntington's Disease Rating Scale is comprised of five domains for assessment: Motor, Cognitive, Behavioral, Independence Scale, and Functional Capacity 16 . Increasing scores in the Motor domain reflect worsening disease features, while decreasing scores in Cognitive, Behavioral, Independence, and Functional measurements reflect worsening disease. 20 cc of blood was withdrawn into sodium heparin tubes and centrifuged at 3400 RPM for 15 min, after which plasma was aliquoted and stored at − 80 °C. Lumbar punctures were performed by study personnel (AM, HS) and withdrew approx. 4 cc of clear cerebrospinal fluid per participant for storage at − 80 °C. QC for blood contamination of CSF was not performed.
Metabolites shown in "Results" section were selected for (1) having a minimum of 10 samples, (2) having nominally significant changes across motor, functional, and cognitive domains, or (3) for special interest as being nominally significant for motor change only or multiple cognitive domains only. Metabolite changes of nominal significance across multiple facets of the disease phenotype were felt to most likely reflect underlying biology. Behavior significance was not required for meeting this threshold, given the allowance of behavioral treatment during the study (participants on anxiolytics or antidepressants). Tetrabenazine treatment for chorea was also allowed in the study, as chorea is only one component of Total Motor Score assessment. www.nature.com/scientificreports/ Metabolomics panels. Metabolites were extracted from plasma and cerebrospinal fluid (CSF) and concentrations obtained using the AbsoluteIDQ kit p180 (Biocrates Life Science AG, Austria) following the manufacturer's protocol for the API5500 LC/MS/MS System (ABSciex, USA), running with Analyst 1.5.2 software equipped with an electrospray ionization source, a Shimadzu CBM-20A command module, LC-20AD pump, and a Shimadzu SIL-20AC-HT autosampler and a CTO-10Ac column oven heater 71 . The Lipidyzer (Sciex, Framingham, MA) platform was used to measure lipids in the phosphatidylcholine (PC), phosphatidylethanolamine (PE), lysophosphatidylcholine (LPC), lysophosphatidylethanolamine (LPE), free fatty acid (FFA), sphingomyelin (SM), diacylglycerol (DAG), cholesteryl ester (CE), ceramide (CER), hexosylceramide (HCER), lactosylceramide (LCER), dihydroceramide (DCER), and triacylglycerol (TAG) groups of lipids on API5500 LC/MS/MS System (ABSciex, USA), following a previously describe method 72 . Lipid species were included in the data analyses if above the limit of quantification in > 90% of the participants was satisfied.
Kynurenine pathway. Separation of the kynurenines was accomplished following a previously published protocol 73 . Briefly, to 40 µl plasma or CSF, 10 µl internal standard and 10 µl 0.1% formic acid in water was added. Solid-phase extraction cartridges (Oasis HLB, Waters Corp) were conditioned with 1 ml methanol, then 1 ml water. The samples were added and washed with 100 µl water. Finally, the metabolites were eluted with 1 ml 0.1% formic acid in 95:5 methanol: water and stream-dried under nitrogen. The samples were reconstituted in 100 µl 0.1% formic acid in 10:90 methanol: water and transferred to auto sampler vials for analysis. Data was acquired using a Nexera XR HPLC (Shimadzu) coupled with a QTRAP 6500 + (SCIEX) and was analyzed with Analyst 1.6 (SCIEX). A linear gradient was run for 30 min at a flow rate of 0.3 ml/min: 0-1 min 5% B, 3 min 23% B, 3.1-5 min 70% B, 5.5-20 min 90% B, 20.1 min 10% B, 21 min 5% B at 40 °C, on an X-Select HSS C18 column (2.1 × 150 mm, 2.5 µm, Waters), with mobile phase A consisting of 0.2% aqueous formic acid and mobile phase B consisting of 0.2% formic acid in methanol. Relative concentrations (abundance) of the metabolites were determined in standard solution and not matrix (matrix effects were not considered) using area ratios calculated using their corresponding deuterated standard with the exception of 3-hydroxykynurenine, and anthranilic acid, where D 4 -Kyn was used as their internal standard. d-Serine. d-Serine levels were measured following a previously developed protocol with slight modifications 74 .
Briefly, plasma or CSF samples (100 μl) were combined with 20 μl aliquot of IS (10 nmol/ml in acetone) and 400 μl acetone and then centrifuged at 13,000×g for 10 min at 4 °C. A 400 μl aliquot of the supernatant was subsequently derivatized with 300 μl (R)-1-Boc-2-piperidinecarbonyl chloride. After derivatization by stirring at 1000 rpm for 2 h at room temperature, 200 μl of triflouroacetic acid was added to each sample and then incubated for 1 h. The terminal product was evaporated to dryness under a stream of nitrogen. The residue was dissolved in 100 μl of methanol/water, 10:90, v/v and transferred to the autosampler for analysis. NAD + metabolome. Separation of the NAD + metabolites was accomplished using a previously described method 32 . Briefly, 20 μl of CSF was solubilized in 60 μl of methanol, including the addition of 5 μl of internal standard. Samples were then centrifuged at 4 °C for 10 min at 13,200×g to remove the protein pellet. The supernatant was collected and placed in an autosampler vial for analysis. Three quality controls with spiked standards was used for determination of the relative concentrations of NAD + and its metabolites. The NAD + metabolites were resolved using an Accucore HILIC column (2.1 × 150 mm, 2.6 μ, Thermo) at 32 °C was used with ammonium acetate [7.5 mM, pH 7.86] as mobile phase A and acetonitrile as mobile phase B with a 10 µl injection volume. The following linear gradient was run for 14.0 min at a flow rate of 0.4 ml/min: 0-1 min 90% B, 1.5 min 72.5% B, 2.5 min 67.5% B, 8.0 min 20% B, 10 min 20% B, 10.1 min 90% B. Relative values for the metabolites were determined using area ratios of the targeted metabolites and the corresponding internal standard using the following heavy standards: C 13 -NAD, C 13 -NADH, C 13 -ADPR, C 13 -AcCoA. Calibration curves were carried out in standard solutions. Matrix effects were accounted for in each targeted matrix by adjusting calculated levels based on three quality controls (low, middle, high).

Statistical analysis.
Pearson correlation coefficients were calculated between the components of the UHDRS as continuous variables and the measured metabolites. P-values obtained were not adjusted for multiple comparisons, and so all values are considered nominally significant. Correlations were categorized as moderate (absolute r-value = 0.4-0.69) or strong (absolute r value = 0.7 to 1.0), and coded accordingly in heat maps (moderate as lighter hue, strong as darker hue). Correlations could be positive (metabolite concentrations changing in same direction as clinical scale) or negative (concentrations changing in opposite direction as clinical scale). Among domains of the UHDRS, TMS and Behavior scores tend to increase with disease progression, while TFC, IS, FA, and cognitive assessments tend to decrease with progression; therefore, with worsening disease, correlations for TMS and Behavior were expected to be of opposite polarity to other assessments. Considerations of normality and confounding variables were not considered, as this was a pilot study with only 12 subjects.