Edited by: Unn Kristin Haukvik, University of Oslo, Norway
Reviewed by: David M. Cole, University of Zurich, Switzerland; Nina Kraguljac, University of Alabama at Birmingham, United States
*Correspondence: Emma K. Towlson,
This article was submitted to Schizophrenia, a section of the journal Frontiers in Psychiatry
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
The study of brain networks, including those derived from functional neuroimaging data, attracts a broad interest and represents a rapidly growing interdisciplinary field. Comparing networks of healthy volunteers with those of patients can potentially offer new, quantitative diagnostic methods and a framework for better understanding brain and mind disorders. We explore resting state functional Magnetic Resonance Imaging (fMRI) data through network measures. We construct networks representing 15 healthy individuals and 12 schizophrenia patients (males and females), all of whom are administered three drug treatments: i) a placebo; and two antipsychotic medications ii) aripiprazole and iii) sulpiride. We compare these resting state networks to a performance at an “N-back” working memory task. We demonstrate that not only is there a distinctive network architecture in the healthy brain that is disrupted in schizophrenia but also that both networks respond to antipsychotic medication. We first reproduce the established finding that brain networks of schizophrenia patients exhibit increased efficiency and reduced clustering compared with controls. Our data then reveal that the antipsychotic medications mitigate this effect, shifting the metrics toward those observed in healthy volunteers, with a marked difference in efficacy between the two drugs. Additionally, we find that aripiprazole considerably alters the network statistics of healthy controls. Examining the “N-back” working memory task, we establish that aripiprazole also adversely affects their performance. This suggests that changes to macroscopic brain network architecture result in measurable behavioral differences. This is one of the first studies to directly compare different medications using a whole-brain graph theoretical analysis with accompanying behavioral data. The small sample size is an inherent limitation and means a degree of caution is warranted in interpreting the findings. Our results lay the groundwork for an objective methodology with which to calculate and compare the efficacy of different treatments of mind and brain disorders.
In recent years, neuroimaging data and graph theory have allowed for the description of the topological properties of large-scale brain networks (
There is a body of evidence demonstrating drug treatments lead to specific localized changes in functional network structure (
Our results show that schizophrenia patients and healthy controls exhibit different network topologies, in agreement with the existing literature (
Twelve people with chronic schizophrenia and 15 healthy, nonpsychotic volunteers were recruited for participation in this study (see
Every subject attended three scanning sessions, each 1 to 2 weeks apart, for collection of functional MRI data and completion of working memory tests (see below). At each visit, they were administered one of three drug treatments: i) an oral placebo, 180 and 90 min before scanning; ii) oral aripiprazole, 15 mg 180 min before scanning and oral placebo, 90 min before; iii) oral placebo, 180 min before scanning and oral sulpiride, 400 mg 90 min before. We used a double dummy design with dosing of aripiprazole 180 min and sulpiride 90 min before the start of fMRI scanning. Both patients and experimenters were blind for the drug condition. The study medication was randomized by a colleague, who was not a member of the study team and stored in envelopes for each patient and testing session. There was a sealed envelope with the drug order for each participant that could be opened in case of a serious adverse effect. Both aripiprazole and sulpiride are antipsychotic medications designed to alleviate the symptoms of schizophrenia. At both time points (−180 min and −90 min), we co-administered 10 mg of domperidone to minimize side effects. Domperidone is a peripheral D2 receptor blocker sometimes used to mitigate nausea in pharmacological functional neuroimaging studies (
At each session, the subjects were required to complete an “N-back” task to assess their verbal working memory (
A General Electric (GE) Signa system scanner operating at 1.5 T at the BUPA Lea Hospital (Cambridge, UK) was used to acquire functional MRI data over 17 min and 12 s, during which time, subjects were asked to lie quietly with their eyes closed. In each session, 516 gradient-echo T2*-weighted echo planar images depicting blood oxygenation level-dependent (BOLD) contrast were generated from 16 noncontiguous near-axial planes: repetition time, 2 s; echo time, 40 ms; flip angle, 70°; voxel size, 3.05 × 3.05 × 7.00 mm; section skip, 0.7 mm; matrix size, 64 × 64; field of view (FOV), 240 × 240 × 123 mm. Four volumes were discarded to allow for T1 equilibration effects, leaving 512 volumes per data set (
Control 2 was missing an anatomical image so was discarded from the study, and patient 11 was missing data for the aripiprazole treatment. Each data set was analyzed for effects of head motion within the scanner (
The two-sample
To examine the effects of the drugs, volunteer type, and task difficulty on cognitive performance, we performed a three-way ANOVA with two repeated measures (
To quantify effect sizes between groups, we calculated Cohen’s
where
The Games-Howell
Then, we consider the difference between the means, which, to be considered significant, must satisfy:
We employ the Games-Howell
For each individual data set, voxel time series were averaged within each of the 325 equally sized anatomical regions in a random driven atlas [see Ref. (
Undirected weighted networks were generated for each individual based on correlating scale 3 wavelet coefficients. The resulting correlation coefficients
The
where
Average clustering and global efficiency values. The box plots display distributions of
The
If the shortest path lengths,
A measure of the
where
This is equivalent to averaging the nodal efficiencies for all nodes in the network.
The
where σ
Motion diagnostics, preprocessing and parcellation of the functional MRI data were completed using the preprocessing pipeline with temporal despiking from (
We first compared the functional brain networks derived from schizophrenia patients and healthy volunteers on placebo treatments and, as expected (
We next employed an ANOVA (two-way, one repeated-measure) to examine any differences in network measures between groups, and the factors underlying them. This confirmed group differences due to subject type on both efficiency and clustering (
Summary statistics for a 2 way ANOVA with 1 repeated measure on the network global clustering values of patients and healthy controls treated with placebo, aripiprazole, and sulpiride. Individuals for which networks were available for all drug treatments were used, equating to n = 12 for healthy controls and n = 9 for patients. There is a significant difference between the network clustering of the HV and SZ groups (p = 0.002), a significant drug effect (p = 0.005) and an additional drug-group type interaction term (p = 0.005)—all highlighted in red. This interaction term stems from the effect of aripiprazole—it greatly increases the clustering of control networks while causing only a small and variable increase in the schizophrenia networks. The placebo and sulpiride treatments have a more consistent effect on the two groups.
Source | SS | df | MS | F | p |
---|---|---|---|---|---|
Subject type | <0.001 | 1 | <0.001 | 8.093 | 0.010 |
Error | 0.001 | 19 | <0.001 | ||
Drug | <0.001 | 2 | <0.001 | 3.480 | 0.041 |
Drug×Subject type | <0.001 | 2 | 0.0001 | 1.005 | 0.376 |
Error | 0.0010 | 38 | <0.001 |
Summary statistics for a two-way ANOVA with one repeated measure on the network global efficiency values of patients and healthy controls treated with placebo, aripiprazole, and sulpiride. Individuals for which networks were available for all drug treatments were used, equating to
Source | SS | df | MS | F | p |
---|---|---|---|---|---|
Subject type | 0.026 | 1 | 0.026 | 12.208 | 0.002 |
Error | 0.040 | 19 | 0.002 | ||
Drug | 0.004 | 2 | 0.002 | 6.023 | 0.005 |
DrugxSubject type | 0.004 | 2 | 0.002 | 5.997 | 0.005 |
Error | 0.012 | 38 | <0.001 |
We found that aripiprazole has a dramatic effect on healthy individuals, with a large variation across individuals. We observe significantly reduced global efficiencies (median
All subjects score consistently highly on the very easy 0-back task, but their performance deteriorates considerably with increasing difficulty of the task, with subjects experiencing profound difficulty with the 3-back version (see
N-back working memory task. Panel
Summary statistics for a three-way ANOVA with two repeated measures on the hit rates during a working memory task with four levels of difficulty of patients and healthy controls treated with placebo, aripiprazole, and sulpiride. Individuals for which data were available for all drug treatments were used, equating to
Source | SS | df | MS | F | p |
---|---|---|---|---|---|
Subject type | 0.007 | 1 | 0.007 | 0.044 | 0.836 |
Error | 3.762 | 23 | 0.164 | ||
Within groups | |||||
Drug | 0.386 | 2 | 0.193 | 5.489 | 0.007 |
DrugxSubject type | 0.011 | 2 | 0.006 | 0.160 | 0.853 |
Error | 1.618 | 46 | 0.035 | ||
Cog. difficulty | 4.828 | 3 | 1.609 | 42.779 | <0.0001 |
Cog. difficultyxSubject type | 0.045 | 3 | 0.015 | 0.403 | 0.751 |
Error | 2.596 | 69 | 0.038 | ||
DrugxCog. difficulty | 0.120 | 6 | 0.020 | 1.160 | 0.331 |
Subject typexDrugxCog. difficulty | 0.162 | 6 | 0.027 | 1.160 | 0.331 |
Error | 2.384 | 138 | 0.017 |
An ANOVA (three-way, two repeated measures) revealed that, naturally, the predominant factor in determining success at the working memory tests was the difficulty of the task (
Patients scored worse than controls in the N-back working memory task, but not significantly so, with an average hit rate of 0.73 ± 0.18 for patients and 0.83 ± 0.19 for controls for the placebo 2-back task. Aripiprazole and sulpiride do not change the performance of patients much (average hit rates of 0.71 ± 0.21 and 0.79 ± 0.28, respectively, for the 2-back task) with an ANOVA showing no significant drug effect (
It has been known for some time that the whole-brain functional brain network organization of people with schizophrenia differs from that of healthy volunteers (
Our results suggest that there is an optimal configuration for a brain network in terms of maximizing cognitive ability: performance worsens given any change (increase
Schizophrenia implicates D2 receptors (among others) (
Consistent with the literature, our results indicated that aripiprazole significantly worsened the performance of healthy subjects at the N-back working memory task, but did not hinder the patients. Sulpiride had no significant detected effects. The main pharmacological difference between the two antipsychotics is that aripiprazole is a partial D2 antagonist and sulpiride is the most selective D2 antagonist. Given this, we might expect that sulpiride have a larger detrimental effect on the healthy controls’ performance, but it is possible that due to the relatively high dose of aripiprazole, it had stronger D2 antagonistic effects due to the relatively high dose of sulpiride. It is also surprising that we observed such small differences between the healthy and patient groups at the working memory tasks. However, we recruited relatively stable and high-functioning patients, and all participants were trained outside the scanner so that their performance was relatively stable. We aimed to match patients and controls for task performance so that drug effects were not confounded by group effects at the baseline (the placebo treatment).
The greatest limitation of this study is the small sample size:
Finally, there are likely confounding effects from other drugs. All patients were prescribed oral antipsychotic medication and asked not to take their medication on the days of the fMRI study (see
This quite unique data set allowed for an investigation into the effects of antipsychotic drugs on both the large-scale functional brain networks and the cognitive performance of people diagnosed with chronic schizophrenia and healthy volunteers. Despite its limitations, clear drug effects were observed on the network topology and performance at the N-back working memory task. We find a reduction in the difference between specific healthy and patient network metrics and that aripiprazole impairs cognitive ability and radically rewires the brain networks of healthy volunteers. It would be highly beneficial for future studies to use state-of-the-art functional MRI data to further investigate the links between disrupted networks in people with brain disorders, how medication influences these, and an “ideal” network topology for the brain (which should be identified by association with an optimal behavioral parameter, such as the best cognitive performance). Such advanced studies have the potential for not just diagnosis of the original brain disorder, but also for the quantification of the effectiveness of a drug in treating the illness. This would then allow for a systematic comparison between alternative treatments.
The datasets for this manuscript are not publicly available because the fMRI data were collected with formal approval from the Addenbrooke’s NHS Trust Local Research Ethics Committee in 2005. They did not approve making anonymized data available. Requests to access the datasets should be directed to ET (
This study was carried out in accordance with the recommendations of the Addenbrooke’s NHS Trust Local Research Ethics Committee with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the Addenbrooke’s NHS Trust Local Research Ethics Committee. All participants were provided with a detailed PIS that explained the nature of the pharmacological experiment and the double dummy design.
ET and SA conceived of the analysis. UM-S designed the clinical trials, recruited the subjects, and gathered the data. ET did the analysis, SA oversaw it, and ET, SA, and PV interpreted the results. ET, PV, and SA wrote the manuscript.
ET was supported by an Engineering & Physical Sciences Research Council (UK) PhD studentship, and is now supported by NSF award number 1734821. PV was supported by the Medical Research Council (MR/K020706/1), and is now a Fellow of MQ: Transforming Mental Health, grant number MQF17_24. The Behavioural & Clinical Neuroscience Institute is supported by the Medical Research Council (UK) and the Wellcome Trust. SA was supported by a Royal Society University Research Fellowship (UF080037 and UF120247), and is currently supported by a Gatsby Career Development Fellowship (PTAG/021). Data collection was supported by a grant from Bristol Myers Squibb to the late Robert Kerwin at King’s College London.
UM-S has received honoraria for advisory board participation, consultancy, and educational talks, travel expenses, and/or support for educational conferences from Astra Zeneca, Bristol Myers Squibb, Eli Lily, Heptares, Lundbeck, Flynn Pharma/Medice, Janssen-Cilag, Shire, and/or Sunovion.
The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
We thank Ameera Patel for discussions about pre-processing the raw fMRI data. We thank Dr. Rebecca Jones and Dr. Peter McKenna for their contribution to patient recruitment.
The Supplementary Material for this article can be found online at: