Causal Connection Between Serum Levodopa Metabolic Prole and Medication in Parkinson’s Disease

No methods to assess ecacy of levodopa-associated therapy by blood sampling in Parkinson’s disease (PD) have been established. In this study, we investigated levodopa associated metabolites to characterize their associations with medication and clinical symptoms in PD patients. Comprehensive metabolome analysis using plasma from PD and controls was performed in two independent cohorts (PD: 109, controls: 32; PD: 145, controls: 45). In another validation cohort [251 PD patients (16 de novo, 17 receiving only dopamine-receptor agonists, 218 receiving levodopa/benserazide or levodopa/carbidopa with/without other parkinsonian drugs) and 40 age-matched controls], serum levels of levodopa and its six metabolites were examined by liquid chromatography-mass spectrometry. The association of each metabolite with clinical parameters, medication, and enzymic genotypes was investigated. Signicant increases in 3-methoxytyrosine and homovanillic acid were observed in PD patients administered levodopa/benserazide or levodopa/carbidopa. Serum levels of levodopa and ve of its metabolites were signicantly increased in PD patients administered levodopa and were related to the levodopa or entacapone dose but not to disease severity. Levodopa levels were more effectively preserved in PD patients given levodopa/benserazide than in those given levodopa/carbidopa, especially when taken with entacapone. Each dopamine or 3-methoxytyramine level was eciently expressed with a numerical model using levodopa, entacapone, and selegiline doses as variables, indicative of its application for drug ecacy monitoring. Benserazide (25 mg) blocked AADC and preserved levodopa levels more effectively than carbidopa (10 mg), and entacapone provided a concomitant effect on levodopa level preservation. The drug ecacy of levodopa-associated medication could be monitored by dopamine or 3-methoxytyramine levels. metabolite metabolic PD seven consecutive levodopa-associated metabolites in serum simultaneously. HCs PD patients. PD novo PD, patients receiving DAs, PD patients receiving levodopa/benserazide (L/B) with/without entacapone, amantadine, hereafter described as “PD with L/B or L/C”]. analyzed the level of each levodopa-associated metabolite in each group and compared the results with the HC group using Steel’s test No signicant differences were detected in mean ages at blood sampling or in the sex ratio between the HC and de novo PD, PD with only DA or PD with L/B or L/C groups. Disease duration was signicantly longer in the PD with L/B or L/C group than in other PD groups. On average, disease severity in the de novo PD and PD with only DA groups was mild-to-moderate according to the H&Y stage and UPDRS-III score. L/C groups reect the mean area under the serum concentration curve. a recent report, 19 a larger AUC for L/B treatment for L/C treatment, consistent with our results. of benserazide or carbidopa only increases the amount of levodopa reaching the brain to an estimated 10% of an administered dose because blocking AADC shunts levodopa into the COMT metabolic pathway, thereby increasing peripheral formation of 3-methoxytyrosine. 10 Among all cohorts of this study, 3-methoxytyrosine levels in PD patients were elevated more than 100 times those in HCs. Although no differences in ecacy between L/B and L/C were detected in PD patients treated with levodopa without COMT-Is, 27–29 our data suggested that concomitant use of benserazide with entacapone preserved levodopa concentrations more than carbidopa. These results are supported by correlated levels of dopamine or its downstream metabolites with entacapone doses in the L/C but not the L/B group, showing a high level of AADC leakage associated with COMT inhibition. According to previous therapeutic studies, 30 switching from L/B to L/C/entacapone produces similar ecacy in PD as switching from L/C to L/C/entacapone; however, no clinical trials comparing L/B/entacapone and L/C/entacapone in PD have been reported. Selegiline is primarily metabolized by the liver P450 system (CYP1A2) with some extrahepatic metabolism occurring in platelets. 31 Although platelet MAOB activity, which is inhibited by more than 85% within 4 hours of selegiline administration (5 mg), can be monitored in vitro, 32 measurement of MAOB-I ecacy using serum/plasma has not been reported. In our study, dopamine and 3-methoxytyramine levels in PD patients receiving levodopa were signicantly enhanced by selegiline, indicating potential responders to MAOB-Is. Further large-scale studies should address possible associations of MAOB SNVs, such as rs1799836, with MAOB Association of metabolism with for in


Introduction
Parkinson's disease (PD) is the second most common neurodegenerative disease. It is characterized by motor symptoms of akinesia, tremor and rigidity, which respond to levodopa, a precursor of dopamine. 1,2 Remarkable effects of oral levodopa administration were rst described for PD by George Cotzias in 1967, and levodopa with peripheral aromatic amino acid decarboxylase inhibitors (AADC-Is) was subsequently established as a standard therapy. [3][4][5] Although various routes of levodopa administration are available, oral administration of levodopa is the central pharmacological therapy against PD. 6,7 Levodopa-associated metabolites including 3,4-dihydroxyphenylacetaldehyde are toxic to nigral dopaminergic neurons and glia. 8 Levodopa is actively absorbed in the duodenum and proximal jejunum and is then metabolized to dopamine by AADC in intestinal epithelial cells or to 3-methoxytyrosine by catechol-O-methyltransferase (COMT) in the liver, muscles, kidneys, and red blood cells ( Supplementary Fig. 1). 9 Because 10% of levodopa actually reaches the brain, 10 the residual levodopa around 90% are degraded by both AADC and COMT during the systemic circulation. Thus, precise monitoring of the systemic levodopa metabolism enables us to perceive its internal alteration, leading to an appropriate adjustment of antiparkinsonian drugs. However, levodopa pharmacokinetics in PD have been investigated since 1975, 11,12 comprehensive metabolic pro les of levodopa have not been investigated in de novo PD or in pharmacologically-treated PD patients. Additionally, associations of common AADC single nucleotide variants with levodopa metabolism have not been reported. Finally, integrated effects of COMT inhibitors and/or monoamine oxidase inhibitors (MAOB-Is) along with levodopa/AADC-Is on levodopa metabolism in PD have never been investigated. Here, we comprehensively analyzed levodopa metabolism using liquid chromatography-mass spectrometry [13][14][15][16][17] and associated metabolite levels with medications and clinical features.

Participants
Comprehensive metabolome data obtained from previously reported double independent cohorts were reanalyzed. 14 Levels of two metabolites downstream of levodopa [3-methoxytyrosine and HVA] were signi cantly higher in PD patients treated with levodopa/AADC-Is (p < 0.0001) than in HCs (Supplementary Table 1). The levels of both metabolites are in uenced by levodopa/AADC-I and COMT-I, 10, 18 and we were unable to estimate medication e cacy using non-consecutive metabolite levels in the same metabolic pathway. To determine the precise e cacy of PD medications, we used an LC-MS/MS system to measure seven consecutive levodopa-associated metabolites in serum simultaneously. We recruited 40 HCs and 251 PD patients. Based on their medication characteristics (Supplementary Table 2), we subdivided the 251 PD patients into three groups [de novo PD, PD patients receiving only DAs, and PD patients receiving levodopa/benserazide (L/B) or levodopa/carbidopa (L/C) with/without other antiparkinsonian medications (DAs, entacapone, selegiline, amantadine, zonisamide, droxidopa), hereafter described as "PD with L/B or L/C"]. We analyzed the level of each levodopa-associated metabolite in each group and compared the results with the HC group using Steel's test (Table 1). No signi cant differences were detected in mean ages at blood sampling or in the sex ratio between the HC and de novo PD, PD with only DA or PD with L/B or L/C groups. Disease duration was signi cantly longer in the PD with L/B or L/C group than in other PD groups. On average, disease severity in the de novo PD and PD with only DA groups was mild-to-moderate according to the H&Y stage and UPDRS-III score.

Analysis of levodopa metabolites
The purpose of administering oral levodopa with an AADC-I and/or COMT-I is to maintain appropriate peripheral blood concentrations of levodopa by reducing conversion to dopamine or 3-methoxytyrosine, respectively. Thus, we measured the concentrations of seven metabolites (levodopa, dopamine, 3-methoxytyrosine, DOPAL, DOPAC, 3-methoxytyramine, and HVA) ( Supplementary Fig. 1). Six metabolites could be reliably measured, while the retention time of the peak associated with DOPAL often uctuated compared with the standard (Supplementary Fig. 2), indicating unsteady chemical characteristics that might include dopamine and/or carbidopa derivatives. Thus, we excluded DOPAL from this study. Changes to the six metabolites in each PD group Dopamine and 3-methoxytyramine levels were below the limit of detection in 39/40 and 39/40 HCs, 15/17 and 17/17 of de novo PD patients, and 14/16 and 15/16 PD patients with only DA, respectively, while both were detected in all patients in the PD with L/B or L/C group. No signi cant changes in levodopa, 3-methoxytyrosine, DOPAC, or HVA were identi ed between HCs and the de novo PD or PD with only DA groups, indicating that they were not suitable diagnostic biomarkers for PD at early stages ( Table 2). As expected, signi cantly increased levodopa, 3-methoxytyrosine, DOPAC, and HVA levels were detected in the PD with L/B or L/C group compared with levels in HCs.

Regulation of levels of levodopa and its downstream metabolites
Because much higher levels of six metabolites were identi ed in the PD with L/B or L/C group compared with other groups (Table 2), we rst estimated levels of the six metabolites for each H&Y stage in the PD with L/B or L/C group. As expected, most variables correlated with disease severity because of progressive medication intensity (Supplementary Table 3); therefore, because of their peripheral mechanisms of action, we investigated the effects of each anti-PD drug (levodopa, entacapone, selegiline, choice of AADC-I, and equivalent dose of DA) on the levels of each metabolite as explanatory variables in linear regression analyses. The levels of ve and six metabolites were signi cantly regulated by each dose of levodopa and entacapone, respectively ( Table 3). Levels of dopamine and 3-methoxytyramine, which are converted to DOPAL/DOPAC or HVA, respectively, by MAOB, were signi cantly correlated with selegiline in a dose-dependent manner ( Table 3). The DA equivalent dose did not contribute to any levodopa, dopamine, or 3-methoxytyrosine level in the PD with L/B or L/C group (Table 3) but contributed to HVA levels in the PD with only DA group with mild signi cance (p = 0.0499) (Supplementary Table 4). Likewise, in the PD with L/B or L/C WITH DA group, DAs mildly contributed only to 3-methoxytyrosine levels with statistical signi cance (p = 0.0402) (Supplementary Table 5), implying little effect of DA on levodopa metabolism.
Levodopa and dopamine levels in the PD with L/B or L/C group were not related to disease severity assessed by H&Y stage (Table 3) or UPDRS-III score (Supplementary Table 6), consistent with the unchanged levodopa pharmacokinetics throughout the course of the illness. 19,20 In Europe and Japan, levodopa (100 mg)/benserazide (25 mg) and levodopa (100 mg)/carbidopa (10 mg) are clinically prescribed. As shown in Table 3, choice of AADC-I signi cantly affected the levodopa level, implying differences in the effects of the administered levodopa and/or AADC-I dose. Therefore, we compared each metabolite concentration between the PD with L/B and with L/C groups without statistical correction (Supplementary Table 7). In contrast to the signi cantly higher dose of levodopa in the L/C group than in the L/B group, lower levodopa and 3-methoxytyrosine levels, as well as higher dopamine, DOPAC, 3-methoxytyramine, and HVA levels, were detected in the L/C group compared with the L/B group, indicative of imperfect AADC inhibition in the L/C group. Consistently, the linear regression analysis shown in Table 3 Table 5). Taken together, we concluded that 25 mg benserazide might have greater inhibitory e cacy on AADC than 10 mg carbidopa.  Concomitant effects of entacapone and/or selegiline with L/B or L/C on levodopa metabolism To con rm the difference in e cacy of AADC-Is, benserazide, and carbidopa, the correlation obtained by linear regression analysis was separately examined for each PD group treated with L/B or L/C. In this cohort, no differences were observed in the COMT SNVs, rs4680 and rs4818, between the L/B and L/C groups (Supplementary Table 7). Although the mean doses of levodopa and entacapone in the L/C group were higher than those in the L/B group, the levodopa levels in the L/C group were inferior to those in the L/B group, indicating a lower AADC inhibitory effect of L/C than with L/B as mentioned above (Supplementary Table 7). Consistently, the entacapone dose signi cantly contributed to levodopa levels only in the PD with L/B group according to both H&Y stage (Table 4) and UPDRS-III score (Supplementary Table 8). Likewise, the leakage tendency of three downstream metabolites (dopamine, DOPAC, and HVA) was more prominent in PD patients treated with L/C with the addition of entacapone. Taken together, the effects of entacapone on levodopa preservation were more pronounced in PD patients treated with 25 mg benserazide than with 10 mg carbidopa. This was con rmed by the observation that levodopa levels were signi cantly in uenced by the choice of AADC-I for PD treated with L/B or L/C with entacapone (Supplementary Table 9, 10) but not WITHOUT entacapone (Supplementary Table 11, 12). Likewise, for PD treated with L/B, a greater correlation was detected between the dopamine and 3-methoxytyramine levels and the selegiline dose, indicating that excessive leakage of dopamine in the PD with L/C group might not be overcome by the peripheral MAOB inhibitory effects of selegiline (Table 4). In accordance with the lack of association between disease stage and levodopa level previously reported, 12 no association of the H&Y stage (Table 4) with levodopa level was detected in this study. However, a mildly signi cant correlation between UPDRS-III score and levodopa level was identi ed only in the PD with L/B group (Supplementary Table 8).  Mathematical expression of two metabolite levels (dopamine and 3-methoxytyramine) As shown in Table 3, each dopamine or 3-methoxytyramine level was signi cantly in uenced by the levodopa, entacapone, and selegiline doses. Thus, we selected appropriate variables with the AIC to clarify which variables are important for estimating the serum concentration of dopamine or 3methoxytyramine. The resultant models are given in (1) and (2) In both cases, levodopa, entacapone, and selegiline were selected by the AIC. For dopamine estimation, all metabolites had positive impacts. However, for 3-methoxytyramine estimation, entacapone had a negative impact, consistent with the levodopa metabolic pathway characteristics, in that 3methoxytyramine production from dopamine is dependent on COMT (Supplementary Fig. 1).

Motor complications and levodopa concentration
The degree of nigrostriatal degeneration and the mode of drug administration are important factors of motor complications [wearing off (WO) and levodopa induced dyskinesia (LID)]. 21,22 The well-established risks of motor complications are age at onset, disease severity, levodopa treatment duration, and levodopa, entacapone, selegiline, or DA dose. 23 Fluctuating serum levodopa levels may be translated into peaks and troughs in striatal dopamine concentration; therefore, we examined the relationship between levodopa concentration and motor complications. As shown in Supplementary Table 13, levodopa levels did not correspond to the presence or absence of WO or LID. AADC rs6263 A/G affects the response to carbidopa The clinical response to L/B or L/C is variable, and approximately 25% of PD patients show a poor response to levodopa because of unknown causes. 24 Thus, we examined six SNVs (rs6950777, rs3735273, rs6263, rs4947580, rs1157457, and rs4490786) in the AADC gene, which are observed in more than 1% of the Japanese population (jMorp, https://jmorp.megabank.tohoku.ac.jp/202001/). As shown in Supplementary Table 14, there was no signi cant change in AADC activity (dopamine/levodopa ratio) for any SNV in all PD patients. Considering the difference in inhibitory (binding) activity against AADC between benserazide and carbidopa, we then determined the difference in AADC activity among groups treated with each AADC-I. AADC activity was not changed by AADC SNVs except that AADC activity was preserved by rs6263 (A→G) in patients treated with L/C, indicative of their poor response to carbidopa (Table 5).   Table 15). No signi cant changes of COMT activity were detected.

Discussion
In this study, we analyzed levodopa metabolic pathway during the systemic circulation, corresponding to about 90% metabolism of it. The absolute serum concentrations of levodopa and ve of its downstream metabolites were signi cantly higher in PD patients receiving levodopa than in HCs. As expected, levels of ve and six metabolites signi cantly correlated with the levodopa and entacapone doses, respectively. The dose of the MAOB-I selegiline did not contribute to the serum levels of levodopa but did contribute to the serum dopamine and 3-methoxytyramine levels, consistent with its blocking point, MAOB, in the pathway. Importantly, signi cant differences in levodopa, 3-methoxytyrosine, DOPAC, and HVA levels between the L/B and L/C groups were detected that were compatible with the higher area under the curve (AUC) of levodopa in response to 25 mg benserazide compared with 10 mg carbidopa. 26 Likewise, concomitant use of L/B with entacapone or selegiline preserved levodopa or dopamine levels, respectively, compared with L/C. Additionally, mathematical models could partially express the correlation of dopamine or 3-methoxytyramine levels with three medications (levodopa, entacapone, and selegiline). Finally, the rs6263 A/G variant in the AADC gene conferred higher AADC activity (dopamine/levodopa ratio) than the wild-type (G/G), indicating a decreased response to carbidopa.
In addition to the positive correlation between total daily levodopa dose and plasma levodopa concentration in PD, 11 pharmacokinetic studies have shown that the Tmax is between 30 and 60 min, and the half-life is approximately 3 hours. 12,19 In this study, blood was obtained within 4 hours after the most recent levodopa administration. The sampling time was randomly allocated every 30 min; therefore, the levodopa levels in the PD with L/B or L/C groups were averaged and might re ect the mean area under the serum concentration curve. According to a recent report, 19 a larger AUC was observed for L/B treatment than for L/C treatment, consistent with our results.
Use of benserazide or carbidopa only increases the amount of levodopa reaching the brain to an estimated 10% of an administered dose because blocking AADC shunts levodopa into the COMT metabolic pathway, thereby increasing peripheral formation of 3-methoxytyrosine. 10 Among all cohorts of this study, 3-methoxytyrosine levels in PD patients were elevated more than 100 times those in HCs. Although no differences in e cacy between L/B and L/C were detected in PD patients treated with levodopa without COMT-Is, [27][28][29] our data suggested that concomitant use of benserazide with entacapone preserved levodopa concentrations more than carbidopa. These results are supported by correlated levels of dopamine or its downstream metabolites with entacapone doses in the L/C but not the L/B group, showing a high level of AADC leakage associated with COMT inhibition. According to previous therapeutic studies, 30 switching from L/B to L/C/entacapone produces similar e cacy in PD as switching from L/C to L/C/entacapone; however, no clinical trials comparing L/B/entacapone and L/C/entacapone in PD have been reported.
Selegiline is primarily metabolized by the liver P450 system (CYP1A2) with some extrahepatic metabolism occurring in platelets. 31 Although platelet MAOB activity, which is inhibited by more than 85% within 4 hours of selegiline administration (5 mg), can be monitored in vitro, 32 measurement of MAOB-I e cacy using serum/plasma has not been reported. In our study, dopamine and 3-methoxytyramine levels in PD patients receiving levodopa were signi cantly enhanced by selegiline, indicating potential responders to MAOB-Is. Further large-scale studies should address possible associations of MAOB SNVs, such as rs1799836, with MAOB activity. 33 In our validation cohort, linear regression analyses revealed no correlation between H&Y stage and levodopa concentration in PD treated with L/B and/or L/C (Table 3, 4), consistent with other reports. 12 Importantly, existence of WO or LID was not in uenced by levodopa concentration in the PD with L/B or L/C group (Supplementary Table 13). These results were compatible with motor complications that occur when the levodopa therapeutic range is narrowed because of the increase in median effective concentrations at the more advanced stage of PD. 34,35 Among three PD groups (untreated, stable, and uctuating), no differences in pharmacokinetics, including mean plasma levodopa peak, were detected. 11,36 However, some patients in the uctuating group showed higher levodopa concentrations, consistent with our results of levodopa levels being related to UPDRS-III scores in the L/B group.
According to a systematic review, 26.9% of pathologically proven PD patients were nonresponsive to oral levodopa treatment. 24 Misdiagnosis, malabsorption because of insu cient acidi cation resulting from Helicobacter pylori infection, and undesired conversion of intestinal levodopa because of excessive bacterial enzymes, including decarboxylase, have been reported; 37, 38 however, effects of AADC SNPs on the enzyme activity have not been identi ed. One of ve haplotypes de ned by 23 SNVs (> 1% minor allele frequencies) without AADC rs6263 is predictive of AADC activity, as measured by [ 18 F]-FDOPA positron emission tomography. 39 In addition, the common AADC polymorphisms (rs921451 and rs3837091) may alter therapeutic responses to levodopa. 40 According to our data, rs6263 (A→G) preserved higher AADC activity than the wild-type allele in the PD with L/C but not the PD with L/B group, indicating insu cient AADC blockage by carbidopa.
A limitation of this study is that it was conducted at a single university hospital, and severe PD cases (H&Y V) were not fully represented because of the history of aspiration pneumonia or cancer exclusion criteria. Although we performed linear regression analysis to evaluate the effect of each medication on each metabolite level, we could not exclude unknown interactions among the drugs. Other antiparkinsonian medications may in uence the metabolism of levodopa; however, at least in our validation cohort, a DA equivalent dose did not correlate with levodopa concentrations. Association studies of levodopa metabolism with genetic background screening for levodopa-associated genes in a large PD cohort should be performed. 41

Ethics statement
The study protocol complied with the Declaration of Helsinki and was approved by the ethics committee of Juntendo University (#2012157). Written informed consent was obtained from all participants.

Participants
All participants were recruited at the Juntendo University Hospital and examined by board-certi ed neurologists. Cohort 1 and 2 were previously reported. 14 PD was diagnosed according to the Movement Disorder Society diagnostic criteria. 1 Exclusion criteria for complications (e.g., dementia) were previously described. 16 Hoehn and Yahr (H&Y) stages and Uni ed Parkinson's disease Rating Scale motor section (UPDRS-III) scores were de ned during the "on" phase for practical and ethical reasons. Entacapone or selegiline were used as COMT-Is or MAOB-Is, respectively. Pramipexole, ropinirole, and rotigotine were used as dopamine receptor agonists (DAs) against parkinsonism, and equivalent DA doses were calculated according to a method previously reported. 42 Sample collection All blood samples were collected at the outpatient department of Juntendo University Hospital between October 2014 and March 2018. Venous blood samples for laboratory analysis were collected between 9:00 am and 12:00 pm. All participants were only allowed to have water and medicines from 12:00 am until sampling. Plasma or serum samples were collected using 7 ml EDTA-2Na blood spits (SRL, Tokyo, Japan) or 8 ml INSEPACK tubes (Sekisui Medical, Tokyo, Japan) with two or three inversions, respectively. Samples were then allowed to incubate for 30-60 min at 4°C followed by centrifugation for 10 min at 2,660 g. The plasma and serum were then separated and placed in collection tubes, which were then stored in liquid nitrogen until analysis.

Sample preparation
Sample preparation of plasma for metabolome analysis was previously described. 14  Genomic DNA analysis During plasma collection, DNA was extracted from peripheral blood according to a standard protocol using a Qiagen kit (Venlo, Netherlands). Primer sequences were designed to amplify the coding exons of AADC and rs4818 and rs4680 of COMT. PCR products were puri ed, and DNA sequences were determined by Sanger Sequencing (Genewiz, South Plain eld, NJ, USA). The frequencies of each variant were evaluated using the Japanese Multi Omics Reference Panel (jMorp, https://jmorp.megabank.tohoku.ac.jp/202001/).

Statistical analysis
All statistical analyses were performed using JMP13 (SAS Institute, Tokyo, Japan). The chi-square test was used to analyze categorical variables.
ANOVA was used to assess the relationships of each clinical parameter among PD groups. Wilcoxon's test or ANOVA was performed to test for statistical signi cance of enzyme activities between two single nucleotide variants (SNVs) or among three SNVs. Steel's test is a nonparametric, multiple-comparison test and was used to examine participant characteristics and levels of levodopa and its metabolites in PD patients and healthy controls (HCs). Linear regression analysis was performed to reveal the in uence of the levodopa, entacapone, or selegiline dose, choice of AADC-I, H&Y stage, and/or UPDRS-III score, followed by variable selection based on the Akaike Information Criteria (AIC). A p-value <0.05 was considered statistically signi cant.
Shin-Ichi Ueno reports no disclosures.