Abstract
Background:
Depression and low mood are leading contributors to disability worldwide. Research indicates that clinical depression may be associated with low creatine concentrations in the brain and low prefrontal grey matter volume. Because subclinical depression also contributes to difficulties in day-to-day life, understanding the neural mechanisms of depressive symptoms in all individuals, even at a subclinical level, may aid public health.
Methods:
Eighty-four young adult participants completed the Depression, Anxiety and Stress Scale (DASS) to quantify severity of depression, anxiety and stress, and underwent 1H-Magnetic Resonance Spectroscopy of the medial prefrontal cortex and structural magnetic resonance imaging (MRI) to determine whole-brain grey matter volume.
Results/outcomes:
DASS depression scores were negatively associated (a) with concentrations of creatine (but not other metabolites) in the prefrontal cortex and (b) with grey matter volume in the right superior medial frontal gyrus. Medial prefrontal creatine concentrations and right superior medial frontal grey matter volume were positively correlated. DASS anxiety and DASS stress scores were not related to prefrontal metabolite concentrations or whole-brain grey matter volume.
Conclusions/interpretations:
This study provides preliminary evidence from a representative group of individuals who exhibit a range of depression levels that prefrontal creatine and grey matter volume are negatively associated with depression. While future research is needed to fully understand this relationship, these results provide support for previous findings, which indicate that increasing creatine concentrations in the prefrontal cortex may improve mood and well-being.
Keywords: Creatine, depression, grey matter volume, neuroimaging, prefrontal
Introduction
Depression and low mood are leading contributors to disability worldwide and can affect more than 300 million people at any one time (Britton, 2017). While most of the negative health and social effects of low mood are attributed to Major Depressive Disorder (MDD), individuals who experience subclinical depression (i.e. those who score below clinical thresholds on depression scales) also experience a significant impact on their daily functioning (Cuijpers et al., 2004). Importantly, the prevalence of subclinical depression/low mood may be increasing in society and may now affect a higher percentage of society (24%) than clinical depression (Van Zoonen et al., 2015). Because depression may be best understood as a continuum of symptoms, and because subclinical depression may be one of the best predictors of future MDD (Rodriguez et al., 2012; Van Zoonen et al., 2015), understanding the mechanisms of subclinical depressive symptoms in non-clinical groups may contribute to improvements in public mental health.
To aid development of novel therapies for depression, studies have attempted to determine the neurochemical mechanisms of depressive symptoms, yet many have ignored the potential importance of brain creatine. This is partly because when using the primary method for quantifying this endogenous compound in vivo (1H-Magnetic Resonance Spectroscopy (1H-MRS)), research groups have often referenced their metabolite of interest to creatine to ‘correct’ for concentrations of other metabolites (e.g. Kumar et al., 2002), on the basis that creatine was considered to be stable within regions of interest and across individuals (Li et al., 2003). However, prefrontal creatine is influenced by a range of factors, including cigarette smoking (Durazzo et al., 2016; Faulkner et al., 2021), cocaine use (Chang et al., 1997), cannabis use (Prescot et al., 2011), anxiety (Yue et al., 2012) and schizophrenia (Öngür et al., 2009), and can be influenced by other pathological conditions such as hepatic encephalopathy (Braissant et al., 2019), and has been shown to be almost absent in the brains of patients with cerebral creatine deficiency syndromes (e.g. Rackayova et al., 2017).
Currently, the potential relationship between brain creatine and depression is poorly understood. Specifically, no difference was found in anterior cingulate creatine metabolite concentrations between 19 depressed individuals diagnosed with MDD and 30 aged-matched controls (Auer et al., 2000), or in thalamic creatine concentrations between 18 young depressed individuals and 18 young non-depressed individuals (Mirza et al., 2006). Conversely, Kondo et al. (2016) reported a weak yet significant negative relationship between depression scores and concentrations of creatine in the frontal cortex of 22 adolescent participants diagnosed with MDD (p = 0.030; actual effect size not reported), indicating that high concentrations of creatine in the prefrontal cortex may be associated with lower levels of depression. Along these lines, Dechant et al. (1999) report that daily administration of 20 g creatine monohydrate for 4 weeks increased total brain creatine by up to 8.7% in a small sample of nine healthy individuals (see Allen, 2012, for a review), while Lyoo et al. (2012) report that daily supplementation of 5 g creatine monohydrate for 8 weeks augmented the antidepressant effects of escitalopram in 25 depressed females. Interestingly, Nemets and Levine (2013) report no significant effect of creatine supplementation when administered daily for only 4 weeks. However, this lack of an effect is likely due to, at least in part, the fact that the authors compared the effects of (a) 5 g creatine supplementation, (b) 10 g creatine supplementation and (c) placebo in small groups of only five, four and nine individuals, respectively. As such, creatine administration may positively influence depressive symptomatology, and determining whether there is a relationship between prefrontal creatine and depression may aid the treatment of this mood disorder.
Structural neuroimaging studies have also indicated that depression and low mood are related to lower grey matter volume in the prefrontal cortex. Three meta-analyses of 20+ studies that examined brain structure using voxel-based morphometry (VBM) revealed that compared to healthy controls, depressed patients exhibit lower grey matter volume bilaterally in the medial prefrontal cortex and anterior cingulate cortex (Bora et al., 2012; Lai, 2013; Wise et al., 2017). Furthermore, studies published more recently support these findings (e.g. Kandilarova et al., 2019). However, it is currently unknown whether this depression-related low prefrontal grey matter volume is related to alterations in prefrontal creatine concentrations.
In the current study, we examined the relationship between depression levels in a non-clinical sample and both prefrontal creatine metabolite concentrations (quantified using 1H-MRS) and grey matter volume (quantified using VBM). It was hypothesized that there would be a significant negative association between depression levels and both creatine concentrations and grey matter volume in the prefrontal cortex. For completeness, and on the basis of findings by Hasler et al. (2007) and Auer et al. (2000), who report that depressed individuals exhibit low prefrontal concentrations of glutamate and γ-aminobutyric acid (GABA) metabolites, we also performed exploratory analyses to determine the relationship between depression levels and concentrations of all metabolites quantified by the 1H-MRS sequence.
Methods
We report an analysis of data collected from two separate studies, both of which had the aim of collecting health-related magnetic resonance imaging (MRI) data in young adults. Both study protocols used the same MRI sequences for volumetric and 1H-MRS data (see below), and all data were acquired on the same 3T MRI scanner at the Combined Universities Brain Imaging Centre.
Participants
Across both studies, 84 participants were recruited via print and online advertisements. Thirty-eight of these subjects participated in Study 1, and 46 subjects participated in Study 2. All participants gave written informed consent after receiving a detailed explanation of their study procedures (approved by the University of Roehampton Research Ethics Committee) and were screened for eligibility. Exclusion criteria for both studies were as follows: self-report of psychiatric diagnoses, current drug use/abuse or dependence (other than tobacco and cannabis use), history of neurological injury or disease, pregnancy and contraindications for MRI (e.g. metal implants). Data from Study 1 were collected from November 2015 to March 2018, while data from Study 2 were collected from September 2017 to December 2019.
Questionnaire measures
Participants completed a demographics form (developed in-house) to determine age, gender, level of education, self-reported psychiatric comorbidity, neurological disorder and use of tobacco, cannabis and illicit drugs. Exposure to cigarettes was inferred from the average number of cigarettes smoked per day as in previous research (e.g. Durazzo et al., 2016; Faulkner et al., 2018, 2019, 2020). Participants also completed the Depression, Anxiety and Stress Scale (DASS), a 42-item questionnaire designed to quantify levels of depression, anxiety and stress; each of these emotional symptoms is assessed by summing the answers to fourteen 4-point Likert-type scale questions (Lovibond and Lovibond, 1995). On the Depression subscale, a score of 0 to 9 denotes no depression, 10 to 13 mild depression, 14 to 20 moderate depression, 21 to 27 severe depression, while 28+ denotes extremely severe depression. On the Anxiety subscale, a score of 0 to 7 denotes no anxiety, 8 to 9 mild anxiety, 10 to 14 moderate anxiety, 15 to 19 severe anxiety and 20+ denotes extremely severe anxiety. On the Stress subscale, a score of 0 to 14 denotes no stress, 15 to 18 mild stress, 19 to 25 moderate stress, 26 to 33 severe stress and 34+ denotes extremely severe stress.
To determine the influence of age, gender, cigarette smoking and cannabis use on depression, anxiety and stress, an analysis of variance (ANOVA) was constructed, in which scores from the DASS Depression, Anxiety or Stress subscales were added as the dependent variable (as relevant), and age, gender, the average number of cigarettes smoked per day and the average number of cannabis joints smoked per day were all added as separate factors.
1H-MRS data acquisition, preprocessing and analysis
All 1H-MRS scans were acquired using the same 3T Siemens Magnetom TIM Trio MRI system using a 32-channel head coil. 1H-MRS in vivo spectra were acquired from the same 20 × 20 × 20 mm3 voxel located in the right medial prefrontal cortex (typical location shown in Figure 1(a)). The structure and function of the medial prefrontal cortex are related to the clinical features of depression (e.g. Farb et al., 2011); this voxel placement therefore allowed us to test our own hypotheses pertaining to relationships of depression and brain chemistry. A medial position was also chosen, as lateral voxels can be harder to place due to tissue boundaries. The voxel was placed manually by referring to the individual subject’s T1-weighted (magnetization-prepared rapid gradient echo (MPRAGE)) scan. Specifically, we ensured that the voxel was placed very close to the midline of the brain, and as anterior as possible while avoiding gyri and cerebrospinal fluid. As such, the voxel was placed both anterior and slightly dorsal to the corpus callosum. Spectra were acquired using a SPin ECho full Intensity-Acquired Localized (SPECIAL; Mlynarik et al., 2006) spectroscopy 1H-MRS sequence with water suppression (repetition time (TR) = 3000 ms, echo time (TE) = 8.5 ms, Phase cycle Auto, 192 averages from the right prefrontal cortex voxel; Godlewska et al., 2015). Water unsuppressed spectra (16 averages) were also acquired. Outer volume suppression slabs were applied 5 mm from the edge of each side of the voxel (six slabs in total), both to suppress signals originating outside of the right medial prefrontal voxel of interest, and to minimize motion artefact effects on spectra within the voxel.
Figure 1.
(a) Typical 1H-MRS voxel placement in the medial prefrontal cortex. (b) Example attained spectrum from the medial prefrontal voxel seen in (a).
Spectra were analysed using LCModel 6.3-1L, with a basis set consisting of 19 simulated spectra, as in Morgenroth et al. (2019) and Faulkner et al. (2021) (for the full basis set, see Supplementary Materials). This basis set was simulated using FID appliance (FID-A) (Simpson et al., 2017) for TE = 8.5 ms, magnetic field strength = ~3 T and assuming ideal radiofrequency pulses. Line widths and signal-to-noise ratios were estimated as less than 8 Hz and greater than 40 Hz, respectively (Godlewska et al., 2015). Cramér–Rao lower bounds (Kreis, 2016), line widths and signal-to-noise ratios did not differ as a function of DASS scores (i.e. levels of depression, anxiety or stress) or between Study 1 and Study 2 (see Supplementary Materials).
Water referencing and eddy current correction were used to quantify metabolite levels. When quantified in this way, such levels are influenced by cerebral spinal fluid, grey and white matter volumes of the region in which spectra are obtained (Srinivasan et al., 2006), as well as by individual differences in whole-cortical grey matter (Huster et al., 2007). We therefore corrected these metabolite levels for grey and white matter content within the right medial prefrontal voxel using the GABA Analysis Toolkit (Gannet 3.1, http://gabamrs.blogspot.co.uk/), adapted to work with Siemens SPECIAL data. Segmentation was performed using ‘new segment’ in Statistical Parametric Mapping 12 (SPM12) (https://www.fil.ion.ucl.ac.uk/spm/software/spm12/). Cerebrospinal fluid and grey and white matter volumes were then accounted for in the expression of creatine using LCModel (Ernst et al., 1993; Gasporovic et al., 2006). For the specific calculation by which metabolites were corrected for using these volumes, see the Supplementary Materials.
To determine relationships between metabolite concentrations and DASS depression, anxiety or stress scores, an ANOVA was constructed in which DASS depression scores were added as the dependent variable, and age, gender, the average number of cigarettes smoked per day and daily cannabis use were all added as separate factors. For completeness, two exploratory ANOVAs were performed in which the dependent variables were (a) DASS anxiety scores and (b) DASS stress scores, to determine relationships between both anxiety and stress, respectively, and metabolite concentrations; these two ANOVAs both corrected for age, gender, daily cigarette smoking and daily cannabis use. Because the primary hypothesis pertained only to the relationship between depression and prefrontal creatine, and because all other analyses were considered secondary, corrections for multiple comparisons were not applied.
Structural MRI data acquisition, preprocessing and analysis
In both studies, high-resolution structural images were acquired using a T1-weighted MPRAGE sequence. Images were analysed using Computational Anatomy Toolbox 12 (CAT12; http://www.neuro.uni-jena.de/cat) implemented in SPM12 (Wellcome Trust Centre for Neuroimaging; fil.ion.ucl.ac.uk/spm/software/spm12/), as per the standard protocol (see http://www.neuro.uni-jena.de/cat12/CAT12-Manual.pdf); for a detailed description of the structural MRI (sMRI) data preprocessing steps, see the Supplementary Materials.
To determine the relationship between DASS depression scores and whole-brain grey matter volume, group-level analyses were performed by constructing a one-sample general linear model (GLM) in SPM12 that contained each participant’s modulated, normalized, segmented, registered and smoothed grey matter tissue segments, and one explanatory variable (DASS depression scores), along with separate variables for age, gender, daily cannabis use, smoking status (smokers vs non-smokers) and total intracranial volume to control for the influence of these variables. Contrasts were performed to identify regions in which whole-brain grey matter volume (a) positively and (b) negatively correlated with DASS depression scores. For completeness, two exploratory GLMs were performed; one that included DASS anxiety scores, and one that included DASS stress scores in place of depression scores. A threshold of p < 0.05 with Family-wise error (FWE) correction for multiple comparisons was applied to all contrasts.
To determine relationships between metabolite concentrations and grey matter volume that was associated with depression severity, bivariate correlations were performed; these models included only metabolite concentrations and values from significant clusters that were identified from the above volumetric contrasts.
All ANOVAs and correlational analyses were performed using both frequentist and Bayesian analyses. Frequentist analyses were performed using the Statistical Package for the Social Sciences (version 26, SPSS, Inc., Chicago, IL, USA). Bayesian analyses were performed using JASP (version 0.11.1; https://jasp-stats.org/download/).
For the frequentist analyses, a significance threshold of alpha = 0.05 (two-tailed) was adopted. For the Bayesian analyses, we adopted the thresholds set out by