Dietary tributyrin supplementation in Parkinson’s disease: An open-label target engagement study

Abstract

Short-chain fatty acids such as butyrate (key signaling molecules that influence the gut-brain axis and modulate inflammatory, mitochondrial, and transcriptional regulatory processes) are attracting interest as potential treatments for neurodegenerative disorders such as Parkinson’s disease. Oral butyrate supplementation in the form of sodium butyrate suffers from limitations, however, as butyrate is rapidly metabolized by colonocytes, resulting in low plasma and brain concentrations. The butyrate prodrug tributyrin, naturally present in butter, is a neutral short-chain fatty acid triglyceride likely to overcome the pharmacokinetic drawbacks of butyrate. Despite these pharmacokinetic advantages, no clinical studies to date have assessed the safety, tolerability, and target engagement of tributyrin as a postbiotic treatment in the setting of Parkinson’s disease. Tributyrin’s safety profile and potential biomechanistic effects were thus investigated in the setting of Parkinson’s disease via an open-label target engagement study. Fourteen individuals with Parkinson’s disease and three normal controls completed a 30-day (±7 days) intervention of dietary tributyrin supplementation (500 ​mg taken orally three times daily), demonstrating a reassuring safety profile with high rates of adherence. Ten subjects completed [¹¹C]butyrate PET imaging before and after the intervention to assess for treatment-related changes in brain, liver, heart, and gastrointestinal uptake of butyrate, confirming target engagement (i.e., organ-specific changes in butyrate availability). Systemic anti-inflammatory effects were also observed. Exploratory cognitive, motor, and neurobehavioral clinical testing was conducted before and after the supplementation period, identifying associated improvements in cognitive and motor features of Parkinson's disease. Given these findings, tributyrin warrants further investigation via larger, placebo-controlled trials as a potential complementary therapy for Parkinson’s disease.

Introduction

Cognitive decline and postural instability and gait difficulties (PIGD) are common features of Parkinson’s disease (PD) that represent a major therapeutic challenge. Cognitive impairment is a major source of disability and lower quality of life among persons with Parkinson’s (PwP) [1,2]. Mild cognitive impairment is already present in 25–30% of newly diagnosed patients and is a risk factor for the development of Parkinson's disease dementia (PDD) [[3], [4], [5]]. The management of PDD represents an unmet clinical need, as FDA-approved drugs, including cholinesterase inhibitors, have limited to modest symptomatic effects at best and do not reverse progressive deterioration of cognition [6]. Similarly, PIGD motor features become increasingly refractory to levodopa (l-DOPA) with progression of disease [7]. Dementia and PIGD motor features represent the most important contributors to impaired quality of life and disability in PD [8]. Given the lack of sufficient efficacy offered by current treatments, there is an urgent need to investigate alternative therapeutic strategies.

Neuroimaging correlates have demonstrated that progression to dementia reflects both progressive loss of subcortical afferents and intrinsic cortical pathologies. We and others showed that cognitive decline is accompanied and heralded by deficits in regional neocortical glucose metabolism, demonstrated with [¹⁸F]fluorodeoxyglucose (FDG) Positron Emission Tomography (PET) imaging [[9], [10], [11], [12]]. Similarly, several studies have demonstrated a link between abnormalities in glucose metabolism and PIGD motor features in PD [[13], [14], [15]]. Deficits in glucose uptake and metabolism are thought to be closely related to loss of synaptic nerve terminals [16], which are differentially vulnerable due to the high metabolic demand imposed by maintenance of their physiological function [17]. However, recent PET studies utilizing synaptic vesicle glycoprotein 2A (SV2A) – a proxy of synaptic terminal density – in subjects with PD and Alzheimer’s disease (AD) have demonstrated that metabolic deficits at early-to-moderate disease stages (as measured via FDG PET and regional perfusion proxy measures) may precede the widespread loss of synaptic terminals observed in more advanced disease [[18], [19], [20], [21], [22]]. Consequently, treatments aimed at supporting neurometabolic function during the early stages of the neurodegenerative process hold promise not only for symptom relief but possibly even as disease-modifying therapies.

Ketone bodies, derived from fatty acid metabolism, represent a major alternative brain energy substrate. Comparative studies of brain ketone body vs. glucose uptake in subjects with mild cognitive impairment (MCI) and early AD indicate relative preservation (or even up-regulation) of regional brain ketone body metabolism despite the abnormal regional glucose uptake associated with more advanced cognitive impairment, further suggesting relative preservation of synapses [18,19].

Accumulated PET imaging and other data is broadly consistent with the concept that mitochondrial-metabolic dysfunction is a major feature of, and likely pathogenetic mechanism in, PD [23], particularly given the profound reliance of neuromodulatory synaptic function (both dopaminergic and non-dopaminergic) on mitochondrial oxidative phosphorylation of glucose to meet its energetic demands [17]. One implication is that enhancing metabolism of alternative energetic substrates may circumvent abnormalities of glucose uptake and initial glucose metabolism in PD, enhancing-restoring neuronal mitochondrial function and mitigating cognitive deficits secondary to cortical dysfunctions.

The brain appears to utilize multiple energy substrates, especially when its metabolic needs fail to be fully satisfied by glucose metabolism. Ketone bodies provide an alternative substrate via their conversion to acetyl coenzyme A (acetyl-CoA), which subsequently leads to energy production within the Krebs cycle, conferring some degree of compensation for potential impairments in the oxidative phosphorylation of glucose. Ketone bodies also confer a multitude of ‘signaling’ properties, enacting anti-inflammatory, epigenetic, and anti-neurodegenerative (e.g., anti-amyloid) effects [18]. The structural (and possibly mechanistic) similarities between beta-hydroxybutyrate (the body’s most abundant ketone body during endogenous ketosis, accounting for approximately 70% of the circulating ketone body pool by some estimates [24]) and short-chain fatty acids (SCFAs) such as butyrate have garnered increasing interest in SCFA metabolism as another unexplored pathway that may target neurometabolic deficits in PD.

SCFAs (such as acetate, propionate, and butyrate) are naturally produced by the gut microbiome as a byproduct of dietary fiber fermentation and may subsequently be converted to acetyl-CoA via beta-oxidation, which, in turn, contributes to energy production via the Krebs cycle. Similarly to ketone bodies, augmenting systemic availability of SCFAs may fuel alternative energetic pathways to compensate for the abnormalities in glucose uptake and initial metabolism observed in PD. SCFAs may also be significant modulators of inflammatory/oxidative, mitochondrial, and transcriptional regulatory processes linked to neurodegeneration in PD, including α-synuclein aggregation [[25], [26], [27], [28], [29]]. Preclinical experiments suggest that SCFAs have neuroprotective effects beyond their role as an alternative energy substrate [[30], [31], [32], [33], [34], [35], [36]].

Butyrate (a four-carbon SCFA) offers a particularly attractive avenue for intervention relative to other SCFAs due to its greater net energetic yield upon beta-oxidation [37,38]. Increasing systemic butyrate levels via sodium butyrate injection has been shown to confer a protective effect on age-related memory impairments in rodents [31]. Similarly, intermittent fasting (which led to increased systemic butyrate levels) in a 1-methyl-4-phenyl-1,2,3,6-tetrathydropyridine (MPTP)-induced mice model of PD resulted in increased levels of brain-derived neurotrophic factor (BDNF) and inhibitory effects on neuroinflammation [39]. These observations have fueled translational research interest in butyrate given its role in the gut-brain axis [25] and potential benefits in the setting of neurodegenerative disorders [25,[40], [41], [42]]. Animal and human ingestion studies of prebiotic fibers associated with brain health, such as inulin and resistant potato starch, demonstrated selective increases in the production of colonic butyrate and acetate [34,43]. The benefits of dietary fiber supplementation, however, may not extend as readily to PwP, in whom the abundance of butyrate-producing bacteria in the gut has been observed to be lower than in healthy controls [44]. Oral butyrate supplementation in the form of sodium butyrate overcomes this limitation by not relying on the integrity of endogenous production through the microbiome, but still exhibits limited efficacy, as butyrate is metabolized rapidly in colonocytes [45], resulting in low plasma concentrations [46].

Tributyrin, naturally present in butter, is a neutral SCFA triglyceride likely to overcome the pharmacokinetic drawbacks of sodium butyrate and dietary fiber interventions [47,48]. Rapidly absorbed and stable in plasma, tributyrin diffuses through cell/organelle membranes and is metabolized by intracellular lipases, releasing butyrate within cells. Unlike butyrate, tributyrin is lipophilic with a plasma half-life of 40 ​min, making it a preferred choice for systemic oral administration to study direct effects of butyrate on the brain. Given tributyrin’s favorable pharmacokinetics [47,49] and reassuring tolerability profile [48], we selected tributyrin as a SCFA intervention for this target engagement study in PwP. To demonstrate target engagement, we utilized a novel radiolabeled butyrate PET ligand with the hypothesis that dietary tributyrin supplementation would alter regional brain butyrate uptake in PwP. In additional exploratory analyses, we assessed a number of relevant clinical endpoints and pertinent biomarkers.

Methods

Design

This pilot study represents the first human study assessing tributyrin as a potential therapy in Parkinson’s disease. Fourteen participants with PD and three normal controls completed an open-label 30-day (±7 days) intervention of dietary tributyrin supplementation, 500 ​mg taken orally three times per day, with cognitive, motor, neurobehavioral outcomes measured before and after the supplementation period (Table 1). Motor evaluations were performed in the dopaminergic ‘OFF’ state and were completed during the morning portion of study visits prior to the participants' first dose of PD medications. Nine participants (8 PD, 1 NC) underwent optional fMRI scans at baseline and post-intervention to assess functional changes in neural networks. Nine participants (7 PD, 2 NC) underwent early dynamic [¹¹C]butyrate brain PET imaging at baseline and post-intervention to assess for treatment-related regional butyrate uptake changes in the brain. Seven participants (5 PD, 2 NC) underwent additional late static [¹¹C]butyrate whole-body PET imaging at baseline and post-intervention to assess for treatment-related changes in the liver, heart, and gastrointestinal system. Participation in PET imaging was optional to accommodate participants who preferred to avoid radiation. Participants wore biometric devices – specifically, Oura™ rings and Dexcom® continuous glucose monitors (CGMs) – to capture baseline and post-intervention sleep, heart rate variability (HRV), and blood glucose level trends. Biometric devices were worn for 7 (±3) days after the baseline visit and during the last 7 (±3) days of the intervention period. Finally, participants underwent high-sensitivity C-reactive protein (hs-CRP) testing via peripheral blood sampling before and after tributyrin supplementation to assess for changes in systemic inflammation. Post-intervention imaging and laboratory assessments were performed in the fasted state (i.e., participants abstained from their morning dose of tributyrin until completing post-intervention imaging assessments or blood draws), allowing biomarker assessments to isolate the cumulative effects of tributyrin supplementation over the preceding month rather than acute effects of supplementation that morning. For a visual overview of this study design, please refer to Fig. 1.

Table 1.

Exploratory cognitive, motor, and neurobehavioral testing batteries at baseline and post-intervention.

Cognitive testing	Montreal Cognitive Assessment (MoCA)
	Parkinson’s Disease Cognitive Rating Scale (PDCRS)
	Weschler Adult Intelligence Scale IV (WAIS-IV): Information, digit span, matrix reasoning
	Weschler Adult Intelligence Scale III (WAIS-III): Digit symbol coding
	Stroop Color-Word Interference Test
	The F-A-S Verbal Fluency Test
	Boston Naming Test 30-Item Version (BNT-30)
	California Verbal Learning Test II (CVLT-II)
	Benton Judgment of Line Orientation (JLO)
	Delis Kaplan Executive Function System (D-KEFS): Verbal fluency test, trail making test, sorting test
Motor testing (completed in dopaminergic ‘OFF’ state)	Purdue Pegboard Test
	Finger tapping
	Foot tapping
	Timed Up & Go (TUG) Test
	Movement Disorder Society Unified Parkinson’s Disease Rating Scale Part III (MDS-UPDRS Part III)
	Mini Balance Evaluation Systems Test (Mini-BESTest)
Neurobehavioral testing & quality of life measures	Movement Disorder Society Unified Parkinson’s Disease Rating Scale Parts I, II, & IV (MDS-UPDRS Parts I, II, & IV)
	Beck Depression Inventory II (BDI-II)
	Spielberger Trait Anxiety Scale
	Epworth Sleepiness Scale
	Mayo Sleep Questionnaire
	Parkinson’s Disease Questionnaire-39 (PDQ-39)
	Parkinson’s Disease Cognitive Functional Rating Scale (PD-CFRS)

Fig. 1.

Flowchart of study events.

Ethics

This study was approved by the Institutional Review Board of the University of Michigan Medical School (HUM00211320). Procedures followed were in accordance with the Declaration of Helsinki as revised in 1983. All human subjects provided informed consent prior to any study-related tasks. The study was registered with ClinicalTrials.gov (NCT05446168).

Participants

Participants were community-dwelling, with a diagnosis of PD (Hoehn & Yahr stages 2–3), using standard criteria, or normal control participants without a diagnosis of PD. PD diagnosis was based on the United Kingdom Parkinson’s Disease Society Brain Bank Diagnostic Research Criteria [50]. Participants were allowed to continue their normal PD medications during the trial. Given unknown effect sizes, a convenience sample of seventeen participants was enrolled. Exclusion criteria included: evidence of large vessel stroke or mass lesion on MRI; regular use of anticholinergics, benzodiazepines, or antipsychotics; a history of significant gastrointestinal disease; a significant metabolic or uncontrolled medical comorbidity; poorly controlled diabetes; pregnancy or breastfeeding; dementia requiring informed assent; and suicidal ideation.

Intervention

Healus Complete Biotic tributyrin supplement, an over-the-counter postbiotic supplement containing 500 ​mg of liquid tributyrin enclosed in vegetarian capsules, was used for the intervention. Ingredients in this formulation included: liquid tributyrin, hypromellose (capsule), and silica. The tributyrin supplement was provided in bottles of 60 capsules (500 ​mg of active ingredient per capsule). Each participant was given two supplement bottles (a total of 120 capsules) to last the 30-day (±7 days) intervention period.

Tributyrin capsules were clear, odorless, and tasteless. Participants were instructed to take one 500 ​mg capsule three times per day by mouth (1500 ​mg daily in total) during the intervention period. It was recommended that participants take each dose with food. Safety monitoring calls were performed by study team members on a weekly basis during the intervention period to document any adverse events.

Adherence measures

To quantify adherence, a daily log was provided for participants. Participants were instructed to provide checkmarks on a daily basis signifying adherence to each supplement dose (i.e., morning, afternoon, and evening doses). Additionally, upon completion of the intervention period, participants returned empty supplement containers so that study coordinators could confirm the exact number of doses exhausted from containers. Participants who consumed 80% or more of the required doses were designated as ‘adherent,’ with participants consuming 65–80% of the required doses designated as ‘semi-adherent’ and those consuming less than 65% of the required doses designated as ‘non-adherent.’

Adverse events

Adverse events (AEs), recorded and reported as per standard Institutional Review Boards of the University of Michigan Medical School (IRBMED) protocols, were assessed for at each study visit and during regular phone check-ins with study coordinators (typically conducted weekly, with reasonable accommodations for patient preferences or travel plans). Participants were also instructed to self-report any concerns or potential adverse effects encountered, with a low threshold for contacting research staff. Adverse events were classified as mild, moderate, or severe based on standard IRBMED criteria. Adverse events were further classified as related, possibly related, or unrelated to the study based on the nature and timing of a given event. Finally, adverse events were classified as expected or unexpected based on potential events proactively identified as risks in the study protocol. For example, if a participant reported nausea after taking their tributyrin dose, it would be classified as a mild, related, and expected AE. If a participant developed a cold during the intervention period, it would be classified as a mild, unrelated, and unexpected AE. Serious adverse events (SAEs) were defined as events resulting in a life-threatening outcome (e.g., requiring emergency department presentation or hospitalization) or death.

Outcome measures and related methodology

Primary outcome measures: [¹¹C]butyrate brain & body PET

The primary outcome measures for this target engagement clinical trial were brain and whole-body [¹¹C]butyrate uptake changes from pre-treatment to post-treatment. Target engagement of tributyrin supplementation was assessed by performing [¹¹C]butyrate PET brain (n ​= ​7) followed by whole body (n ​= ​5) imaging on participants with PD. [¹¹C]butyrate was prepared as previously described [51]. 3D imaging mode on a Siemens Biograph Vision 600 PET/CT scanner (Siemens Molecular Imaging, Inc., Knoxville, TN), was used to acquire 159 transaxial slices (slice thickness: 1.646 ​mm) over a 26.2-cm axial field-of-view. Dynamic early PET brain imaging was performed for 90 ​min (18 frames) following a bolus intravenous injection of 18 ​mCi [¹¹C]butyrate. For two participants, the PET scan (one pre-treatment and one post-treatment) was terminated prematurely (40–50 ​min post-injection) due to participant request, resulting in only 13 and 14 out of 18 frames, respectively, being acquired for those participants. Static whole-body PET imaging was subsequently performed (only among the subjects with full brain PET acquisitions). Exposure time was 0.6 ​s for body CT, and 1 ​s for brain CT. For whole body PET, the duration of exposure to acquire a single frame was 5 ​min.

Dynamic PET images were motion-corrected and aligned to MRI images using frame-by-frame rigid body transformation. Parametric distribution volume ratio (DVR) images of brain [¹¹C]butyrate uptake were obtained via the voxelwise graphical Logan plot analysis method [52], with perpendicular least squares fitting [53]. Morphologically eroded (2-mm radius sphere) supraventricular cerebral white matter was chosen as a reference region due to relatively low uptake. Within-subject stability of the selected reference region was confirmed using a paired-samples t-test comparing the mean standardized uptake value (SUV) of dynamic imaging frames 10 ​min post-injection between post-treatment and pre-treatment scans. Start time for the linear phase of the Logan plot was set at 10 ​min post-injection (9 frames) and reference k₂ value was set to 0.1 as previously described [51]. For participants with early-termination of the brain PET scan, end time for Logan plot parametric image estimation was set to the latest frame time of the terminated scan (for both pre- and post-treatment images, to avoid potential within-subject bias). Volume-of-interest Logan plots for the two scans with early termination were closely examined and a decision was made to retain them in the analysis, given that the available imaging frames were sufficiently late to reach dynamic equilibrium. For voxelwise analyses, parametric DVR images were warped to Montreal Neurological Institute (MNI) standard space (using a transform estimated from structural MRI images), all voxels other than gray matter were masked out, a 6-mm full-width half-maximum Gaussian smoothing kernel was applied, and lastly the images were resampled to a 2-mm isotropic resolution using linear interpolation. Volume-of-interest (VOI) mean [¹¹C]butyrate values were extracted for a post-hoc effect size regional mapping analysis.

Whole-body [¹¹C]butyrate parametric standardized uptake value ratio (SUVR) PET images were obtained by normalizing the image to the mean of the morphologically eroded (1-mm radius sphere) right ventricular cardiac blood pool signal. TotalSegmentator software was used to first segment CT images which were taken together with the PET, with the standard whole-body segmentation and the cardiac subregions segmentation [54]. After obtaining the SUVR parametric images, mean organ-specific values were extracted for the spleen, liver, pancreas, left and right adrenal glands, duodenum, and myocardium.

Secondary & exploratory outcome measures

Secondary outcome measures included fMRI network flexibility for seven a priori defined networks and changes in systemic inflammation as measured via hs-CRP levels. Blood glucose trends (determined via Dexcom® continuous glucose monitoring) and Oura™ biometric ring sleep metrics (i.e., total sleep duration, REM sleep proportion, deep sleep proportion, and light sleep proportion) were assessed as exploratory outcomes. Cognitive, motor, and neuropsychological clinical tertiary outcome measures were also included in exploratory analyses. Global cognitive Z-score (a composite score of multiple neuropsychiatric domains: memory, visuospatial cognition, verbal fluency, and executive function) was used as an exploratory cognitive outcome, MDS-UPDRS Part III total score was used as an exploratory motor outcome, and PDQ-39 total score was used as an exploratory quality of life outcome. The remainder of the clinical and laboratory test battery served as exploratory outcomes, which are detailed in the supplementary materials section. Safety measures included the Beck Depression Inventory II (BDI-II) to screen for suicidality, patient self-reporting of adverse events, and weekly check-ins with study coordinators to monitor for adverse events.

MRI methods

Magnetic resonance imaging (MRI) was performed on a 3 ​T Philips Achieva system (Philips, Best, The Netherlands). A 3D inversion recovery-prepared turbo-field-echo was performed in the sagittal plane using TR/TE/TI ​= ​9.8/4.6/1041 ​ms; turbo factor ​= ​200; single average; FOV ​= ​240 ​× ​200 ​× ​160 mm; acquired Matrix ​= ​240 ​× ​200 ​× ​160 slices and reconstructed to 1-mm isotropic resolution. Structural images were segmented using the FreeSurfer open-source software. Open-source ANTsPyNet Python module function was used to obtain a brainmask for the structural MRI. A non-linear symmetric normalization transform from the brainmasked structural MRI images to MNI standard space was estimated using ANTsPy open-source Python module function.

Resting-state fMRI scans were acquired using a 32-channel head coil and multi-band sequence, with the following nominal parameters: TR/TE/FA ​= ​720/34/52, 2-mm isotropic resolution, 72 slices. Data was collected with eyes fixated on a cross for 8 ​min. Pre-processing was performed using a standardized fMRIprep pipeline for motion correction and confounder estimation [55]. The first five volumes in the 4-dimensional fMRI image were dropped to exclude magnetic field instability artifacts, and open-source NiLearn Python module [56] was used to implement signal detrending, band-pass filtering (0.01–0.08 ​Hz range), and global signal regression with tri-axial rotation and translation, global, CSF, and white matter averaged signals as confounder regressors, as described in prior work on intrinsic functional connectivity networks [57].

Network flexibility quantification was performed based on procedures previously described in a study of ketogenic diet and exogenous ketone ester fMRI target engagement [58]. Seven resting state brain networks were defined a priori based on FreeSurfer volumes of interest: default mode network (DMN: precuneus, posterior cingulate, medial orbitofrontal cortex, and inferior parietal cortex), dorsal attention network (DAN: superior parietal, inferior parietal, and superior frontal cortices), salience attention network (SAN: rostral anterior cingulate and insula), somatomotor network (SMN: putamen; thalamus; precentral, postcentral, and paracentral cortices), visual network (VIS: cuneus; pericalcarine and lateral occipital cortices), limbic network (LIMB: hippocampus, parahippocampal cortex, and amygdala), and frontoparietal network (FPN: rostral middle frontal and inferior parietal cortices). Network flexibility was computed by extracting the regionally averaged blood-oxygen-level-dependent (BOLD) signal of these pre-defined regions (separately for each hemisphere), splitting the signal into non-overlapping time windows of 24 ​s each, estimating Pearson’s R correlation functional connectivity (FC) matrix within each window, and finally calculating the mean scalar norm of the difference between temporally adjacent FC matrices [58]. The obtained measure conceptually represents the magnitude of temporal fluctuation in functional connectivity between regions within a network.

Continuous glucose monitoring methods

Participants wore Dexcom® continuous glucose monitors for 7 (±3) days after the baseline visit (prior to intervention initiation) and during the last 7 (±3) days of the intervention period. Three measures of glucose lability were quantified from the Dexcom® CGM raw readouts (obtained every 5 ​min): mean glucose levels, mean proportion of time spent within normative range (60–140 ​mg/dL), and mean number of peaks above 170 ​mg/dL. The measures were calculated at a daily level, and then averaged across all days before vs. after intervention onset. The effect of tributyrin treatment on proportion of measurements below the hypoglycemic threshold (blood sugar <55 ​mg/dL) was also assessed.

Oura™ biometric ring monitoring methods

Oura™ biometric devices were worn for 7 (±3) days after the baseline visit (prior to intervention initiation) and during the last 7 (±3) days of the intervention period. Sleep outcomes included: total sleep duration, REM sleep proportion, deep sleep proportion, and light sleep proportion. Additional biometric evaluations included in supplementary analyses assessed trends in mean heart rate, mean HRV, mean nocturnal heart rate, and mean nocturnal HRV.

Global cognitive Z-score calculation

An external sample of 55 cognitively normal (MoCA score greater than 25) control subjects were utilized to obtain the global cognitive Z-score (included in main analysis) and cognitive domain Z-scores (included in supplementary analyses). For the F-A-S Verbal Fluency Test, JLO, CVLT short-term memory, CVLT long-term memory, modified Stroop, and Trail Making Test B (letter and number switching), normative linear regression models were estimated in the control sample, including confounding effects of age and sex as covariates. Normative z-scores for each test among patients were then computed by subtracting from their raw scores the value predicted by the control model (based on the patient’s age and sex) and dividing this difference by the standard deviation of control model residuals. The resulting measure represents the signed (directional) deviation of patient values from the mean of controls in control standard deviation units after holding constant the biological effects of age and sex. To obtain cognitive domain normative z-scores, JLO normative z-score was used to define the visuospatial domain, F-A-S Verbal Fluency Test z-score was used to define the verbal fluency domain, Trail Making Test B and modified Stroop were multiplied by negative one (so that higher values reflect better cognition) and averaged to define the executive domain, and WAIS backwards digit span and CVLT short-term and long-term memory normative z-scores were averaged to define the memory domain.

Statistical analysis

Brain butyrate uptake was compared within-subject before and after treatment with a voxelwise paired-samples t-test using the SPM12 software. Resulting statistical parametric maps were thresholded at a voxelwise significance threshold of p ​< ​0.05, and a cluster-forming threshold of 25 voxels. Both positive and negative comparison contrasts (increases and decreases in [¹¹C]butyrate uptake as a function of treatment) were examined. Cluster-level findings were considered statistically significant at a familywise error rate (FWE)-corrected significance threshold of p ​< ​0.05. Both positive and negative comparison contrasts for the biasing effect of early PET scan termination were examined for statistical significance. After identifying clusters with statistically significant changes in [¹¹C]butyrate uptake, average uptake in these clusters was extracted from spatially normalized parametric images for additional analysis (see next paragraph). Lastly, regional localization of statistically significant voxels was performed in MNI152 standard space using the cerebellar SUIT (cerebellar lobules and deep nuclei) [59], BigBrain (subcortical regions) [60], and Brodmann area (cortical regions) atlases [61].

Tributyrin supplementation-related within-subject changes in clinical and biological variables were tested using a random intercept mixed linear model approach. The null hypothesis model was defined as containing only a random intercept term and relevant covariates, while the alternative hypothesis model was defined as containing an additional fixed effect of treatment, which was used to compare the post-treatment visit measurements against the baseline pre-treatment visit. Relevant covariates included age, sex, and disease duration for all models. Education level was included as an additional covariate when modeling cognitive outcome measures. Levodopa equivalent dose (LED) was included as an additional covariate when modeling motor outcome measures. Model comparisons were performed using a likelihood ratio test. Statistical significance of model comparisons for primary target engagement outcome measures (central and peripheral [¹¹C]butyrate regional uptakes) was determined based on a Bonferroni-adjusted α significance threshold (0.05/9 comparisons ​= ​0.00556). An uncorrected α significance threshold of 0.05 was used for secondary and tertiary outcome measures. Model residuals were checked for normality, and alternative generalized mixed linear model families and link functions were used in case non-normality of residuals was observed. Statistical models for all analyzed exploratory outcome measures are presented in supplementary materials. Exploratory measures with an absolute value of confounder-adjusted standardized treatment effect regression coefficient (β) greater than 0.3 are reported. An additional post-hoc VOI-based effect size mapping analysis was performed (comparing pre-vs. post-treatment uptake) on regional brain DVR extracted from native space PET images using the standard FreeSurfer VOI. β coefficients corresponding to the standardized effect size of treatment were mapped unto cortical and subcortical visualizations using the open-source R programming language ggseg library.

Results

Recruitment and baseline participant data

From April of 2022 to June of 2023, 46 subjects were screened. 28 subjects did not meet inclusion criteria. 18 subjects were enrolled in the open-label study (15 PD; 3 NC). One PD subject withdrew from the study prior to baseline testing due to concerns about the imaging requirements. One NC subject was withdrawn prior to completing the post-intervention testing due to repeated inappropriate behavior directed at staff (furthermore, his examination findings were suggestive of undiagnosed parkinsonism, and he was therefore excluded from the NC analysis). The remaining 16 subjects (14 PD; 2 NC) completed the study protocol, with Table 2 detailing the baseline characteristics of these subjects.

Table 2.

Baseline characteristics of study subjects.

Variable	PD (n ​= ​14)	NC (n ​= ​2)
Average age (years)	66 ​± ​5.7 (57–77)	62.5 ​± ​9.2 (56–69)
Sex (M:F)	9:5	0:2
Average education (years)	17.9 ​± ​2.7	17.5 ​± ​2.1
Average BMI (kg/m2)	26.7 ​± ​4.1	24.6 ​± ​3.0
Average H&Y stage	2.5 (2–3)	0 (0–0)
Average UPDRS-III score	46.9 ​± ​12.7	16.3 ​± ​3.9
Average MoCA score	26.8 ​± ​2.9	28 ​± ​1.4
Average PDCRS score	92 ​± ​12.3	95 ​± ​4.2
Carbidopa/Levodopa status (taking: not taking)	12:2	0:2
Median levodopa dose (mg)	425 (0–2450)	N/A

Tributyrin supplement characteristics

All Healus Complete Biotic used in this study came from the same lot number (6105741). The batch utilized for this study was manufactured on 3/26/2021. Purity was verified by Capsugel’s Certification of Analysis on 6/4/2021.

Adherence outcomes

14/17 participants (82.35%) were adherent to the study protocol (defined as 80% or more of doses administered). The remaining 3/17 (17.65%) participants were semi-adherent to the supplementation protocol (defined as 65–80% of doses administered). Zero participants were non-adherent to the supplementation protocol (defined as <65% of doses administered).

Adverse events

Tributyrin 500 ​mg TID PO for 30 days was well tolerated, with only transient mild gastrointestinal upset or bloating (both expected side effects) reported by seven subjects. There were no related moderate or serious adverse events.

Dropouts

There were zero dropouts. One NC subject was withdrawn from the study prior to his post-intervention imaging visit due to inappropriate behavior directed at study staff.

Functional imaging results

Primary outcome measures: [¹¹C]butyrate brain & body PET

No statistically significant within-subject differences were observed in the mean SUV of the morphologically eroded supraventricular white matter reference region (t ​= ​0.087, p ​= ​0.934, df ​= ​6). Group averaged SUV time activity curves (see Fig. 2) demonstrate that supraventricular white matter reliably exhibits lower uptake than gray matter target regions and on average remains below an SUV of 1, indicating no significant uptake of the radiotracer. Regionally averaged DVR values from the obtained parametric images (see Fig. 3) show relatively high uptake in the mesial posterior and midcingulate cortices, thalamus, putamen, and cerebellum. Intermediate uptake was observed in the lateral fronto-temporal cortices, mesial frontal cortices, hippocampus, and amygdala. Relatively low uptake was observed in the sensory-motor cortices, brainstem, pallidum, and caudate nucleus.

Fig. 2.

Group-averaged standardized uptake value time activity curves.

Standardized uptake values (SUV) were obtained by dividing the image voxel intensities (MBq/mL) by a ratio of injected dose (in MBq) over body weight (in grams). Time activity curves were averaged across 6 PD patients with full acquisitions (2 patients with early termination of PET scan not included). A dashed gray line at 10 minutes after injection indicates the start time of the Logan plot linear phase used to obtain distribution volume ratio (DVR) images.

Fig. 3.

Whole-sample level pre-treatment average of cerebral [¹¹C]butyrate distribution volume ratio (DVR) by region in the sample of 6 PD subjects with full-duration scan. Higher [¹¹C]butyrate is observed in the posterior mesial cortices and cerebellum, followed by the neocortex, putamen, thalamus, and limbic system. Lowest uptake was observed in the pericentral cortices, pallidum, and brainstem.

Standardized uptake values (SUVs) were obtained by dividing the image voxel intensities (MBq/mL) by a ratio of injected dose (in MBq) over body weight (in grams). Time activity curves were averaged across six PD subjects with full acquisitions (one patient with early termination of baseline PET scan not included). A dashed gray line at 10 ​min after injection indicates the start time of the Logan plot linear phase used to obtain DVR images.

Statistically significant (N_vox ​= ​2130, Cluster-FWE p ​= ​0.007) reduction in brain [¹¹C]butyrate uptake was observed after one month of tributyrin treatment, suggesting that greater regional brain uptake of “cold” (i.e., non-radiolabeled) butyrate induced by supplementation resulted in increased competition with “hot” (i.e. radiolabeled) [¹¹C]butyrate uptake (in other words, after a month of supplementation, the radiolabeled butyrate administered for PET imaging appeared to be competing with supplemented, non-radiolabeled butyrate for binding sites in the brain – these observed changes from baseline [¹¹C]butyrate PET imaging to post-intervention [¹¹C]butyrate PET imaging suggest bioavailability in the brain). No statistically significant increases in brain [¹¹C]butyrate uptake were observed as a function of treatment. Across cortical, subcortical, and cerebellar subregions, treatment-related decreases in [¹¹C]butyrate uptake were primarily lateralized to the right cerebral and cerebellar hemispheres, including cerebellar lobules Crus I/II, VI, VII, VIII, and IX, along with cortical fusiform and parahippocampal gyri (see Fig. 4). SPM cluster DVR treatment-related changes by subject are visualized in Fig. 5, with relevant statistics presented in Table 3. An additional post-hoc effect size mapping of treatment across regional mean uptakes are presented in supplementary materials (see Figs. S1 and S2). Effect size mappings support greater involvement of the cerebellum and the right (more than left) cerebral hemisphere.

Fig. 4.

Within-subject paired t-test SPM group comparison performed in a sample of 7 subjects with PD. The colorbar values range between 0 and 8.

Fig. 5.

Mean change in [¹¹C]butyrate brain uptake for 7 PD subjects.

Table 3.

Within-subject group comparisons of [¹¹C]butyrate PET organ uptake values before and after tributyrin treatment.

Variable	N	β	χ2	P
Brain	7	−0.484 [-0.716, −0.252]	10.698	0.001∗
Spleen	5	−0.542 [-0.848, −0.236]	9.053	0.003∗
Liver	5	−0.299 [-0.403, −0.196]	11.845	0.001∗
Pancreas	5	−0.32 [-0.656, 0.017]	3.525	0.06
Left adrenal gland	5	−0.525 [-1.404, 0.355]	1.542	0.214
Right adrenal gland	5	0.045 [-1.009, 1.098]	0.008	0.927
Duodenum	5	−0.353 [-0.969, 0.264]	1.604	0.205
Myocardium	5	−0.892 [-1.456, −0.327]	7.736	0.005∗
Colon	5	0.001 [-0.617, 0.62]	0.000	0.996

Effect size of tributyrin treatment is presented as standardized regression coefficients (β) with 95% confidence intervals, adjusted for confounding effects of age, sex, and disease duration. Likelihood ratio test model comparison P-values relative to a null model with no effect of treatment and only covariates are presented.

Within-subject comparisons for organ-specific [¹¹C]butyrate uptake before and after tributyrin intervention were performed in a sample of five PD subjects for peripheral organs and seven PD subjects for the brain (Table 3). The directionality of change in uptake as a function of treatment for peripheral organs was the same as for the brain (decreased uptake), so the same interpretation is applied to non-brain organs, wherein decreased post-treatment uptake suggests greater competition of supplemented “cold” butyrate with “hot” [¹¹C]butyrate. The most notable statistically significant changes were observed in the brain, spleen, liver, and myocardium. Treatment-related changes in the pancreas also trended toward a decrease, but failed to achieve statistical significance.

Secondary outcome measure: fMRI neural network flexibility

Within-subject comparisons of fMRI neural network flexibility before and after tributyrin intervention were performed in a sample of eight PD subjects (Table 4). The most pronounced treatment-related increase in network flexibility was observed in the visual network, with a more modest increase observed in the default mode network. Notably, a disease duration-related decrease in default mode network flexibility was observed, which appeared equivalent in magnitude to the treatment-related increase in flexibility.

Table 4.

Within-subject group comparisons of fMRI neural network flexibility before and after tributyrin treatment.

Network	βDuration	βVisit	χ2	P
Default mode	−0.616 [-1.063, −0.169]	0.698 [0.067, 1.329]	4.500	0.034∗
Dorsal attention	−0.18 [-0.718, 0.358]	0.3 [-0.585, 1.185]	0.533	0.466
Salience attention	−0.234 [-0.709, 0.241]	0.043 [-0.861, 0.948]	0.010	0.920
Somatomotor	0.223 [-0.28, 0.725]	−0.262 [-1.219, 0.695]	0.322	0.570
Visual	−0.209 [-0.517, 0.099]	1.299 [0.713, 1.886]	13.551	<0.001∗
Limbic	−0.069 [-0.712, 0.574]	0.129 [-0.742, 1.001]	0.108	0.742
Frontoparietal	−0.083 [-0.547, 0.381]	0.776 [-0.107, 1.66]	3.043	0.081

Peripheral biomarker outcomes

Secondary outcome measure: high-sensitivity C-reactive protein (hs-CRP) results

Baseline vs. post-intervention high sensitivity C-reactive protein (hs-CRP) comparisons were performed for thirteen PD participants (one participant was excluded due to having an upper respiratory infection during the post-intervention assessment). Model comparison between a null model containing only sex, age, and disease duration as covariates and the alternative model containing the additional effect of treatment demonstrated that treatment accounted for an additional proportion of variance in hs-CRP levels (β ​= ​−0.586 [−1.131, −0.04], χ² ​= ​4.561, p ​= ​0.0327). An exponential mixed linear regression model (family ​= ​Gamma, link ​= ​log) appeared to provide a substantially better fit to the data (AIC ​= ​27.216) than the standard mixed linear regression (AIC ​= ​74.604). Exponential model comparison still demonstrated a substantial improvement in fit from a confounder model to a treatment model (χ² ​= ​10.680, p ​= ​0.0011). Treatment-related effect according to the exponential model was a 34.9% reduction in hs-CRP levels from pre-treatment to post-treatment (e^β ​= ​0.651 [0.511, 0.830], p ​= ​0.002). Hs-CRP values before and after treatment by adherence subgroup are visualized in Fig. 6.

Fig. 6.

Changes in systemic inflammation as a function of treatment by adherence group.

In a sample of thirteen PD subjects, we performed an additional post-hoc exploratory analysis to examine whether an association existed between hs-CRP levels and treatment-related improvement in H&Y stage. A model assuming independent effects of treatment and hs-CRP was compared with a model assuming an interaction effect between treatment and hs-CRP. The interaction model offered a better fit to the data than the comparator model assuming independent effects of treatment and hs-CRP (χ² ​= ​21.108, p ​< ​0.001). An exponential model again offered better fit to the data (AIC ​= ​−17.49) than a linear mixed model (AIC ​= ​55.08). Higher hs-CRP at baseline was associated with worse baseline H&Y scores (e^β ​= ​1.04 [1.01, 1.08], p ​= ​0.016). Coefficients of the exponential model suggested that among participants with average hs-CRP levels at baseline, an improvement in H&Y scores of 15% was observed as a function of treatment (e^β ​= ​0.85 [0.82, 0.89], p ​< ​0.001). Lastly, individuals with one standard deviation higher hs-CRP values at baseline were found to have a greater improvement in H&Y scores with treatment, on average 30.3% (e^β ​= ​0.82 [0.78, 0.86], p ​< ​0.001).

Exploratory sleep outcomes

No change in total sleep duration (χ² ​= ​2.500, p ​= ​0.114), REM sleep proportion (χ² ​= ​0.409, p ​= ​0.522), or light sleep proportion (χ² ​= ​0.693, p ​= ​0.405) was observed. Linear mixed model of deep sleep proportion did not indicate significant changes (χ² ​= ​1.757, p ​= ​0.185), however some evidence of non-normality in model residuals was observed, so a Gamma generalized linear mixed model (GLMM) with log link function was used. The GLMM offered a better fit to the deep sleep proportion data (AIC ​= ​−83.338) than the linear mixed model (AIC ​= ​72.546). Model comparison indicated statistically significant change in the deep sleep proportion outcome (χ² ​= ​5.479, p ​= ​0.0192). Tributyrin treatment was associated with a 20.8% increase (e^β ​= ​1.208 [1.034, 1.412], p ​= ​0.02) in deep sleep proportion (∼15 ​min per night).

Exploratory continuous glucose monitor (CGM) outcomes

CGM biomarkers were examined in a sample of eleven PD participants (for whom data acquisition was successful). Mean glucose levels (χ² ​= ​0.916, p ​= ​0.338), proportion of measurements within the normative range (χ² ​= ​0.573, p ​= ​0.449), and mean number of peaks above 170 ​mg/dL (χ² ​= ​0.00218, p ​= ​0.963) did not differ pre- and post-treatment. Hypoglycemia was, on average, observed in 0.4% of measurements before treatment, and in 0.01% of measurements after treatment, suggesting that the risk of tributyrin treatment-induced hypoglycemia is negligible.

Exploratory clinical outcomes

In exploratory analyses of cognitive outcomes, tributyrin supplementation was associated with improvements in global cognition, short-term and long-term memory, and visuospatial reasoning. In exploratory analyses of motor outcomes, tributyrin supplementation was associated with improvements in MDS-UPDRS Part III scores (global assessment of motor features), Hoehn & Yahr disease staging scores, and backwards walking speed.

Cognitive: The treatment model accounted for a greater portion of variance in global cognitive z-score than the confounder model (χ² ​= ​10.547, p ​= ​0.00116). Treatment-related improvements in global cognitive z-score were observed (β ​= ​0.405 [0.184, 0.625], p ​= ​0.001). Notably, greater age (β ​= ​−0.526 [−0.984, −0.068], p ​= ​0.027) was a significant independent predictor of lower global cognitive z-score and greater disease duration trended in the same direction (β ​= ​−0.402 [−0.885, 0.081], p ​= ​0.098). Additional exploratory analyses revealed that changes in the global cognitive z-score were primarily reflective of improvements in the memory and visuospatial domains (see Supplementary Table S2). Exploratory analyses of individual test battery items revealed strong trends for improvement, specifically on the MoCA, WAIS-IV matrix reasoning, CVLT short-term and long-term memory, and JLO assessments (see Supplementary Table S3).

Motor: The treatment model accounted for a greater portion of variance in MDS-UPDRS Part III than the confounder model (χ² ​= ​4.670, p ​= ​0.031). Treatment-related decreases in MDS-UPDRS Part III were observed (β ​= ​−0.609 [−1.199, −0.020], p ​= ​0.043). Notably, greater age (β ​= ​0.399 [−0.020, 0.818], p ​= ​0.061) and levodopa equivalent dose (β ​= ​0.497 [−0.064, 1.058], p ​= ​0.079) strongly trended toward a positive association with MDS-UPDRS Part III total scores. Additional exploratory analyses revealed promising trends for improvements in Hoehn & Yahr scores and backward walking speed (see Supplementary Table S4).

Neuropsychiatric: Treatment model trended toward accounting for greater portion of variance in PDQ-39 total score than the confounder model (χ² ​= ​2.733, p ​= ​0.0983), but failed to reach statistical significance. A trend for treatment-related improvements in self-reported quality of life was observed (β ​= ​−0.270 [−0.620, 0.067], p ​= ​0.109). Notably, greater age (β ​= ​0.475 [0.014, 0.936], p ​= ​0.044) was a significant independent predictor of poorer self-reported quality of life. Exploratory analyses yielded no additional promising trends for neuropsychiatric outcomes (see Supplementary Table S5).

Discussion

Summary of key findings

Target engagement studies represent a critical first step in the development of novel therapeutics by testing the basic hypothesis that a potential therapy engages the expected mechanism of action [62]. Neuroimaging methods utilizing magnetic resonance spectroscopy have been successfully applied in the PD population to test hypotheses about target engagement [63]. This pilot study provides the first in-human data on the effects of tributyrin as a potential therapy for Parkinson’s disease. A novel radiolabeled butyrate PET tracer was synthesized to evaluate regional target engagement in the brain and body associated with a 30-day tributyrin intervention, demonstrating successful target engagement (i.e., increased availability of non-radiolabeled butyrate) in the brain, myocardium, liver, and spleen. Exploratory clinical analyses revealed that tributyrin supplementation was associated with improvements in both cognitive and motor features. A concise summary of these results can be found in Table 5.

Table 5.

Summary of key findings.

Domain	Measure(s)	Outcome(s)
Target engagement	[11C]butyrate regional uptake	Target engagement in brain, myocardium, liver, & spleen
Safety profile	AE & SAE reporting	Reassuring safety profile
Systemic inflammation	Hs-CRP	Reduced systemic inflammation
Cognitive features	Global cognitive z-score, CVLT short-term and long-term memory scores, JLO score, MoCA score, WAIS-IV matrix reasoning score	Improved global cognition, visuospatial reasoning, short-term memory, and long-term memory; trends for improvements in MoCA score and WAIS-IV matrix reasoning score
Motor features	MDS-UPDRS Part III score, H&Y disease stage score, backward walking speed, foot tapping speed	Improved global motor feature scores; trends for improvements in H&Y disease stage scores, backward walking speed, and foot tapping speed
Neural network flexibility	fMRI-derived neural network flexibility	Improved neural network flexibility observed in visual network and default mode network
Blood glucose	Blood glucose levels	No significant changes in blood glucose levels, suggesting mechanistic differences in comparison to beta-hydroxybutyrate
Quality of life	PDQ-39 score	Trend for improved quality of life

Strong trends were noted for improvement of both cognitive features – specifically, visuospatial skills, memory, and global cognition – and motor features – specifically, global motor features as assessed via the MDS-UPDRS Part III, bradykinesia as assessed via foot tapping speed, and coordination/balance as assessed via backward walking speed. These findings indicate that tributyrin warrants further investigation as a potential adjunctive therapy for cognitive and motor features of PD. Pertinently, this exploratory clinical analysis is inherently limited in terms of statistical power and by lack of placebo control in this open-label pilot study, though suggestive of the possibility that tributyrin supplementation improves cognitive and motor features in PD while maintaining a reassuring safety profile. Larger, placebo-controlled studies are needed to further elucidate the clinical and biomechanistic effects of tributyrin supplementation in the setting of PD.

Mechanistic insights

The observed reduction in systemic inflammation, as indicated by decreased hs-CRP levels, aligns with the known anti-inflammatory properties of butyrate. This suggests that tributyrin may mitigate systemic and possibly neuroinflammation in PD [[64], [65], [66], [67], [68]]. Moreover, our subanalyses revealed that individuals with higher levels of systemic inflammation at baseline demonstrated worse clinical status at baseline and more pronounced clinical improvement (as ascertained via H&Y stage) with tributyrin treatment. Convergence of peripheral biomarker assessments with functional imaging techniques targeting the central nervous system and gut-brain axis are necessary to further elucidate the biomechanistic effects of butyrate supplementation in the setting of PD.

This target engagement study represents the first utilization of radiolabeled [¹¹C]butyrate to assess target engagement of tributyrin supplementation in individuals with PD via PET imaging. Statistically significant changes in uptake patterns were observed in the brain, myocardium, liver, and spleen when comparing post-intervention [¹¹C]Butyrate uptake to pre-intervention [¹¹C]Butyrate uptake. Specifically, lower binding of “hot” (i.e., radiolabeled) butyrate with regional specificity was observed in the post-intervention condition, suggesting that greater bioavailability of “cold” (i.e., non-radiolabeled) butyrate in the brain and peripheral organs following supplementation resulted in increased competition with “hot” butyrate for binding sites. Treatment-related changes of peripheral butyrate availability were also observed, notably in the spleen, liver, and myocardium, with trends for availability changes observed in the pancreas, left (but not right) adrenal gland, and duodenum. These findings support the hypothesis that tributyrin supplementation leads to elevated availability of butyrate in the brain and other target organs, where it might serve either as an alternative fuel source or a pleiotropic signaling molecule. While 95% of naturally produced butyrate is absorbed by colonocytes [69], the excess butyrate produced via breakdown of the tributyrin supplement may enter systemic circulation and reach downstream targets in greater quantities than naturally occurring butyrate. Our observation of changes in brain uptake agrees with known pharmacokinetics of short-chain fatty acids, with in-vitro studies demonstrating that butyrate is able to cross the blood-brain barrier primarily via diffusion and, in part, via fatty acid translocase transporter proteins [70].

One potential set of molecular targets engaged by butyrate both centrally and peripherally are free fatty acid receptors (FFARs), which are G-protein coupled receptors that trigger downstream metabolic signaling cascades upon ligand binding [71]. Among the four known FFARs, FFAR1 is most abundantly expressed in the brain, while FFAR2-4 are more abundant in peripheral tissues [71]. We observed an increased availability of butyrate in peripheral organs following tributyrin supplementation, primarily in the spleen, liver, and myocardium. The spleen [72] and liver [73] are known to express FFAR2/3 receptors, through which butyrate may exert effects on global bioenergetics and systemic inflammation. Target engagement observed in the myocardium may reflect binding to FFAR3 expressed on sympathetic nerve terminals innervating the heart, which have been shown to modulate the release of norepinephrine in response to increased SCFA concentration [74,75]. Cardiac noradrenergic denervation is a well-recognized early signature of Parkinson disease [76,77], and has recently been shown to associate with the severity of neuropsychiatric symptoms [78] and gait impairment [79] in patients. Taken together, these findings suggest that elevated availability of butyrate induced by tributyrin supplementation may result in target engagement across a range of peripheral targets known to express receptors with high affinity for SCFAs and through which butyrate may exert effects on global bioenergetic, immune, and neuroendocrine processes.

Immunohistochemistry studies in non-human primate models demonstrate relatively greater prevalence of FFAR1 receptors in the cerebral cortex, hippocampus, amygdala, and cerebellum [80], which broadly aligns with the distribution of regional [¹¹C]butyrate uptake observed pre-treatment among human participants in the present work. Accumulating evidence appears to suggest that FFAR1 receptors in the brain play an important role in processes related to neurogenesis, memory formation, and visuospatial cognition [81], which may help contextualize the domain-specific improvements in visuospatial cognition and memory observed following tributyrin supplementation. Medium-to-long-chain fatty acids have greater affinity for FFAR1 than short-chain fatty acids – in isolation, short-chain fatty acids such as butyrate fail to elicit changes in intracellular calcium levels of FFAR1-expressing cell cultures (binding at the FFAR1 active site enacts the classical Gq/11 protein pathway, leading to phospholipase C activation and intracellular calcium release, though FFAR1 can also enact calcium-independent signaling pathways via other mechanisms) [[82], [83], [84]]. Notably, the arginine residues that coordinate the binding and activity of butyrate at FFAR2 are largely conserved in FFAR1 [85]. Arginine residues contribute to allosteric modulation of FFAR1 activity at multiple secondary binding sites [86]. Lastly, recent computational work on molecular docking reports that, in comparison to the SCFAs acetate and propionate, butyrate is expected to have the greatest affinity for all 4 FFARs, including FFAR1 [87]. In conclusion, despite the lack of conclusive evidence for specific activity of butyrate at FFAR1, the possibility of indirect contributions to FFAR1 activity as an allosteric modulator cannot be excluded. Moreover, supplementation-induced abundance of systemic butyrate availability may centrally target FFAR2/3 receptors, which are relatively less abundant in the brain but have much higher affinity for SCFAs [71].

There is a substantial overlap between the brain regions where we observed tributyrin target engagement and those where we previously reported correlation between cholinergic system function as assessed by [¹⁸F]fluoroethoxybenzovesamicol ([¹⁸F]FEOBV) vesicular acetylcholine transporter PET imaging and incidence of freezing [88] as well as cardinal motor symptoms of rigidity and tremor [89]. Furthermore, cortical regions where tributyrin supplementation led to increased availability of butyrate include right-lateralized Brodmann areas 2, 6, 7, 10, 19, 21, 22, 37, 40, 44, 45, and bilateral Brodmann area 47, which almost fully recapitulates the cortical pattern of brain [¹¹C]methylpiperidin-4-yl propionate ([¹¹C]PMP) acetylcholinesterase PET uptake associated with impaired multisensory integration-dependent postural control among PD freezers that we previously reported [90]. Our finding of improved backward walking speed is noteworthy, since backward walking is strongly dependent on effective somatosensory-vestibular integration due to the absence of reliable visual signals. These improvements in multisensory gait and balance control may be due, at least in part, to the beneficial effect of greater butyrate availability on energetically costly neuromodulatory signaling in multisensory integration-related regions.

In conjunction with evidence of greater cerebral butyrate availability following tributyrin supplementation, the fMRI network flexibility analysis suggests that increased availability of butyrate in the brain may support the flexible reconfiguration of long-range functional connectivity networks. More flexible reconfiguration of resting state functional networks as observed on fMRI has been previously associated with critical brain dynamics [91], which give rise to enhanced long-range spatio-temporal correlation across brain regions and bolster neural information integration capacity [92]. Given the important role of neuromodulation in the maintenance of global brain activity patterns [93], it is possible that increased availability of butyrate following tributyrin intervention may lead to enhanced network flexibility by supporting neuromodulatory mechanisms, some of which may be enacted via the gut-brain axis [[94], [95], [96], [97]].

Blood glucose lability (as ascertained via average blood glucose spikes per day and mean blood glucose levels) did not significantly change with the intervention, despite evidence supporting the glucose-lowering effects of beta-hydroxybutyrate [[98], [99], [100]]. Despite the structural overlap between butyrate and beta-hydroxybutyrate, these findings suggest either mechanistic divergence or power limitations in this pilot study given the small sample size, as our outcomes did not replicate the glycemic benefits of sodium butyrate and inulin supplementation that were observed in a larger sample of individuals with type 2 diabetes mellitus [101]. Nonetheless, no safety concerns (e.g., severe hypoglycemic episodes) were identified with continuous glucose monitoring, indicating a reassuring safety profile.

Sleep and heart rate data obtained via Oura™ biometric ring monitoring were largely unremarkable, although deep sleep proportion increased significantly (on average, 21.9 ​% greater proportion of nightly deep sleep) as a function of treatment. Ketogenic diets are known to increase deep sleep proportion [102,103], suggesting that ketone bodies and related molecules such as butyrate may enact convergent sleep-modulating mechanisms.

In conclusion, while these initial findings are promising, they underscore the need for further investigation into the mechanistic pathways and broader clinical effects of tributyrin supplementation in PD. This will involve larger, controlled studies to validate these preliminary results and explore the potential for adjunctive therapies targeting the gut-brain axis.

Considerations for future research

While this pilot study provides the earliest in-human data evaluating the effects of tributyrin in the setting of PD, the most established body of evidence for butyrate supplementation currently falls in the domain of gastrointestinal disorders – namely, Inflammatory Bowel Disease (IBD) and Irritable Bowel Syndrome (IBS) [[104], [105], [106], [107], [108], [109], [110], [111], [112]]. Anti-inflammatory mechanisms have been postulated to play a central role in these clinical applications [69,113,114]. Although neuroinflammation has been shown to play a key role in PD [115,116], to assume that similar butyrate dosing strategies might be applicable to both neurodegenerative and inflammatory gastrointestinal disorders would be an extrapolation. Theoretically, it is feasible that higher doses of butyrate might be necessary in the setting of PD to saturate butyrate concentration in the small bowel and thereby drive relatively greater systemic circulatory uptake, promoting delivery to the brain and other target organs beyond the gut. Moreover, butyrate is postulated to enact multiple mechanisms of action beyond its anti-inflammatory signaling properties, including effects on energy metabolism and mitochondrial turnover [[117], [118], [119]], which may be of particular relevance to PD given the mitochondrial mechanisms implicated in the pathogenesis of PD [68]. Notably, the improvements in neural network flexibility observed in this sample are suggestive of improvements in mitochondrial function, as patterns of spreading activation throughout neural networks are inherently dependent on modulation of excitability at the level of individual neurons. In turn, modulation of neuronal excitability is a function of mitochondrial integrity, given the striking energetic requirements of the Na⁺/K⁺ ATPase responsible for maintaining optimal neuronal membrane potential (by some estimates, this ATPase consumes nearly half of the ATP in the brain) [120,121]. Beta-hydroxybutyrate, a molecule structurally related to butyrate (Fig. 7), has been shown to improve mitochondrial function and neural network stability [58,[122], [123], [124]]. The structural overlap between butyrate and beta-hydroxybutyrate suggests the possibility of mechanistic overlap, with the two therapies potentially conferring complementary effects in combination based on preliminary data [26], though further research is needed to assess this.

Fig. 7.

Structural comparison: butyrate and beta-hydroxybutyrate.

In addition, future research should further assess gut permeability and related effects on systemic inflammation, given that systemic inflammation may drive neuroinflammation, ultimately contributing to the pathogenesis and clinical course of PD. It is possible that improvements in gut permeability were a pivotal factor underlying the anti-inflammatory effects of tributyrin observed in this pilot study (which would be consistent with the literature pertaining to the utility of butyrate in intestinal disorders, where improvements in gut permeability were observed with butyrate administration) [[125], [126], [127]], though more targeted measures such as intestinal permeability testing or zonulin measurements will be necessary in future studies to elucidate this. Butyrate may modulate the gut-brain axis via other mechanisms as well, including modulation of neurotransmitter synthesis, astrocyte bioenergetics, and vagal tone [[128], [129], [130]]. Measures of gut permeability should be utilized alongside functional neuroimaging outcomes to assess modulating effects of the gut-brain axis in its entirety. Fecal sampling may be of additional utility to determine the quantity of butyrate-producing bacteria in the colon at baseline and post-intervention. Deficiencies in colonic butyrate-producing bacteria have been identified in the setting of PD [44,[131], [132], [133]]. Notably, a recent study by Melis et al. found that levodopa treatment is associated with a reduction in butyrate-producing bacteria in PwP – specifically, the study found a reduction in the abundance of Blautia and Lachnospirae in the levodopa-treated group compared to a group of PD subjects not on antiparkinsonian medications [134]. This suggests the possibility that the ameliorative effects of levodopa may be limited in the long term by detrimental effects on gut microbiota composition. Alternatively, this could be interpreted to imply that severity of PD symptoms (e.g., a greater degree of symptomatic burden necessitating symptomatic treatment with levodopa) may be associated with the presence vs. absence of butyrate-producing microflora. Complementary therapies targeting the gut-brain axis by augmenting butyrate-producing microflora or increasing the availability of butyrate via other means may thus offer additional benefit in the setting of PD.

Strengths

This pilot study offers the first in-human data evaluating tributyrin as a potential treatment for PD, revealing promising safety and efficacy outcomes that warrant further investigation. Strengths of this pilot study include utilization of a robust clinical test battery and state-of-the-art equipment allowing for fine characterization of cognitive and motor features, the development and first-ever application of the novel [¹¹C]butyrate PET ligand to quantify regional metabolic changes in the brain and body in conjunction with fMRI neural network stability measurements, and incorporation of peripheral biomarkers to further characterize biomechanistic changes and partially account for the subjective factors inherent to a non-placebo-controlled, open-label target engagement study. Although the sample size in this pilot study was small, several key outcomes reached both statistical and clinical significance.

Limitations

While the open-label nature of this study design facilitates feasibility and implementation, it inherently creates a vulnerability to the placebo effect that can only in part be accounted for by utilizing objective functional imaging outcomes and peripheral biomarkers. Furthermore, the small sample size utilized for feasibility purposes in the setting of unknown effect sizes appears to be limited in regard to statistical power, as there were a number of outcomes that trended toward improvement but did not reach statistical significance. Future studies can address these limitations by incorporating expanded sample sizes, placebo control groups, and blinding protocols. Finally, from an outcome measure sensitivity standpoint, performing motor examinations in the dopaminergic ‘OFF’ state improves sensitivity (i.e., allows for greater capacity to detect changes in motor features related to tributyrin supplementation) at the cost of external validity, as tributyrin supplementation would be intended to serve as a complementary rather than alternative therapy for PD. Conversely, non-motor evaluations were performed in a dopaminergic ‘ON’ state to minimize participant burden throughout the testing day. Evidence indicates that while dopaminergic medications, including levodopa, produce clear improvements in motor function, their acute effect on global cognitive scores is minimal. Minor changes may occur in specific cognitive domains – such as slight improvements in naming and abstraction and a mild worsening in delayed recall – but these are not expected to translate into significant changes in overall cognitive test scores or quality of life measures on the day of assessment [[135], [136], [137]]. Moreover, testing in the dopaminergic ‘ON’ state for non-motor measures was consistent between baseline and post-intervention assessments, and no changes in dopaminergic medications occurred for any subjects during the course of the intervention, controlling for this potential effect.

Conclusion & directions for future research

Collectively, these results are consistent with the fundamental premise of this pilot study, that cortical metabolic deficits reflect synaptic hypofunction secondary to energy deficits that can be partially remediated by alternative energy substrate provision. In addition, given topographic brain correlations and marked reductions in systemic inflammation, some effects could also be explained by the signaling properties of butyrate, anti-inflammatory mechanisms, and modulation of the gut-brain axis. These preliminary findings justify further investigation of tributyrin as a potential therapy in PD via more highly powered, placebo-controlled phase 2 clinical trials.

Directions for future research.

• 1.Future studies should expand on our use of functional neuroimaging correlates, such as [¹¹C]butyrate PET and fMRI, in more highly powered, placebo-controlled phase 2 clinical trials.
• 2.Measures of gut permeability should be utilized alongside functional neuroimaging outcomes to assess modulating effects of tributyrin on the gut-brain axis in its entirety.
• 3.Future studies should also investigate more clinically accessible peripheral biomarkers, such as hs-CRP or zonulin, which may provide opportunities to help guide clinical care or predict treatment response with a minimally invasive approach.
• 4.Complementary therapies targeting the gut-brain axis by augmenting butyrate-producing microflora or increasing the availability of butyrate via other means may help offset detrimental effects of levodopa on gut microbiota composition in the setting of PD.

Author contributions

Jeffrey L.B. Bohnen: conceptualization, literature review, study design, data collection, clinical assessments, data interpretation, writing, figures, review & revision. Stiven Roytman: conceptualization, literature review, study design, functional neuroimaging analyses, data interpretation, writing, figures, review & revision. Travis P. Wigstrom: conceptualization, data interpretation, writing, review & revision. Robert K. Vangel: conceptualization, data collection, clinical assessments, data interpretation, writing, review & revision. Jaime E. Barr: conceptualization, data collection, clinical assessments, data interpretation, writing, review & revision. Giulia Carli: conceptualization, functional neuroimaging analyses, data interpretation, writing, figures, review & revision. Sean M. Parks: conceptualization, writing, review & revision. Hitasha Mittal: conceptualization, data interpretation, literature review, writing, review & revision. Claire Martino: review & revision. Prabesh Kanel: conceptualization, study design, functional neuroimaging analyses, data interpretation, writing, figures, review & revision. Roger L. Albin: conceptualization, literature review, study design, data interpretation, writing, figures, review & revision, funding acquisition, study oversight. Nicolaas I. Bohnen: conceptualization, literature review, study design, data collection, clinical assessments, functional neuroimaging analyses, data interpretation, writing, figures, review & revision, funding acquisition, study oversight.

Declaration of competing interests

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Jeffrey L.B. Bohnen reports financial support and article publishing charges were provided by the Farmer Family Foundation. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article: