Abstract
ABBREVIATIONS
Parkinson’s disease
liquid chromatography/mass spectrometry
principal component analysis
orthogonal projections to latent structures
4-fluoro-7-nitro-2,1,3-benzoxadiazole
catechol-O-methyltransferase
INTRODUCTION
Insidious development of motor symptoms in Parkinson’s disease (PD) indicates a long-term pathophysiological process in PD. Dopaminergic neurons in substantia nigra are decreased by 30 to 50% before clinical diagnosis of PD is made [1]. A biomarker representing an underlying pathological process at the pre-motor stage of the disease should be explored [2].
According to the six pathological stages of PD proposed by Braak and colleagues [3], incidental Lewy body pathologies (Lewy bodies and Lewy neurites) in the autonomic nervous system and the lower brainstem nuclei precedes Lewy bodies in substantia nigra, and represents the pre-motor neuropathological process. Clinical and pathological studies corroborated that PD is always accompanied with Lewy body pathologies in the peripheral autonomic nervous system [4], skin [5], colon [6], epicardium [7], the salivary gland [8], and the olfactory bulb [9]. Lewy bodies in skin biopsy are diagnostic of autonomic dysfunction in PD [10]. In addition, biopsies of the olfactory bulb [11] and the submandibular gland [8] are also proposed to diagnose PD. However, the invasiveness and possible complications make biopsies less likely to be widely accepted in the future. Development of less invasive biomarkers for PD is expected. Two biomarkers for presymptomatic PD pathology are alpha-synuclein oligomers [12] and DJ-1 [13] in the cerebrospinal fluid. In an effort to search for possible biomarkers in sweat and serum in PD, we analyzed amino acids and other hydrophilic molecules in PD using liquid chromatography/mass spectrometry (LC/MS). We found that the serum Tyr/Phe ratio is low in 49% of PD patients.
METHODS
Ethical statements
All studies were approved by the Ethical Review Committee of Nagoya University Graduate School of Medicine; Graduate School of Pharmaceutical Sciences, University of Tokyo; and Takamatsu Neurology Clinic. The studies were performed after an appropriate informed written consent was obtained.
Study subjects and evaluation methods
Diagnosis of PD was made based on the United Kingdom Parkinson’s Disease Society Brain Bank criteria. Controls were spouses of PD patients. Lack of neurological deficits in controls was examined by M.H. The severities of PD were evaluated using the Hoehn and Yahr scale and the Unified Parkinson’s Disease Rating Scale (UPDRS) parts I-IV. We evaluated the smell function using Odor Stick Identification Test for Japanese (OSIT-J), which has been validated and is widely used for evaluation of PD [14]. The OSIT-J included 12 odorants familiar to Japanese subjects: India ink, wood, perfume, menthol, Japanese orange, curry, cooking gas, rose, Japanese cypress (Hinoki), sweaty-smelling clothes, condensed milk, and roasted garlic. The maximum OSIT-J score is 12 points (normal range, 8–12 points).
Acquisitions of sweat samples and analysis using liquid chromatography/mass spectrometry (LC/MS)
Sweat samples were collected as previously described [15]. Two ml of 10% ethanol was poured into a 3.5-cm culture dish. The patient’s right palm was placed over the dish, and the dish was turned to dissolve the palm sweat in 10% ethanol. The dish was held on the patient’s palm for 30 sec, and was turned again to collect the solution in the dish. The amount of sweat secreted from the left index finger during the acquisition of palm sweat was estimated by a hand-assembled perspiration meter (Pico-Device, Nagoya, Japan) [15].
Analytical conditions of LC/MS
LC/MS studies were performed with the maXis HD Ultra-High Resolution QTOF system (Bruker Daltonics, Bilerica, MA, USA). The amount of sweat sample injected into LC was adjusted by the amount of sweat secreted from the left index finger stated above. Samples were ionized by electrospray. MS conditions were as follows: nebulizer at 2.0 bar; dry gas flow of 8.0 L/min; dry temperature at 180°C; positive ions; and a mass range of 50 to 800 m/z. The samples were separated using either of the following methods. First, we used L-column2 ODS column (CERI) with a particle diameter of 3.0 μm and column size of 2.1×150 mm in 5% acetonitrile aqueous solution containing 0.1% formic acid. The flow rate was 0.3 ml/min. Second we also used ACQUITY UPLC BEH Amide column (Waters, Milford, MA, USA) with a particle diameter of 1.7 μm and column size of 2.1×150 mm. The mobile phase was 15% acetonitrile aqueous solution containing 0.1% formic acid with a flow rate of 0.2 ml/min. The mass number was automatically adjusted by injecting standard compound directly to MS every time after injecting each sample to LS but before emergence of the first peak of the sample in MS. The fluctuation of retention time was less than 1.5% with our system. Each sweat sample was analyzed three to four times. A signal with a signal-to-noise ratio of more than 5 was analyzed. Identity of each signal was estimated using the Metlin mass spectral library (The Scripps Research Institute, La Jolla, CA, USA). Identities of Tyr and Phe were confirmed by analyzing standard compounds. The principal component analysis (PCA) and the orthogonal projections to latent structures (OPLS) analysis were performed with SIMCA 13 (Umetrics, Umea, Sweden).
Identification of phe and tyr in the serum
Concentrations of Phe and Tyr were measured using the previously reported methods [16, 17]. Briefly, the serum amino acids were derivatized with 4-fluoro-7-nitro-2,1,3-benzoxadiazole (NBD-F), and then separated with an ODS column (Inertsil ODS-4, 150 mm×3.0 mm I.D., GL Sciences). The mobile phase was composed of a mixture of 25 mM citrate buffer containing 25 mM sodium perchlorate (pH 6.2) and acetonitrile. NBD-amino acids were fluorescently detected with excitation and emission wavelengths of 470 and 530 nm, respectively. Limits of detection for NBD-Phe and NBD-Tyr were 5 and 10 fmol/injection, respectively. Limits of quantitation for NBD-Phe and NBD-Tyr were 17 and 35 fmol/injection, respectively. The relative standard deviations of NBD-Phe and NBD-Tyr within a day were below 5%.
Statistical analyses
Statistical significance was calculated by Student’s t-test, Fisher’s exact test, and one-way ANOVA followed by post-hoc Tukey test using SPSS Statistics (IBM, Armonk, New York, USA). P < 0.05 was used as a threshold for significant difference.
RESULTS
Demographic and clinical data
For screening of biomarkers in the sweat, we recruited 42 PD patients [16 males and 26 females, 65.0 ± 7.6 years (mean and SD)] and 16 controls [11 males and 5 females, 56.2 ± 18.2 years (mean and SD)]. All PD patients were examined by M.H., and were treated with L-Dopa and other anti-PD drugs.
For confirmation of biomarkers in the serum, we recruited 28 de novo PD patients who had not started any drug treatment [13 males and 15 females, 64.2 ± 10.9 years (mean and SD)]. We also recruited 52 treated PD patients [22 males and 30 females, 67.8 ± 9.3 years (mean and SD)] and 27 controls [15 males and 12 females, 67.9 ± 8.7 years (mean and SD)]. All PD patients were examined and followed up by M.H or M.Y. We thus analyzed a total of 80 PD patients [35 males and 45 females, 66.5 ± 10.0 years (mean and SD)] and 27 controls.
LC/MS analysis of sweat samples
To search for biomarkers unique to PD, we analyzed amino acids and other hydrophilic molecules in the sweat of 42 PD patients and 16 controls using LC/MS with the retention time (RT) ranging from 0.5 to 7.0 min. All PD patients were treated with L-Dopa and most were also treated with additional anti-PD drugs. As signals (molecular ions or their fragments) with RT < 1.5 min were not as discriminative as signals with RT≥1.5 min, we analyzed signals with RT≥1.5 min. We performed PCA and OPLS analyses using the individual signals and their intensities. Score plots of both PCA (Fig. 1A) and OPLS (Fig. 1C) were able to discriminate PD and controls. The loading plots showed that seven molecules had high effect sizes to differentiate PD and controls (Fig. 1B and D). Among them, four molecules were amino acids (Tyr, Phe, Leu or Ile, and Asp), whereas only chemical formulas could be estimated for the other three molecules. Taken together, both PCA and OPLS of LC/MS signals generated decent models to differentiate PD patients from controls.
HPLC-fluorescence-based analysis of serum Phe and Tyr levels
Among the amino acids detected in PCA and OPLS using the sweat, Phe and Tyr are precursors of dopamine. As PD is a systemic disease that is not restricted to the nigrostriatal pathway, we analyzed the serum levels of Phe and Tyr. As L-Dopa and other anti-PD drugs may change dopamine metabolisms, we recruited 28 de novo PD patients who had not started any drug treatment, in addition to 52 treated PD patients and 27 controls. The disease duration and the Hoehn-Yahr score were significantly higher in treated patients [9.3 ± 4.7 years and 2.8 ± 0.7 (mean and SD), respectively] than in de novo patients [2.6 ± 1.7 years and 1.8 ± 0.9 (mean and SD), respectively] (p < 0.0001 for both the disease duration and the Hoehn-Yahr score).
We quantified serum concentrations of Phe and Tyr using the HPLC-fluorescence detection system that we previously reported [16]. We found that the serum Phe levels in de novo patients were higher than those in controls (Fig. 2A). In contrast, the serum Phe levels in treated patients were not different from those in controls. The serum Tyr levels in de novo and treated patients were marginally and slightly lower than those in controls, respectively, but without statistical significance (Fig. 2B). We calculated the Tyr/Phe ratio and found that the Tyr/Phe levels were significantly lower in both de novo and treated patients (Fig. 2C). Receiver operating characteristic (ROC) curves show that Phe in de novo PD (Fig. 3A) and Tyr in treated PD (Fig. 3D) differentiate PD and controls to some extents. ROC curves also show that there is a threshold where sensitivity = 0.5 and specificity = 1.0 for both de novo and treated patients (Fig. 3E, F). The threshold of the Tyr/Phe ratio was 0.818 for both de novo and treated patients. When de novo and treated PD patients were combined, subjects with a Tyr/Phe ratio of less than 0.82 were all PD patients (specificity = 27/27 = 1.00), although only 49% of PD patients were below this threshold (sensitivity = 0.49) (Table 1). Similarly, the positive predictive value was 39/39 = 1.00, and the negative predictive value was 27/68 = 0.40 (Table 1).
As half of PD patients had a Tyr/Phe ratio in the normal range and the other half had a lower Tyr/Phe ratio, we next looked into clinical features that were possibly associated with the Tyr/Phe ratio in PD. We compared 12 demographic, clinical, and pharmacological features in two groups of patients by setting the threshold of the Tyr/Phe ratio to 0.82 (Table 2). Two parameters were significantly associated with low Tyr/Phe ratios: (i) male and (ii) lower doses of entacapone, which is an inhibitor of catechol-O-methyltransferase (COMT). The Tyr/Phe ratios in males and females, however, were 0.775 ± 0.172 (mean and SD, n = 35) and 0.860 ± 0.182 (mean and SD, n = 45), respectively, which were not statistically different. Similarly, the Tyr/Phe ratios in patients with and without entacapone were 0.813 ± 0.144 (mean and SD, n = 26) and 0.827 ± 0.199 (mean and SD, n = 54), respectively, which were not statistically different. Additionally, the Pearson’s correlation coefficient between the amount of entacapone and the Tyr/Phe ratio was as low as 0.304 (p = 0.13). Taken together, neither gender nor entacapone was likely to be associated with the serum Tyr/Phe ratios in PD patients. Although there was no statistical difference in the smell test scores in patients with the Tyr/Phe ratio below and above 0.82 (Table 2), the smell test scores tended to be lower in patients with the Tyr/Phe ratio less than 0.82. A plot of the smell test score against the Tyr/Phe ratio showed a weak positive correlation between them (Fig. 4).
DISCUSSION
We analyzed amino acids and hydrophilic compounds in sweat samples of PD patients and controls using LC/MS to screen for biomarkers that distinguish the patients from controls. PCA and OPLS discriminated sweat samples of patients and controls (Fig. 1). The effect size analysis indicated that Tyr, Phe, Leu (Ile), and Asp were informative markers. As signals with the retention time less than 1.5 min in LC/MS were discarded due to low discrimination of these signals, we might have missed essential compounds that differentiate PD patients and controls.
After the first round of screening using sweat samples, we found that collecting sweat samples imposed psychological and physical burdens on the patients, because the patients could not move their hands as instructed and they could not hold their hand/fingers still. As serum samples can be easily obtained in clinical practice, we used serum samples in the subsequent studies. As Phe and Tyr are precursors of dopamine, we measured serum concentrations of Phe and Tyr using a sensitive technique that we previously reported [16]. Our analysis revealed that Tyr/Phe ratios are lower in both de novo and treated patients than in controls (Fig. 2C). We then asked which patients have lower Tyr/Phe ratios, and analyzed 12 demographic, clinical, and pharmacological features (Table 2). However, no parameter was explicitly associated with the lower Tyr/Phe ratio. Serum metabolite profiles in Alzheimer’s disease using gas chromatography and mass spectrometry showed that the Tyr and Phe levels in Alzheimer’s disease were decreased to 0.86 and 0.72 of normal, respectively, yielding a 1.19-fold higher Tyr/Phe ratio than controls [18]. Thus, the low Tyr/Phe ratio is unlikely to be commonly observed in neurodegenerative disorders.
Phe is one of essential amino acids. Phe is enzymatically hydroxylated to yield Tyr by phenylalanine hydroxylase. Tyr is then further hydroxylated to produce Dopa by tyrosine hydroxylase. Dopa is a precursor of dopamine, norepinephrine, and epinephrine. Tyrosine hydroxylase is the rate-limiting enzyme in the biosynthesis of catecholamines. In phenylketonuria, a germline mutation in phenylalanine hydroxylase causes elevated serum Phe levels [19]. In de novo PD patients, serum Phe levels were higher than in controls, but not to the levels observed in hyperphenylalaninemia. The lower Tyr/Phe levels in PD may be due to minimal or insignificant decrease in the amount or the enzymatic activity of phenylalanine hydroxylase.
The lower serum Tyr/Phe ratios in PD can be accounted for by three possible mechanisms. First, Fumimura and colleagues analyzed 783 consecutive autopsy cases and found that all 18 PD cases with or without dementia, as well as 33 out of 38 cases with dementia with Lewy bodies (DLB), had Lewy bodies in the adrenal gland [20]. The adrenal medulla is the principal site of the conversion of Tyr to catecholamines. The release of catecholamines from the adrenal medulla occurs pulsatively and the basal catecholamine levels are variable [21]. Thus, measurement of serum catecholamine levels is unlikely to be able to represent catecholamine metabolisms in adrenal medulla. On the other hand, Phe and Tyr are precursors of catecholamines in the adrenal gland, and serve as reservoirs for producing catecholamines. The low serum Tyr/Phe ratios in PD patients may represent pathological changes of chromaffin cells in the adrenal medulla. Second, the body weight of a PD patient is stable up to several years before the diagnosis of PD, and then declines afterwards [22]. PD patients consume more energy even at the resting state than those without PD, likely due to muscle rigidity and dyskinesia [23]. The low Tyr/Phe ratios in PD may thus be associated with the increased energy expenditure, although precise alteration of metabolisms in PD remains to be elucidated. Third, Lewy bodies are frequently observed in the enteric nervous system (ENS) [24]. Furthermore, Lewy bodies in the ENS may commence 20 years before the onset of degenerative changes in the central nervous system and the associated motor symptoms in PD [25]. In PD, intestinal absorption of sucrose is reduced [26]. Although abnormal absorption of amino acids has not been reported in PD to our knowledge, the low Tyr/Phe ratios may be partly due to alterations in the absorption ofamino acids.
According to the National Institutes of Health Biomarkers Definitions Working Group in 1998, a biomarker is defined as “a characteristic that is objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes, or pharmacologic responses to a therapeutic intervention.” [27]. Validity of a biomarker is estimated by specificity, sensitivity, and positive/negative predictive values [28]. All these validation parameters should be as high as possible, although no explicit thresholds are proposed for these parameters to the best of our knowledge. When the threshold was set to 0.82, the serum Tyr/Phe ratio had sensitivity = 0.49, specificity = 1.00, positive predictive value = 1.00, and negative predictive value = 0.40 (Table 1). These parameters indicate that 100% of individuals with the serum Tyr/Phe ratio in the abnormal range were all PD (positive predictive value = 1.00), whereas only 40% of individuals with the serum Tyr/Phe ratio in the normal range were controls (negative predictive value = 0.40). Conversely, only 49% of PD patients have abnormal Tyr/Phe ratios (sensitivity = 0.49), whereas 100% of controls have normal Tyr/Phe ratios (specificity = 1.00). Accordingly, low
Tyr/Phe ratios are strongly supportive of PD, while 51% of PD patients will be underestimated with this threshold.
Pathological processes causing PD are likely to occur several years prior to the development of classic motor symptoms that are required for clinical diagnosis of PD. Recognizing these pre-motor symptoms could help identify individuals at a high risk for developing PD. However, the sensitivity and specificity of individual features including the smell test and constipation are not high enough to identify such individuals with pre-motor PD. As premotor symptoms generally develop in parallel [29], an integrated analysis of the multiple pre-motor features might be an effective strategy to identify individuals at a high risk for PD [30]. We observed a positive correlation, although weak, between the smell test scores and the Tyr/Phe ratios in PD. This implies that the Tyr/Phe ratio may be low even in pre-motor PD. Inclusion of the serum Tyr/Phe ratio in the integrated analysis of pre-motor features is likely to improve a model to predict pre-motor PD. In this communication, we analyzed a limited number of patients and controls, and we did not analyze other Parkinsonian syndromes or other neurodegenerative disorders. Further studies will elucidate the validity and applicability of the serum Tyr/Phe ratio in predicting pre-motor PD.
COMPETING INTERESTS
Nothing to declare.
Footnotes
ACKNOWLEDGMENTS
The authors thank Goro Funakoshi at Aichi Center for Industry and Science Technology; Tateru Asai and Mahoro Hisanaga at Pico-device; and Fumiko Ozawa, Yui Okada, Sae Goto, Satoru Hasegawa, and Saki Maeda at Nagoya University Graduate School of Medicine for their technical assistance. This study was supported by Grants-in-Aid from the Ministry of Education, Culture, Sports, Science, and Technology of Japan; the Ministry of Health, Labor, and Welfare of Japan; and the Aichi Science and Technology Foundation.