Introduction

Parkinson’s disease (PD) is one of the common neurodegenerative disorders clinically characterized by motor and non-motor symptoms in aged individuals. PD is predicted to affect more than 10 million people worldwide by the year 2030 [1]. Pathohistological hallmark of PD is the presence of abnormally aggregated α-synuclein (Lewy bodies) in the central nervous system (CNS). Lewy body pathology in PD is found in various regions in the CNS, particularly in the substantia nigra. Lewy body pathology is also observed in the peripheral autonomic and enteric nervous system (ENS) in the gastrointestinal tract, the olfactory system, and the cardiac sympathetic system, which suggests that α-synuclein pathology in PD may start in these tissues [2]. PD patients frequently exhibit non-motor symptoms, including signs and symptoms of gastrointestinal dysmotility (e.g. delayed gastric emptying [3] and constipation [4]). Similarly, in PD, accumulation of α-synuclein in the ENS could commence 20 years before the onset of degenerative changes in the CNS and the associated motor symptoms [5]. In accordance with these observations, the smell test [6] and cardiac meta-iodobenzylguanidine scintigraphy [7] are sensitive methods to diagnose early PD. In addition, constipation is the most common premotor symptom in PD, and constipation antedates the development of parkinsonian symptoms by 10 or more years [8]. These findings suggest that Lewy body pathology in the ENS might cause gastrointestinal symptoms in PD. In PD, intestinal permeability is increased, and is correlated with increased intestinal staining for Escherichia coli, α-synuclein, and nitrotyrosine, which is a marker for protein oxidation [9]. Similarly, intraperitoneal administration of lipopolysaccharide (LPS) induces hyperpermeability of the large intestine, and abnormal aggregation of phosphorylated α-synuclein in the large intestine [10]. Additionally, transplantation of gut microbiota of 6 PD patients to transgenic mice over-expressing α-synuclein causes worsening of motor deficits [11]. Oxidative stress produced by macrophages in the luminal wall due to a hyperpermeabilized intestinal wall may account for the accumulation of α-synuclein in the intestinal mucosa. As the intestinal microbiota is likely to have a substantial effect on the hyperpermeability-induced oxidative stress, the intestinal microbiota may be causally associated with α-synuclein pathology in the ENS in PD.

Intestinal dysbiosis in PD has been reported in nine articles from six countries: three from Germany [12–14]; two from the USA [15, 16]; and one from Finland [17], Japan [18], Russia [19], and China [20]. The first report was from Finland, in which Enterobacteriaceae were more abundant in PD with a PIGD (postural instability and gait difficulty) phenotype than in tremor-dominant PD patients [17]. The second report was from the USA, in which Lachnospiraceae was negatively correlated with disease duration [15]. We reported in Japan that the count of Lactobacillus gasseri subgroup was high in advanced PD patients, and the count of Clostridium coccoides group was high in early PD patients [18]. We also reported that bacteria associated with constipation are different from those associated with disease duration. Gut dysbiosis shared by most but not all reports is that Prevotella and butyrate-producing bacteria are decreased in PD. These reports support the notion that gut dysbiosis is associated with the development and progression of PD. In order to identify gut microbiota that are associated with progression of PD, we analyzed changes in gut microbiota in PD in 2 years. We also correlated gut microbiota at year 0 with the progression of PD in 2 years.

Patients and methods

Results

Clinical and laboratory features: Comparison between the deteriorated and stable groups

The 36 patients were evenly divided into 2 groups based on a change in the total UPDRS scores in 2 years. Each group was comprised of 18 patients. Clinical features and serum LBP levels were compared between the deteriorated and stable groups (Table 1). Total UPDRS scores were higher in the deteriorated group than the stable group even at year 0 (p < 0.05). The difference in UPDRS scores between the 2 groups became more prominent at year 2 (p < 0.01). Gender, age, duration, and HY were not different between the 2 groups at year 0. The ratios of patients with habitual consumption of fermented milk were not different between the 2 groups at years 0 and 2. MMSE, MOCA-J, and FAB were not different between the 2 groups at years 0 and 2. Similarly, serum LBP levels were not different between the 2 groups at years 0 and 2. LED tended to increase in two years in both groups but without statistical significance.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Comparison of demographic features between the deteriorated and stable groups.

https://doi.org/10.1371/journal.pone.0187307.t001

Bacterial counts between the deteriorated and stable groups

The detection rate of each bacterium was not different between the deteriorated and stable groups at years 0 and 2. Similarly, there was no difference in total fecal bacterial counts between the 2 groups at years 0 and 2. The deteriorated group had lower Bifidobacterium counts (q < 0.05, FDR correction) than the stable group at year 0 (Fig 1A), suggesting that lower counts of Bifidobacterium at year 0 may be predictive of progression of PD in 2 years.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Prediction of worsening of UPDRS scores using gut microbiota at year 0.

(A) Geometric plot of the counts of Bifidobacterium at year 0 in the deteriorated and stable groups. P-value was corrected by the Benjamini and Hochberg method to calculate the false discovery rate (FDR) (q-value). (B) A stepwise linear regression analysis for predicting a change of UPDRS scores in 2 years using the counts of 10 bacterial groups/genera/species at year 0 yielded the plotted prediction model, which is comprised of the counts of Bifidobacterium (standardized β = -0.45, p < 0.05) and Atopobium cluster (standardized β = 0.50, p < 0.01). (C, D) Correlation between a change of UPDRS I scores in 2 years and the count of Bifidobacterium at year 0 (C) and the count of B. fragilis group at year 0 (D). (E) Correlation between a change of thought disorder scores (Item 2 in UPDRS I) in 2 years and the count of Bifidobacterium at year 0. (F) Correlation between a change of motivation/initiative scores (Item 4 in UPDRS I) in 2 years and the count of B. fragilis group at year 0. (G, H) Correlation between a change of LBP concentrations in 2 years and the count of L. brevis at year 0 (G) and the count of L. plantarum at year 0 (H). (B, C, D, E, and F) Pearson’s correlation coefficients (r) are indicated with respective p-values.

https://doi.org/10.1371/journal.pone.0187307.g001

Bacterial counts between years 0 and 2

Total fecal bacterial counts were decreased in 2 years in all PD patients, as well as in both the deteriorated and stable groups (Table 2 and S2 Table). In the deteriorated group, the count of L. gasseri subgroup became lower in 2 years (S2 Table). In the stable group, the counts of 7 bacteria (Bifidobacterium, Clostridium leptum subgroup, Bacteroides fragilis group, Atopobium cluster, Enterococcus, L. gasseri subgroup, and Lactobacillus reuteri subgroup) became lower in 2 years (Table 2 and S2 Table). When all PD patients were analyzed together, in addition to the 7 bacteria that were decreased in the stable group, Prevotella became lower in 2 years (Table 2 and S2 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Difference of bacterial counts between years 0 and 2.

https://doi.org/10.1371/journal.pone.0187307.t002

Discussion

Emergence of intestinal α-synuclein in aged non-PD subjects [32], as well as in presymptomatic PD patients [33], implies that the PD pathology starts from the intestine. Accumulating evidence suggests involvement of intestinal microbiota in the development of intestinal α-synuclein [9–11]. We [18] and others [12–17, 19, 20] indeed reported intestinal dysbiosis in PD patients. However, the effects of intestinal microbiota on progression of PD have never been addressed. In an effort to disclose intestinal bacteria that are associated with progression of PD, we examined the effects of intestinal microbiota at year 0 on the progression of PD evaluated at years 0 and 2.

We evenly divided the 36 patients into the deteriorated and stable groups based on a change in the total UPDRS scores in 2 years. Higher total UPDRS scores and higher LEDs even at year 0 in the deteriorated group suggest that the deteriorated group already had advanced PD pathology at year 0. As the disease duration was rather shorter in the deteriorated group, the deteriorated group might have had accelerated PD pathology from the beginning. In addition, the difference in intestinal bacteria between the deteriorated and stable groups may be accounted for by the difference in genders in both groups. Although the deteriorated and stable groups were comprised of 44.4% and 77.8% females, respectively (Table 1), there was no statistical significance (p = 0.086 by Fisher's exact test) and bacterial counts at year 0 were not different between males and females [18].

In 2 years, total fecal bacterial count, as well as the counts of 6 out of 10 representative intestinal bacterial groups/genera/species, were decreased in all PD patients (Table 2). A decreased number of total fecal bacteria was previously reported in patients with constipation-type irritable bowel syndrome [34] and aged subjects [35]. The number of bacteria decreased in 2 years was more in the stable group (4 out of 10 bacteria) compared to the deteriorated group (0 out of 10 bacteria). As far as we know, no other reports have addressed temporal profiles of gut microbiota in PD patients. We previously reported that the counts of C. coccoides group, C. leptum subgroup, and B. fragilis group were lower in PD patients than healthy cohabitants [18]. The counts of Bacteroides were also low in PD in Germany [12], China [20] and Russia [19], but not in Finland [17] or the USA [15], and high in the USA [14]. In contrast, the counts of Clostridium were rather high in PD in Finland [17], low in Germany [13] and the USA [14, 15], and not different in Northern Germany [14]. There seems no intestinal bacterium that is specifically changed in PD. Among the 3 bacteria that were low in PD in our previous study [18], the counts of C. leptum subgroup and B. fragilis group were further decreased in 2 years in the stable group, but not in the deteriorated group. One possibility is that there are lower plateau levels for these 2 bacteria, and that the deteriorated group might have already come close to the plateau levels at year 0. Alternative possibility is that the counts of C. leptum subgroup and B. fragilis group were indeed decreased in both groups, but statistical significance was observed only in the stable group by chance. Similar to C. leptum subgroup and B. fragilis group, which were decreased in the stable group, the count of C. coccoides group was also decreased in the stable group but without statistical significance. Thus, the count of C. coccoides group may also rapidly decrease early in the course of PD. In contrast to the 3 bacteria, the counts of total Lactobacillus were higher in PD patients compared to healthy cohabitants in our previous study [18]. Lactobacillus, however, remained unchanged in 2 years in either group. Reevaluation of the count of Lactobacillus at year 0 and the degree of constipation at year 0 revealed that the correlation coefficient between Lactobacillus and stool frequency was -0.475 (p < 0.001), indicating that Lactobacillus was increased with constipation. Lack of change in the severity of constipation in 2 years in both groups (Table 1) may be correlated with lack of change in the counts of Lactobacillus.

Retrospective analysis of intestinal microbiota at year 0 revealed that the deteriorated group had a lower count of Bifidobacterium compared to the stable group (Fig 1A), although we previously reported no difference between PD patients and healthy cohabitants at year 0 [18]. Similar to C. leptum subgroup and B. fragilis group stated above, the Bifidobacterium counts at year 2 were not different between the deteriorated and stable groups (Table 2). This may suggest that the Bifidobacterium count rapidly decreases in patients with accelerated PD pathology. Presumptive differential declines of the counts of C. leptum subgroup, B. fragilis group, and Bifidobacterium between the deteriorated and stable groups suggest that gut microbiota may be critically influenced by the time after onset. Indeed, we previously reported that the count of C. coccoides was lower in PD patients compared to healthy cohabitants [18]. However, the counts of C. coccoides were not different between the deteriorated and stable groups at years 0 and 2 in our current study. The count of C. coccoides might have been decreased shortly after the onset of PD in the deteriorated group, whereas the count might have been decreased sometime after the onset of PD in the stable group. The time when we analyzed the counts of C. coccoides in the two groups might have been too late to detect the difference.

Stepwise linear regression analysis enabled us to make a model to predict a change of total UPDRS scores in 2 years using Bifidobacterium and Atopobium at year 0 (Fig 1B). However, no individual bacterial groups/genera/species at year 0 were correlated with worsening of total UPDRS scores in 2 years. Nonetheless, a low count of Bifidobacterium at year 0 was associated with worsening of UPDRS I (Fig 1C), especially worsening of hallucination/delirium (Fig 1E), in 2 years. Similarly, a low count of B. fragilis at year 0 was associated with worsening of UPDRS I (Fig 1D), especially lack of motivation/initiative (Fig 1F), in 2 years. While Bifidobacterium constitutes the largest group in the infant intestine, Bifidobacterium becomes the third or fourth largest group in the adult intestine and gradually declines with aging [36]. We indeed showed that Bifidobacterium constituted the fourth largest group in healthy cohabitants [18]. Formula-fed babies have a lower number of Bifidobacterium than breast-fed babies [37], which might account for a higher risk of diarrhea and allergies in formula-fed infants [38]. Probiotic administration of Bifidobacterium species ameliorates irritable bowel syndrome [39]. Bifidobacterium produces lactic acid, which regulates intestinal microbial homeostasis, inhibits growth of pathogenic and harmful bacteria, and modulates local and systemic immune responses [40]. Bifidobacterium additionally produces vitamins B1, B2, and K [41]. Probiotic administration of Bifidobacterium for 12 weeks has favorable effects on MMSE score, malondialdehyde, hs-CRP, markers of insulin metabolism, and triglycerides levels in patients with Alzheimer’s disease (AD) [42]. We previously reported that an oxidative stress marker (urine 8-OHdG) is elevated in PD patients with hallucination [43]. As UPDRS I represents severities of mentation, behavior, and mood including hallucination, the beneficial effects of Bifidobacterium on PD and AD, if any, may be partly attributed to its anti-oxidant effect. Bacteroides also constitutes one of the most dominant bacterial species in mammalian gastrointestinal tract. Transplantation of human Bacteroides to the mouse intestine increases the degradation efficiency of branched chain amino acids (BCAA) in the gut, reduces production of BCAA, and decreases circulating levels of BCAA [44]. Since BCAAs compete for entering into the brain with aromatic amino acids (phenylalanine and tyrosine), the level of aromatic acids in the brain decreases, which is followed by a decreased serotonin (5-HT) level in the brain. Worsening of UPDRS I in patients with low Bacteroides counts may represent decreased serotonin levels in the brain.

Low concentrations of LBP enhance LPS-induced expression of interleukin (IL)-1 and tumor necrosis factor (TNF)-α in membrane-bound CD14 (mCD14)-positive cells, or activates expression of IL-6 and IL-8 in mCD14-negative cells. In contrast, high concentrations of LBP neutralize bacterial endotoxin, and downregulates expression of proinflammatory cytokines such as TNF-α [45]. L. plantarum subgroup has lipoteichoic acid, which constitutes a major cell wall component of this species. Lipoteichoic acid of L. plantarum exerts anti-inflammatory effects on human intestinal epithelial cells by blocking IL-8 production [46]. Similarly, L. brevis improves intestinal barrier function in in a murine model of colitis [47]. Oral administration of L. brevis in aged mice for 8 weeks reduces the LPS level in the colon fluid and the blood, and inhibits the expression of inflammatory markers such as myeloperoxidase, TNF-α, and IL-1β, and suppresses NF-κB activation [48]. Elevation of serum LBP in 2 years in PD patients with high intestinal counts of L. brevis and L. plantarum subgroup suggests that these bacteria are likely to be protective against production of proinflammatory cytokines. We previously analyzed the serum levels of IL-6 and TNF-α [18] As we observed no difference in these cytokines between PD and control groups, we did not measure them at year 2. Our current study additionally revealed that there was no difference in these cytokines at year 0 between the deteriorated and stable groups.

As speculated in all sporadic diseases, PD is likely to be caused by an interaction between genetic and environmental factors. According to meta-analysis, prevalence of PD in 70- to 79-year-old population in Asia is significantly lower than those in North America, Europe, and Australia [49]. Similarly, prevalence of PD in all other age groups in Asia is low compared to other regions. Bifidobacterium is dominant in Asian people compared to Western people [50]. We showed that a low count of Bifidobacterium was associated with worsening of PD symptoms. Bifidobacterium may be protective against progression of PD. However, Bifidobacterium was not different between PD patients and controls in Japan [18] and the USA [15]. Additionally, Bifidobacterium was rather higher in PD patients than controls in Germany [12]. A small number of PD patients precluded us from drawing a decisive conclusion. Studies with a larger cohort for a longer time in multiple countries using the identical method for intestinal microbiota analysis [51] may elucidate the causal association of Bifidobacterium and other bacteria in the development and progression of PD.

Supporting information

Acknowledgments

We thank Sae Goto and Satoru Hasegawa at Nagoya University for their technical supports and helpful critiques.