Abstract
Background:
Metabolic abnormalities have long been predicted in Huntington’s disease (HD) but remain poorly characterized. Chronobiological dysregulation has been described in HD and may include abnormalities in circadian-driven metabolism.
Objective:
Here we investigated metabolite profiles in the transgenic sheep model of HD (OVT73) at presymptomatic ages. Our goal was to understand changes to the metabolome as well as potential metabolite rhythm changes associated with HD.
Methods:
We used targeted liquid chromatography mass spectrometry (LC-MS) metabolomics to analyze metabolites in plasma samples taken from female HD transgenic and normal (control) sheep aged 5 and 7 years. Samples were taken hourly across a 27-h period. The resulting dataset was investigated by machine learning and chronobiological analysis.
Results:
The metabolic profiles of HD and control sheep were separable by machine learning at both ages. We found both absolute and rhythmic differences in metabolites in HD compared to control sheep at 5 years of age. An increase in both the number of disturbed metabolites and the magnitude of change of acrophase (the time at which the rhythms peak) was seen in samples from 7-year-old HD compared to control sheep. There were striking similarities between the dysregulated metabolites identified in HD sheep and human patients (notably of phosphatidylcholines, amino acids, urea, and threonine).
Conclusion:
This work provides the first integrated analysis of changes in metabolism and circadian rhythmicity of metabolites in a large animal model of presymptomatic HD.
Keywords
INTRODUCTION
Huntington’s disease (HD) is an inherited neurodegenerative disease characterized by a wide range of symptoms including motor dysfunction and cognitive decline, as well as psychiatric problems such as depression [1]. Patients with HD, in common with patients with other neurodegenerative diseases, often exhibit disturbed circadian rhythms and sleep disruption [2–6]. In HD, the central hypothalamic pacemaker located in the suprachiasmatic nuclei (SCN) has been implicated in this disturbance [7], but the precise nature of chronobiological dysregulation and the timing of onset in HD remains unclear.
With regards to circadian rhythms and sleep in HD, loss of form and definition in the rest-activity profiles in patients have been identified, as have abnormal night-day ratios [6]. As well as changes to sleep and other circadian controlled behaviors, alterations in metabolism have been described in HD rodent models. Metabolic dysregulation, however, has been more difficult to study in patients. Whilst energy imbalances have long been considered a possibility in HD because symptomatic patients suffer from cachexia of unknown origin [1], clear metabolic deficits have been elusive. Nevertheless, a number of metabolic changes have been identified in HD patients. These include reduced levels of circulating growth factors, abnormal cholesterol metabolism, and elevated levels of C-reactive protein [8–15]. Furthermore, a targeted metabolomics study with analysis of serum by liquid chromatography mass spectrometry (LC-MS) identified deregulation of phosphatidylcholine (PC) metabolism and selected amino acids as characteristic of HD, both versus healthy controls and in progressive severity of HD [16].
While most transgenic research is carried out on rodents, the short lifespan of mice and rats is a disadvantage when investigating the progression of late-onset neurodegenerative diseases such as HD. Sheep (Ovis aries) offer a long-lived alternative model species [17]. A transgenic sheep model of HD has been created [18] that shows HD-relevant brain pathology [19] and measurable phenotype [8, 20–23] but to date no overt symptoms. This model is ideally positioned to study early stages of HD in a time-frame relevant to the human condition. Interestingly, this model shows early circadian behavioral changes [20], as well as sleep and electroencephalography (EEG) abnormalities [21]. Regarding sleep specifically, EEG examination found significant differences between HD transgenic (hereafter called HD) and normal (control) sheep in EEG power and its pattern of distribution during non-rapid eye movement (NREM) sleep [22]. Metabolomic analysis of brain and liver found significant and sex specific changes in amino acids in HD sheep [8], as well as disturbances in plasma melatonin and cortisol levels [23] and dysregulation of metabolism [24]. This large animal model of HD therefore represents an excellent diurnal species in which to study metabolic changes relevant to HD.
In this study, we aimed first to assess whether the metabolic profiles of control and HD sheep could be separated by machine learning, analyzing samples taken at 5 and 7 years of age. These time points were late enough to allow for biological progression of the disease, given that there are already observed brain pathologies [8], but early enough to be considered presymptomatic, since even at 10 years of age there are no overt signs of disease. We found significant metabolic differences between HD and control sheep. We then assessed chronobiological dysregulation in the HD versus control sheep by examining circadian rhythmicity of a wide range of metabolites in both age cohorts. We found that the changes in acrophase (the time at which the rhythms peak) seen in HD sheep compared to controls became progressively larger; the number of dysregulated metabolites also increased with age.
MATERIALS AND METHODS
Animals
We used age-matched HD and control female sheep (ewes) from the OVT73 line that carries a full-length human transgene with a mutation (a CAG repeat of 73) that in humans would typically cause juvenile-onset HD. Sheep were aged 5 years (14 HD, 13 control) or 7 years (16 HD, 14 control). The 7-year-old group included 8 of the control and 10 of the HD sheep that had previously been sampled at 5 years of age. The sheep were obtained from a mixed genotype (HD and control) flock maintained outdoors at the South Australian Research and Development Institute (SARDI) in accordance with OGTR approval NLRD-1037/2003 and NLRD-7574/2015. None of the sheep showed overt behavioral abnormalities of any kind and thus are considered to be at a ‘presymptomatic’ stage.
Animal husbandry
Animal husbandry and blood sampling was performed in accordance with AEC approvals SAM93 (SAHMRI) and 07/14 (PIRSA). For circadian blood sampling, sheep were transported from SARDI to the Preclinical Imaging and Research Laboratory (PIRL) at Gilles Plains, South Australia. To ensure minimal dietary change throughout the period prior to blood sampling, for at least 4 weeks prior to transport and following relocation to PIRL, sheep were fed a defined ration of food. At PIRL, the sheep were transferred to indoor holding pens for a week of acclimatization.
For blood sampling, the sheep were housed in individual indoor metabolic crates. During indoor acclimatization, the light/dark cycle matched the outdoor light/dark cycle. On the day of cannulation, the lights were turned off in the evening (at sunset) and remained off until the end of the blood collection period 45 h later (see below). All samples were collected in dim light using red head torches.
Sample collection and processing
Blood collection commenced at 13:00 h the day following placement of the cannula, as previously described [24]. Blood samples were collected on the hour (±5 min) for 27 h. An appropriate dead volume of blood was discarded, then 10 mL samples of blood were withdrawn into a fresh syringe and transferred to lithium heparin tubes. Samples were centrifuged immediately at 1620 g for 10 min at 4°C. Plasma was aliquoted and stored at –20°C (for melatonin/cortisol/urea) and –80°C (for metabolomics), prior to being shipped on dry ice to the UK for analysis. Melatonin and cortisol concentrations were measured by radioimmunoassay as previously described [25]. The limit of detection for the plasma melatonin assay was 4.4±2.2 pg/mL (5-year-old sheep) and 4.6±2.0 pg/mL (7-year-old sheep). The limit of detection for the plasma cortisol assay was 1.3±0.7 nmol/L (5-year-old sheep) and 1.0±0.5 nmol/L (7-year-old sheep). For both assays, the inter-assay coefficients of variation (CVs) of low, medium, high, and very high QCs were < 15% (range 6.1% –14.7%). One 7-year-old control sheep had abnormally high (2-3-fold higher than normal) melatonin levels and was excluded from further analysis. Urea nitrogen measurements were also performed on hourly plasma samples across 27 h using the ADVIA® 1800 Chemistry System (Siemens Healthcare Diagnostics Inc., Camberley, UK), method based on the Roch-Ramel enzymatic reaction using urease and glutamate dehydrogenase. The inter-assay CVs for the 5-year-old sheep for the plasma urea nitrogen assay were 4.5% at 3.0 mmol/L, 1.5% at 10.2 mmol/L, and 1.5% at 19.4 mmol/L (n = 6 at each concentration). The equivalent inter-assay CVs for the 7-year-old sheep were 4.6% at 5.7 mmol/L and 2.2% at 25.3 mmol/L (n = 6 at each concentration).
A set of plasma metabolites (≈130) was separately measured using the AbsoluteIDQ® p180 targeted metabolomics kit (Biocrates Life Sciences AG, Innsbruck, Austria), and a Waters Xevo TQ-S mass spectrometer coupled to an Acquity UPLC system (Waters Corporation, Milford, MA, USA) as previously described; some metabolites were excluded from analysis using our standard exclusion criteria [24, 26]. One control sheep that was an outlier in a PCA plot of the metabolite data was confirmed to be ill during sampling and died soon after and was thus excluded from any subsequent analysis. The two cohorts (5-year-old and 7-year-old) were sampled in February two years apart, in 2015 and 2017 respectively. Plasma samples were stored at –80°C until LC-MS metabolomics analysis 7–13 months later. Investigation by PCA of possible batch effects, i.e., changes in metabolite identifications or concentrations caused by processing samples in different batches under different conditions, revealed a clear unsupervised separation of the two groups due to changes in the manufacturer’s platform. While this does not influence our analysis of ‘within’-cohort separation (HD versus control sheep), because of this we could not compare the two ages directly.
Data analysis
The sampling protocol generated two data arrays (HD and control) per age cohort, with metabolite concentrations tabulated by sheep and by time point. To control for inter-animal differences in circadian phase, all hormone and metabolite time point data were aligned to each animal’s dim light melatonin onset time (DLMO) that is considered to be a reliable marker of circadian SCN-phase. The onset of melatonin (DLMO) was calculated using the 25% threshold method as described by Sletten et al. [27].
As an initial step in assessing whether there were measurable metabolic differences between the HD and control sheep, multivariate analysis of the datasets was performed by constructing an orthogonal partial-least squares-discriminant analysis (OPLS-DA) model for each cohort (5- and 7-year-old sheep). OPLS-DA is a supervised multivariate technique that reduces high-dimensional data into a smaller number of orthogonal components which can be used to represent the full dataset [28]. In the case of metabolomics, many thousands of features can be reduced into a small number of components, which can then be used to classify and make predictions about the status of participants. This was done using MetaboAnalyst [29], using the two age groups as data frames; both data frames were log-transformed and pareto-scaled before analysis. Feature selection was conducted as a prior step using logistic regression to identify the most important variables and avoid overfitting ‘noise’ variables. For the models obtained, the results were assessed by the R2Y metric (to measure explained variation). High R2Y scores may be due to overfitting, so the Q2Y metric was also used to assess the predictive ability of the model. A Q2Y > 0.5 is regarded as good and a Q2Y > 0.9 as excellent, while differences between R2Y and Q2Y larger than 0.2–0.3 indicate the presence of many irrelevant model terms or a few outlying data points [30].
Variable Importance in Projection (VIP) scores for each model were also extracted to highlight which metabolites were responsible for separating HD from control sheep in each cohort. The VIP feature scores are based on the amount of variance between the positive and negative participants that is explained by each metabolite across the components. To all intents and purposes, VIP scores reflect the relative importance of a metabolite in classifying participants as cases or controls.
To assess differences between the HD and control sheep for each cohort, rhythm analysis algorithms were employed to detect circadian ( 24 h) oscillating periods (rhythms) and acrophase (rhythm peak) in metabolite concentrations. The two aforementioned data arrays, post DLMO correction and Z-scoring, were independently analyzed by six rhythm detection algorithms available as R packages (R – Version 4.0.5, Cosinor – ‘DiscoRhythm’ (Version 1.6.0), [31] JTK – ‘MetaCycle’ (Version 1.2.0), [32] ARS – ‘MetaCycle’ (Version 1.2.0), ECHO – ‘echo.find’ (Version 4.0.1), [33] RAIN – ‘rain’ (Version 1.24.0), [34] and BioCycle – (Version 0.9.3) [35].
Results from all six algorithms were cross validated and individual metabolites were classified as ‘rhythmic’ if q < 0.05 (post Benjamini-Hochberg FDR correction) in≥3 algorithms. Rhythmic metabolites were ordered by ascending acrophase (as determined by Cosinor analysis) and then plotted as heatmaps to visualize and confirm the presence of 24 h rhythms. Heatmaps were produced via MetaboAnalyst with no additional filtering or data transformations save for group averaging data per time point. The same dataset was used to compare acrophase differences between the control and HD groups for each cohort.
RESULTS
Metabolite concentrations and separation analysis
The combination of the Biocrates targeted metabolomics platform and the additional hormone assays (melatonin, cortisol, urea) yielded an overall feature set of 136 and 130 metabolites and hormones for the 5- and 7-year-old sheep, respectively. As an initial step in identifying whether the overall metabolic profiles of the HD and control sheep were separable, logistic regression with recursive feature elimination and cross validation was performed. This allowed identification of the most statistically useful metabolites and removed noise, thus reducing overfitting of the model. For both age cohorts, the 20 most significant metabolites were identified in this way and OPLS-DA models for both age cohorts were then constructed using these reduced feature sets (Fig. 2A, B).
Graphical Abstract. Assessing chronobiological dysregulation of metabolism in HD sheep by targeted liquid chromatography-mass spectrometry metabolomics. Created with BioRender.com.
Separation of HD and control sheep by multivariate analysis. OPLS-DA multivariate separation of 5-year-old HD (n = 14) and control (n = 13) sheep shown in A. Score in arbitrary units on each axis represents the amount of variation in the dataset captured by each component. Ellipses represent the 95% confidence interval within which the observations are expected to cluster. OPLS-DA multivariate separation of 7-year-old HD (n = 16) and control (n = 14) sheep shown in B. Score in arbitrary units on each axis represents the amount of variation in the dataset captured by each component. Ellipses represent the 95% confidence interval within which the observations are expected to cluster.
For the 5-year-old sheep, good separation was achieved (Fig. 2A). R2Y was 0.82 and Q2Y was 0.68, indicative of good fit. The 20 metabolites used in the OPLS-DA model together with their VIP scores are listed in Table 1.
Metabolites contributing to the OPLS-DA separation of HD and control sheep at 5 years old ordered by VIP score
For the 7-year-old sheep, good separation was again achieved (Fig. 2B). R2Y was 0.76 and Q2Y was 0.55, again indicative of good fit. The OPLS-DA separation result of the 7-year-old model is summarized in Fig. 3. The 20 metabolites used in the OPLS-DA model together with their VIP scores are listed in Table 2.
Comparison of rhythmic metabolites in control sheep to the same metabolites in HD sheep. Rhythmic metabolites shown as heatmaps displaying Z-scored and group averaged changes in metabolite concentrations (red = high, blue = low) relative to dim light melatonin onset (DLMO) in 5- (control group, A; HD group, B) or 7- (control group, C; HD group, D) year-old sheep. DLMO measurements were as follows: for 5-year-old control sheep (n = 13), DLMO = 17.6±4.5 h (A); 5-year-old HD sheep (n = 14), DLMO = 18.8±2.6 h (B); 7-year-old control sheep (n = 14), DLMO = 19.4±0.6 h (C); 7-year-old, HD Sheep (n = 16); DLMO 20.7±0.7 h (D). Brackets in C, D indicate metabolites with rhythm patterns that are distinctively different in control and HD sheep. Arrowheads in C, D indicate the approximate peak times of the rhythms.
Metabolites contributing to the OPLS-DA separation of HD and control sheep at 7 years old ordered by VIP score
In the 5-year-old cohort, the most significant metabolites for differentiating between control and HD sheep were PC aa C42 : 6, PC aa 34 : 1, methionine, C5, PC ae C32 : 1, PC ae C30 : 0, and PC ae 34 : 1 (measured by VIP score≥1.0). In the 7-year-old cohort, the most significant metabolites differentiating between control and HD sheep were PC aa C40 : 3, threonine, PC aa C40 : 4, PC aa C38 : 3, SM (OH) C22 : 2, urea, PC aa C42 : 6, and lysoPC a C20 : 4 (measured by VIP score≥1.0).
Metabolite rhythms
Given that the OPLS-DA models showed marked metabolic alterations between the HD and control sheep, chronobiological differences were subsequently investigated. As a first step, rhythmic metabolites were identified in both age groups. In the 5-year-old group, 43 and 38 metabolites were rhythmic in the control and HD group, respectively, of which 26 were commonly rhythmic between the groups. In the 7-year-old group, 67 and 35 metabolites were rhythmic in the control and HD group, respectively, of which 24 were commonly rhythmic between groups. Melatonin was rhythmic in both groups, but cortisol and urea were not (time plots for these three compounds which were analyzed independently of the Biocrates platform are provided in Supplementary Figure 2). Note that because of the observed batch effect, the number and rhythmicity of metabolites in each age cohort are not directly comparable.
Rhythmic metabolites from each control group were visualized as heatmaps and compared against their equivalent age HD groups (Fig. 3A–D). The 5-year-old year control and HD sheep present a similar profile (Fig. 3A, B, respectively), although for specific lipids, such as the sphingomyelins and some phospholipids, metabolite peaks (dark red, Fig. 3A, B) appear wider and less well-defined in HD sheep (Fig. 3B), resulting in a narrower and less-well defined nadir (dark blue;Fig. 3B).
The differences between control and HD sheep were more apparent at 7 years of age. Notably, there were marked differences in rhythm period between genotype groups. Phospholipids peaks were seen at approximately – 4 h and+18– 20 h relative to DLMO for control sheep (indicated by ▽ in Fig. 3C), resulting in a near 24 h rhythm (Fig. 3C). By contrast, in the HD sheep phospholipids either had an ill-defined peak or they peaked later, at approximately -4h, -1h and +3 h and again at +18– 20 h relative to DLMO (indicated by ▾ in Fig. 3D). This may be the reason why many of these phospholipids were not classified as ‘rhythmic’ for a 24 h period in the 7-year-old HD sheep. These circadian changes were observed despite there being no significant difference in the timing of melatonin ‘phase’ as measured by DLMO between control and HD sheep at either age.
Metabolite concentrations were z-scored and DLMO-corrected prior to rhythm analysis. To visualize phase differences between the control and HD group, we plotted the acrophase time (with respect to DLMO) of the rhythmic metabolites for the 5-year-old (n = 55; Supplementary Figure 1) and 7-year-old (n = 78; Fig. 4) sheep. Those metabolites where the acrophase differed by more than two and four hours between controls and their respective HD group are shown in yellow and red respectively; for a full list see Table 3. A full list of the rhythmic metabolites in control sheep only, in HD sheep only and in both control and HD sheep for the 5-year-old groups is presented in Supplementary Table 1.
Comparison of metabolite acrophases in 7-year-old HD and control sheep. Acrophase (peak) time of metabolites rhythmic in both control and HD groups (n = 24, main panel), rhythmic in control only (n = 43, top) and rhythmic in HD only (n = 11, right) with respect to DLMO for 7-year-old sheep. The solid black line indicates where metabolites would fall if the difference in time of acrophase (phase difference) between groups = 0 h. Phase differences > 2 h and > 4 h are shown in yellow and red, respectively. A full list of the rhythmic metabolites in controls only, in HD only and in both controls and HD for the 7-year-old sheep is presented in Table 3.
Rhythmic metabolites in 7-year-old HD and control sheep
DISCUSSION
Using targeted LC-MS based metabolomics, significant metabolic alterations were identified in HD compared to the flock control sheep. In both the 5-year-old and 7-year-old cohorts, the significant features for separation (measured by VIP score≥1.0) were dominated by PCs, with threonine and urea additionally significant in the 7-year cohort. These metabolites were identified by a fully validated targeted metabolomics platform that is reproducible and selective, with low relative standard deviations, thanks to use of internal standards and calibration curves that are not available in untargeted analyses. This provides confidence that the metabolites identified by this method are real and that the data generated can be replicated in other facilities [36]. As the assay uses orthogonal identifying information (accurate m/z mass-to-charge ratio, isotopically labelled standards and retention time), the metabolites can be considered fully identified in accordance with the Metabolomics Standards Initiative [37].
The metabolites identified as driving separation of HD from control sheep include a large proportion of phosphatidylcholines (PCs). Notably these have previously been identified as being involved in HD pathology in humans. Two metabolomic studies observed downregulation of PC aa 36 : 0 in HD symptomatic and presymptomatic patients, with one of these studies also identifying dysregulation of PC C38 : 6, PC ae C38 : 0, PC aa C38 : 0, PC ae C38 : 6, PC ae C42 : 0, PC aa C36 : 5, and PC ae C36 : 0 [16, 38]. Overlap with specific PCs identified in the sheep model was limited, but the dominance of PCs suggests that, at least in part, human pathways of dysregulation are mirrored in the HD sheep. PCs were the dominant differentiating class for both the 5-year-old and 7-year-old female sheep used here. Interestingly, the differentiating features identified showed limited overlap with those seen in HD rams [24], consistent with previously described sex differences in the manifestation of HD in the sheep model [8].
The differences between HD and control sheep across the two age groups were also evident in the circadian rhythm analysis of the two cohorts. While in the 5-year-old sheep the number of rhythmic metabolites and their respective profiles were similar, the same did not hold true for the 7-year-old sheep. Several metabolites in the HD group showed distinct alterations in normal rhythms at DLMO+3 h, which corresponds to previous observations of disturbed evening behavior in HD sheep [19], similar to the ‘sundowning’ effect seen in some patients with dementia [39].
Increased melatonin levels in blood samples from HD sheep have been described previously [23]. Notably the circadian changes described in the 7-year-old sheep were seen despite there being no significant difference in melatonin ‘phase’ between control and HD sheep at either age. This suggests that the observed differences in circadian phase described above are not directly regulated by the SCN, but rather may be regulated by aberrant behavioral cycles and/or timing of peripheral clocks. The liver is the most likely driver of this, due to its central role in lipid metabolism. It is interesting to note that significant end stage hepatic atrophy is present in HD postmortem [40, 41]. Although there is little evidence for frank hepatic disease in HD, there is evidence of circadian hepatic dysfunction in HD mouse models [42] and subclinical hepatic dysfunction in HD patients by presymptomatic stages [43].
Glycerophospholipids, specifically PCs and lysophosphatidylcholine (LPC), are the most prominent metabolite classes displaying differences in rhythmicity within our data. The biological significance of these differences is not known. A potential metabolic process of interest, however, is the Land’s cycle, a process which regulates interconversions between phospholipids and lysophospholipids through deacetylation/acetylation and occurs jointly within the liver and the blood stream. Interestingly, temporal activity for specific enzymes that regulate the Land’s cycle have been observed in vitro such as LPC acyltransferase (LPCAT) which regenerates PCs from LPCs in the presence of acyl-CoA [44]. Moreover, phospholipaseA2 that is responsible for converting PCs to LPCs in the Land’s cycle has been shown to have increased activity in leukocytes in HD patients and may partly explain why serum concentrations of specific PCs decrease in conjunction with HD progression [16, 38]. Temporal regulation is a mechanism that can separate incompatible metabolic processes and is potentially required to some degree for optimal regulation of the Land’s cycle and associated outcomes such as clearing excess LPCs from regulation and regenerating PCs [45]. It can be inferred from our data that there is some degree of temporal dysregulation in the HD sheep although its source must remain a matter for speculation. Should phospholipaseA2 activity be increased in HD patients, as previously observed, and/or temporal dysregulation of other components of the Land’s cycle occurs, a potential consequence may be diminished circulating PCs and excessive circulating LPCs. Circulating LPCs have numerous roles and effects on different cell types, such as promoting inflammation and inducing demyelination both of which are associated with HD pathophysiology [38, 46]. Moving forward, the Land’s cycle, and its potential temporal regulation of phospholipids, should be considered a metabolic pathway of interest in relation to HD progression. (For discussion of alternative hypotheses on the dysregulation of lipid and PC metabolism see [47]).
Two metabolites dominated the separation of HD and control groups at 7 years, urea and threonine. Interestingly, both have been shown previously to be altered in HD. Urea is increased in both human and sheep brain [48, 49], and in HD sheep serum [24]. Threonine is an essential amino acid that can be converted to pyruvate and can undergo thiolysis to produce acetyl-coA. It has been suggested that pyruvate is neuroprotective [50] and has been shown to be so in a quinolinic acid rat model of HD [51]. Increased threonine concentrations in serum from the 7-year-old HD sheep agrees with a previous study showing increased levels of this essential amino acid in serum from human HD patients [16]. Those authors suggested that this could represent a compensatory mechanism as an attempt to produce more substrates to generate pyruvate. This is an interesting avenue of study that could be pursued further in the HD sheep.
It should be noted that in this work only plasma was investigated, but peripheral clocks are active in most tissues, including many involved in metabolism (liver, adipose tissue, muscle, pancreas, and gut); any of these organs could well contribute to the observed metabolite changes, as could exogenous factors such as gut microbiota (for example as discussed recently by Kong et al.) [52]. Furthermore, only ewes were analyzed for this study. We would expect to find generally dysregulated chronobiology if male sheep were to be studied, albeit we would also expect them to be sexually dimorphic, and this would be an interesting area for future research. Nevertheless, the integration of separation analysis based on concentrations/amplitude and high resolution rhythm analysis across the 24 h day for the ewe data analyzed here provides a more comprehensive picture than can be achieved by examining separation and chronobiology independently. For example, average amplitude of urea plasma concentrations shows a significant change between HD and control sheep in the 7-year-old cohort, but the urea rhythm does not exhibit a shift in acrophase. The rhythm analysis presented here provides a first comprehensive link between metabolic changes and circadian dysregulation in HD sheep. Together our data strongly support the idea that a plethora of metabolic changes and alterations in the timing of metabolite rhythms occur early (and presymptomatically) in HD, and that this mismatch may contribute to the progression of symptoms. We show that metabolite circadian rhythm changes are progressive even in presymptomatic sheep, with marked increases in the number and magnitude of changes in acrophase between the ages of 5 to 7 years. We also find similarities between the metabolites identified in the sheep model and the human model striking, notably in the dysregulation of PCs, urea and threonine.
Overall, our study provides direct evidence for disruption of two important— and linked— systems that are necessary for optimal physiological function, namely metabolism and circadian rhythms. There are numerous publications providing evidence suggesting that metabolism is dysregulated in HD (see introduction for references), although it has been challenging to study metabolomics in HD patients. We and others have also shown previously that not only are there disruptions to circadian behavior, but also circadian regulation of gene expression. Indeed, there is growing evidence that circadian disruption in HD is widespread. Most recently a study mapping brain gene coexpression in daytime transcriptomes revealed diurnal molecular networks that showed perturbations in gene signatures in HD mice [53]. The mechanisms underlying the changes we see in the rhythms of metabolites in HD are unknown. Nevertheless, while untangling the precise role of these disturbances in HD is a challenge for future studies, our study shows that the HD sheep provided an excellent and controllable model system in which to study presymptomatic changes in HD.
Footnotes
ACKNOWLEDGMENTS
We thank the team at SAHMRI for animal care and collection of the plasma samples, in particular C. Fraser, S. Porter, and S. Moore. We thank the team at SARDI for animal care and translocation of the sheep, in particular S. Bawden and S. Rudiger. The authors also thank Dr. J. Pennings (National Institute for Public Health and the Environment, Bilthoven, the Netherlands) for statistical assistance and advice. The graphical abstract was made using BioRender (Biorender.com).
This work was funded by CHDI Inc (AJM) and supported in part by a UK Biotechnology and Biological Sciences Research Council (BBSRC) Grant BB/I019405/1 (DJS).
DATA AVAILABILITY
The analytical protocols used as well as mass spectrometry.raw files, sample and participant data will be available on request.
CONFLICT OF INTEREST
DJS and BM are past co-directors of Stockgrand Ltd. The other authors have no interests to declare.
