First-episode psychosis (FEP) is a psychiatric disorder, characterised by positive and negative symptoms, usually emerging during adolescence and early adulthood. FEP represents an early intervention opportunity for intervention in psychosis. Redox disturbance and subsequent oxidative stress have been linked to the pathophysiology of FEP. Exercise training can perturb oxidative stress and rebalance the antioxidant system and thus represents an intervention with the potential to interact with a mechanism of disease. The aim of this study was to assess the effect of exercise on markers of redox status in FEP. Twenty-two young men were recruited from Birmingham Early Intervention services and randomised to either a 12-week exercise programme or treatment as usual (control). Measures of blood and brain glutathione (GSH), markers of oxidative damage, inflammation, neuronal health, symptomology and habitual physical activity were assessed. Exercise training was protective against changes related to continued psychosis. Symptomatically, those in the exercise group showed reductions in positive and general psychopathology, and stable negative symptoms (compared to increased negative symptoms in the control group). Peripheral GSH was increased by 5.6% in the exercise group, compared to a significant decrease (24.4%) (p = 0.04) in the control group. Exercise attenuated negative changes in markers of neuronal function (brain-derived neurotrophic factor), lipid damage (thiobarbituric acid-reactive substances) and total antioxidant capacity. C-reactive protein and tumour necrosis factor-α also decreased in the exercise group, although protein and DNA oxidation were unchanged. Moderate-intensity exercise training has the ability to elicit changes in markers of oxidative stress and antioxidant concentration, with subsequent improvements in symptoms of psychosis.
First-episode psychosis (FEP) is a psychiatric disorder, characterised by positive and negative symptoms, usually emerging during adolescence and early adulthood1. It is well described that a delay in FEP treatment can lead to cognitive alterations and worsened functional outcome2. FEP patients are highly heterogeneous in symptom profile, treatment response and cannabis consumption, and there is a need for improved characterisation of biomarkers related to disease pathophysiology in order to better understand pathology of disease. Many of the biochemical perturbations observed in this population are associated with the psychosis phenotype and thus represent targets for treatment.
Several aspects of (neuro) physiology are altered in FEP and schizophrenia patients, including dysfunctional neurotransmitter systems3,4, decreased synaptic plasticity5 and reduced hippocampal volume6. In addition, inflammation and cellular redox status is also altered. This is often termed oxidative stress, and represents an imbalance between reactive radicals or oxidants, which are naturally produced in aerobic metabolism, and antioxidants, which balance and quench oxidants to allow essential signalling processes, but prevent damage to biomolecules leading to cell injury or death7. Brain-derived neurotrophic factor (BDNF), which has an essential role in neural signal transmission and synaptic plasticity, is depleted in FEP8. It has also been suggested that BDNF is responsible for protecting against oxidative stress by upregulating the expression of antioxidant enzymes9, highlighting the value of BDNF as a candidate biomarker of psychosis.
Glutathione (GSH) is the most abundant antioxidant that acts as a scavenger of reactive oxygen species, with a primary role of maintaining the intracellular redox balance10. GSH has been implicated in the pathogenesis of a range of neurodegenerative and psychotic diseases11. FEP patients have shown as much as a 52% reduction in GSH compared with controls12,13. Indeed, evidence of perturbed redox homeostasis in FEP and schizophrenia is plentiful. There have been reports of reduced total GSH13,14,15, increased oxidised GSH (GSSG)15, increased lipid16,17, protein18,19 and DNA damage20,21. However, conflicting data by Reddy et al.22 also showed no impairment of antioxidant defences in comparison with a healthy control group, hypothesising that disease stage is related to progression of oxidative stress accumulation. A study by Fraguas et al.14 assessed the relationship between grey matter volume and GSH in the brains of schizophrenia patients. The study highlighted a progressive blood GSH decline over the 2-year follow-up period, as well as a relationship between grey matter loss and blood GSH.
The brain is a particularly vulnerable target for free-radical-mediated damage due to a high level of oxygen consumption23,24, and high lipid content19. When this high oxygen concentration is coupled with a modest endogenous antioxidant concentration25, and inability of GSH to cross the blood–brain barrier (BBB), particularly from exogenous supplementation or circulating pools, the antioxidant defence system is stretched. Assessment of brain redox state in vivo is not easy, and peripheral markers are usually employed as an estimate of whole-body oxidative environment and antioxidant concentration. However, recent methodological advances in magnetic resonance spectroscopy (MRS) have allowed the quantification of GSH in the cerebral tissue26.
Exercise has been identified as an adjunctive treatment for psychosis as it has the ability to improve clinical symptoms27 and has the potential to restore redox homeostasis, as demonstrated in healthy populations28. However, no study has proven the efficacy of exercise in altering the underlying pathology of disease. Studies employing moderate-intensity aerobic exercise29 and combined aerobic and resistance training30 across 16 and 12 weeks, respectively, have resulted in reduced psychotic symptoms assessed by the Positive and Negative Symptom Scale (PANSS). Acil et al.31 described a 14% increase in quality of life, as well as a decrease of positive and negative symptoms that define the schizophrenic phenotype.
Mechanisms underpinning exercise adaptation are well established in studies of healthy participants and are characterised by an antioxidant response to radical signals32. Elokda and Nielsen33 demonstrated adaptations to either aerobic training or combined aerobic and circuit weight training (standardised workload). Both training programmes led to a significant increase in GSH, and reduced oxidised GSH (GSSG). Additionally, exercise training has the potential to modify other indices of redox status. Studies in health volunteers have shown that exercise training may elevate BDNF34 and reduce oxidative damage markers, including malondialdehyde (lipid peroxidation) and protein carbonyl concentration35.
Furthermore, exercise training is beneficial in reducing circulating proinflammatory cytokines36. There is robust evidence to implicate a systemic proinflammatory state in FEP. Inflammation and oxidative stress are strongly linked in a number of pathologies. Systematic review37 confirms elevated proinflammatory cytokines in drug-naive FEP, including interleukin-6 (IL-6), tumour necrosis factor-alpha (TNF-α), IL-1β and sIL-2R, and there is clear evidence of elevated IL-6 in childhood, predicting risk for both psychosis and metabolic dysfunction. Regular physical activity can lead to reductions in circulating IL-638,39, TNF-α40 and C-reactive protein (CRP)41,42.
The aim of this exploratory pilot study was to assess the effect of 12 weeks of exercise training on the GSH system in FEP, via in vivo brain and peripheral blood measures. To add context around changes in GSH, measures of oxidative damage, inflammation and neuronal health were assessed, as well as symptoms of psychosis.
Male patients, aged 16–35 years, with a diagnosis of FEP (as identified by a psychiatrist in keeping with ICD-10 F 20-29, F31.2, 32.3 diagnostic criteria), were recruited from the community-based Birmingham Early Intervention service. Patients were within 3 years of first presentation of illness. Male patients only were recruited, since the oestrogen cycle has a significant effect on antioxidant concentration, in particular GSH43. Eligibility criteria were assessed initially by the primary care coordinators for each patient, followed by assessment of habitual activity to ensure a sedentary lifestyle, as well as the use of a general health questionnaire to assess cannabis use and confirm the patient was free from medical conditions that prevent participation in moderate-intensity aerobic exercise. A sample size of 28 was calculated, using the G*Power software44, based on intervention-associated change in GSH. Exclusion criteria included failure to adhere to pretesting requirements, for example, provide a blood sample, or significant risk to self of others as identified by the clinical team. Study assessments took place at baseline, mid-point (for the exercise group only) and post intervention. This study was commenced following approval from the NIHR HRA ethics committee (West Midlands- Edgbaston REC 17/WM/0412). Intervention design, quality and patient-oriented outcomes were assessed and are summarised by Fisher et al.45.
Following consent, participants were randomised to either the exercise intervention group or the control arm (treatment as usual) of the study. A block randomisation method (http://www.randomization.com)46,47 was used to allow for equal group distribution in the event of poor recruitment.
Exercise sessions were designed and supervised by a trained researcher at the School of Sport, Exercise and Rehabilitation Sciences, University of Birmingham. The intervention was 12 weeks long, with each participant required to exercise at least 2 times per week, for 40–60 min per session. In an effort to maximise attendance and compliance, which has historically been difficult in this group48, participants were given a choice of different activities to undertake at each session (available exercise: running; cycling; swimming; tennis; squash; badminton; circuit training and football). Each training session was standardised by heart rate target zone, based on 70–80% HRmax (maximum heart rate). The minimum training intensity for improvement in aerobic fitness is 55–65% HRmax49; therefore, in order to observe a meaningful and significant effect of exercise, intensity was set above this. The ACSM (American College of Sports Medicine) recommends 50% VO2max (maximum rate of oxygen consumption) as a minimum intensity for exercise training50, which corresponds to 65% HRmax. The relationship between HR and VO2 reflect energy expenditure in a linear fashion, up to 85% HRmax51. Energy consumption, resting heart rate, active calories and intensity minutes (equivalent to moderate-intensity exercise) was tracked at three time-points in the intervention period (baseline, mid-intervention and post intervention), and for the duration of each exercise bout by Garmin VivoSmart™ HR activity monitor (Garmin, USA).
Cephalic/cubital venous blood samples were taken at pre-intervention (0 weeks), mid-intervention (6 weeks) and post-intervention (13 weeks) time-points, into BD Vacutainer® MAP K2EDTA 1.0 mg tubes (BD, USA). Following removal of whole blood (2 × 1 mL) for comet assay and GSH analysis, samples were centrifuged (2000 r.p.m., 15 min, 10 °C) and the resultant plasma was aliquoted for subsequent determination of FRAP (ferric-reducing ability of plasma), TBARS (thiobarbituric acid-reactive substances), 8-isoprostane, BDNF, IL-6, CRP, TNF-α and protein carbonyl. All samples were stored at −80 °C for a maximum of 9 months until analysis.
MRS GSH measurement
Scans were conducted at the Birmingham University Imaging Centre, using a 3 Tesla Phillips Achieva MRI (magnetic resonance imaging) scanner, with a 32-channel head coil. The 1H single-voxel Mescher–Garwood point-resolved spectroscopy method of acquisition was employed, with repetition time = 2 s; echo time = 131 ms; 55 Hz bandwidth editing pulse at 4.56 p.p.m.; and 1024 complex data points acquired at a sampling frequency of 2000 Hz, followed by water suppression. The MRS protocol included a T1-weighted structural MRI for MRS planning (5 min). The volume of interest was located in the anterior cingulate cortex (30 × 30 × 20 mm3) (Fig. 1). Total MRS scan time for GSH measurement was 18 min. Spectral alignment was completed using the RATS (Robust Alignment to a Target Spectrum) method52, implemented in R (v3.5.0) (Vienna, Austria), and integrated into the SPANT (SPectroscopy ANalysis Tools) package (v0.12.0) for MRS analysis. GSH was then fitted using TARQUIN. An example spectrum is detailed in Fig. 1, following post-processing, with the GSH peak highlighted.
Whole-blood GSH concentration was determined using a commercially available luminescence-based assay (GSH-Glo™ Glutathione Assay, Promega, WI, USA). The assay was undertaken according to the manufacturer’s instructions.
Cells were prepared for assessment of single-strand DNA strand breaks, characterised by Singh et al.53. Thawed whole-blood samples were used for analysis, breaking away from the conventional use of isolated mononuclear cells, replicating the protocol described by Akor-Dewu et al.54. Active SYBR-GOLD nucleic acid gel staining solution (1:1000 dilution of stock solution in neutralisation buffer) was distributed across the slides and incubated at room temperature before scoring (tail length representing single-strand break density), using a fluorescent microscope (Zeiss Axiovert 10, Germany).
8-Isoprostane and TBARS concentrations were used to assess lipid peroxidation in the plasma. For TBARS, plasma samples and standards (1,1,3,3-tetramethoxypropane) (100 µl) were mixed with trichloroacetic acid (410 mM, 100 µl) and colour reagent (4.6 mM thiobarbituric acid, 1.74 M glacial acetic acid and 0.67 M butylated hydroxytoluene, 800 µl) in Eppendorf tubes. After boiling vigorously in water (100 °C) for 1 h, the reaction was stopped by placing the tubes in an ice bath for 10 min. Supernatants were transferred to a multiwell plate and the absorbance was measured at 540 nm. The concentration of 8-isoprostane in plasma samples was determined by competitive enzyme-linked immunosorbent assay using a commercially available kit (8-isoprostane EIA Kit 516351, Cayman Chemical, Ann Arbor, MI), according to the manufacturer’s instructions. After pretreatment of the plasma sample to concentrate the total 8-isoprostane content, affinity sorbent (8-isoprostane affinity sorbent 401113-1, Cayman Chemical, Ann Arbor, MI) was used.
Protein carbonyl concentration was determined in the plasma. Anti-DNP (2,4-dinitrophenol) antiserum primary antibody (1:1000) and peroxidase-labelled secondary antibody (rat anti-mouse IgE in blocking buffer 1:5000) were used, following the protocols described by Buss et al.55 and Alamdari et al.54.
Plasma IL-6, CRP and TNF-α concentrations were determined using commercially available immunoassay kits (Human IL-6 Quantikine ELISA Kit (D6050), Human TNF-alpha Quantikine ELISA Kit (DTA00C) and Human C-Reactive protein/CRP Quantikine ELISA Kit (DCRP00)), from R&D Systems (Minnesota, USA). Concentrations were determined according to the manufacturer’s instructions.
Total antioxidant capacity using the FRAP method
The FRAP method was developed by Benzie and Strain56. Standards were prepared using a 0–1000 μM concentration range of ascorbic acid. FRAP reagent—consisting of acetate buffer (300 mM sodium acetate at pH 3.6 (3.1 g) into neat 16 mL glacial acid per litre of buffer solution), TPTZ (2, 4, 6-Tris (2-pyridyl)-S-triazine) solution (160 mM: 0.05 g/mL–0.1 g TPTZ in 2 mL methanol, then 2 mL into 30 mL 40 mM HCl) and FeCl3·6H2O solution (0.332 g ferric chloride in 100 mL ddH2O)—was added to samples/standards on the plate, incubated for 8 min at room temperature and read at 650 nm. FRAP concentrations were determined by linear regression relative to ascorbic acid.
Brain-derived neurotrophic factor
Plasma BDNF concentration was determined using a commercially available immunoassay kit (Human BDNF ELISA Kit (ab99978), Abcam, Cambridge, UK). Results using this sandwich ELISA Kit were determined according to the manufacturer’s instructions.
At each assessment time-point, participants completed the PANSS57, via a structured clinical interview designed to monitor symptoms of psychosis. The PANSS interview assesses positive and negative symptoms, and is widely considered the ‘gold-standard’ method of quantifying psychotic behaviour. Interviewers were trained in the completion of PANSS.
Data analysis was performed using GraphPad Prism 8 software (version 8.0.1, 2018). At baseline, relationships between markers were determined using linear regression, and to assess any difference at baseline between the two groups, two-sample t tests were used. To assess changes between different time-points in the study, paired t tests were used. To compare relationships between marker pre-intervention/control period vs. mid or post, Pearson’s correlation coefficient was employed. Biomarkers were assayed in triplicate, with average values calculated. Outlying values were identified using the ROUT (Robust regression and Outlier removal) method (Q = 1%). Shapiro–Wilk test for normality was used for Gaussian distribution, with α significance level set at 0.05. Standardised mean difference (SMD) test was also used to assess effect size between time points and intervention groups, using the standard deviation of paired differences.
Twenty-two early intervention service users were recruited and randomised into the study, aged between 17 and 34 (average length of service use, at recruitment, was 19 months). Baseline characteristics are presented below (Table 1). From a potential caseload of 134, 67 patients did not meet eligibility criteria. Of the 67 remaining, with potential for inclusion into the study, 22 were randomised. Reasons for eligible patients not being randomised included discharge from the early intervention services, work/university time commitments, no interest in taking part in the study and an inability to make initial contact with the patient. Fifteen participants completed the trial, across the exercise (n = 7) and control (n = 8) groups. Peripheral biomarker analysis and symptom assessment was completed for all participants, whereas only four members of each intervention group were able to complete the MRS brain scan.
Brain and blood GSH
Peripheral whole-blood GSH increased by 6.13% in the exercise group, but significantly decreased by 24.37% in the control group (p = 0.04), in a group–time interaction (SMD = 0.47).
Ineligibility for scanning included foreign metal in the body (n = 2), anxiety/fear of enclosure in the scanner (n = 6) and injury preventing prolonged periods of time in the supine position (n = 1). Brain GSH concentration increased by 8.50%, and 8.80% in the exercise and control groups, respectively (SMD = −0.06). There was no significant change in brain GSH data as a result of 12 weeks of exercise (p = 0.55), but these data are limited by the small sample of the cohort that were able to be scanned (exercise group n = 4 and control group n = 4 for pre–post scans).
The Pearson’s correlation coefficient for the relationship between blood and brain GSH showed that in the control group, pre and post r values were 0.37 and 0.23, respectively, indicating weak positive correlation at both time points. The r value for the exercise group pre-intervention was 0.91, and post intervention 0.06. These changes indicate that, in this exercising group of participants, exercise affected peripheral and brain GSH regulation differently. Both blood and brain GSH changes are presented in Fig. 2. Additionally, in the exercise group, the relationship between GSH change and intensity minutes increase, as a result of being in the intervention, was strongly positively correlated, for both blood (r = 0.61) and brain (r = 0.68) measures.