High long-term test–retest reliability for extrastriatal ¹¹C-raclopride binding in healthy older adults

Abstract

In vivo dopamine D2-receptor availability is frequently assessed with ¹¹C-raclopride and positron emission tomography. Due to low signal-to-noise ratios for ¹¹C-raclopride in areas with low D2 receptor densities, the ligand has been considered unreliable for measurements outside the dopamine-dense striatum. Intriguingly, recent studies show that extrastriatal ¹¹C-raclopride binding potential (BP_ND) values are (i) reliably higher than in the cerebellum (where D2-receptor levels are negligible), (ii) correlate with behavior in the expected direction, and (iii) showed good test–retest reliability in a sample of younger adults. The present work demonstrates high seven-month test–retest reliability of striatal and extrastriatal ¹¹C-raclopride BP_ND values in healthy, older adults (n = 27, age: 64–78 years). Mean ¹¹C-raclopride BP_ND values were stable between test sessions in subcortical nuclei, and in frontal and temporal cortices (p > 0.05). Across all structures analyzed, intraclass correlation coefficients were high (0.85–0.96), absolute variability was low (mean: 4–8%), and coefficients of variance ranged between 9 and 25%. Furthermore, regional ¹¹C-raclopride BP_ND values correlated with previously determined ¹⁸F-fallypride BP_ND values (ρ = 0.97 and 0.92 in correlations with and without striatal values, respectively, p < 0.01) and postmortem determined D2-receptor densities (including striatum: ρ = 0.92; p < 0.001; excluding striatum: ρ = 0.75; p = 0.067). These observations suggest that extrastriatal ¹¹C-raclopride measurements represent a true D2 signal.

Introduction

The dopamine (DA) system is implicated in a broad range of functions, including motor control, cognition, and reward.^1–3 It is a target of investigation in several disciplines, such as neuroscience, neurology, and psychiatry. Ever since its discovery in the 1980s,^4–6 the DA D2 receptor (D2DR) antagonist ¹¹C-raclopride has been a widely used radioligand in positron emission tomography (PET) studies of the DA system, e.g.^7-17. Its wide use is due to its pharmacological profile and high specificity for D2DRs,^18–21 its high reproducibility in measurements of D2DR availability at rest,²² and its high reliability in measurement of changes in synaptic DA levels via ligand displacement upon tasks and interventions.^23–27

¹¹C-raclopride is a moderate affinity ligand with low signal-to-noise ratios in areas with low D2DR levels. Seminal work demonstrated that raclopride binding is several-folds higher in the striatum compared to cortical regions²⁸ and only a few percent higher in frontal and temporal cortices compared to its inactive enantiomer, FLB472.²⁹ Therefore, it has not been considered reliable for D2DR assessments outside the DA-dense striatal compartment.^5,29,30 However, cortical raclopride binding depicted a rostral-caudal gradient, which is consistent with DA levels in these brain areas and was displaceable by D2-antagonists.³¹ More recent studies show that extrastriatal ¹¹C-raclopride binding potential (BP_ND), even though being low, is reliably above zero and within expected ranges^32,33 when comparing with D2 receptor levels quantified postmortem.^28,34 Additional support for the reliability of extrastriatal ¹¹C-raclopride BP_ND comes from recent findings showing that striatal, hippocampal, and cortical ¹¹C-raclopride BP_ND were interrelated, similarly linked to brain activation during a memory task, and predictive of episodic memory performance.^32,33,35 Furthermore, individuals with lower working memory performance had lower frontal cortical ¹¹C-raclopride BP_ND along with less beneficial brain activation patterns at rest and task.³⁶ In patient groups with DA disorders characterized by frontostriatal dysfunction (Parkinson's and Huntington's diseases), parallel reductions of striatal and cortical ¹¹C-raclopride binding have been reported.^37,38 Moreover, striatal as well as extrastriatal ¹¹C-raclopride displacement was observed in DA release paradigms in healthy individuals.^10,39–43 Taken together, these observations suggest that extrastriatal ¹¹C-raclopride may represent a meaningful D2DR signal, rather than noise.

Measurement reliability can be assessed by consecutive testing within a group of individuals. High test–retest reliability has been shown for striatal ¹¹C-raclopride measurements,^44–46 but also for extrastriatal ¹¹C-raclopride binding in cortical regions and thalamus.^47,48 Alakurtti and colleagues performed their investigations with a high-resolution brain-designated PET scanner in a small sample of younger adults. Consequently, it remains unclear whether their findings can be generalized to other settings, such as when using a scanner with lower resolution or when examining samples of older ages with varying degrees of age-induced DA decline.⁴⁹ For this reason, the present work examined ≥6-month test–retest reliability of extrastriatal ¹¹C-raclopride BP_ND in a sample of healthy, older adults (n = 27, ages: 64–78 years, 10 men) belonging to the Physical Influences on Brain in Aging (PHIBRA) study.⁵⁰ To elucidate the validity of our extrastriatal ¹¹C-raclopride signal, we compared regional ¹¹C-raclopride BP_ND values with values determined with the high-affinity ligand ¹⁸F-fallypride,⁵¹ and also, with D2DR density values (B_max) determined with ³H-raclopride in postmortem autoradiography assessments.²⁸

Material and methods

The study was approved by the Swedish Ethical Review Authority (Umeå, Sweden; registration number: 2013-238-31M) and was carried out in accordance with The Code of Ethics of the World Medical Association (Declaration of Helsinki). Written informed consent was obtained from all participants prior to any testing was done.

Sample

The analyses were performed using the participants randomized to the control group of the PHIBRA study, a six-month aerobic exercise intervention (see Jonasson et al. (2017) for details).⁵⁰ Between test sessions, individuals included in the present work engaged in exercises of stretching and toning to improve muscle strength, flexibility, and balance. Participants underwent brain assessments, including magnetic resonance imaging (MRI) and PET with ¹¹C-raclopride, at two occasions separated by approximately seven months (mean: 7.3 months; min: 6 months; max: 8 months). Exclusion criteria included factors that affect brain and cognition, such as diabetes; medication known to influence the DA system; neurological, psychiatric, and motor diseases; head trauma; and Mini Mental State Examination scores below 27. Further exclusion criteria consisted of MR-incompatible factors, e.g. claustrophobia and metal implants. All structural MRI images were inspected by a radiologist to screen for structural abnormalities.

BP_ND values were excluded for two individuals (from of the original 29), as the percent change between test session was found extreme according to the outlier labeling rule with 2.2 interquartile ranges.⁵² Thus, the effective sample consisted of 27 healthy older adults (age: 64–78 years, 10 men; Table 1).

Table 1.

Sample descriptives.

	Frequency or mean (SD)	min, max
Men	10	64, 78
Women	17	7, 25
Age	69.0 (2.9)	19.4, 37.3
Years of education	13.7 (4.7)	28, 30
Years of education	26.7 (3.5)
MMSE	29.5 (0.6)

BMI: body mass index; max: maximum; min: minimum; MMSE: Mini Mental State Examination; SD: standard deviation.

Brain imaging

MRI

MRI was performed with a 3T Discovery MR 750 scanner (General Electric, WI, US), equipped with a 32-channel phased-array head coil. A 3D fast spoiled gradient-echo sequence was used to obtain high-resolution anatomical T1-weighted images. Imaging parameters were 176 slices with 1 mm thickness, TR = 8.2 ms, TE = 3.2 ms, flip angle = 12°, and field of view = 25 × 25 cm.

The longitudinal image processing pipeline in Freesurfer, version 6⁵³ was used to process T1 images and achieve subcortical⁵⁴ and cortical gray-matter⁵⁵ segmentations of regions-of-interest (ROIs) for the two time points. Following cortical parcellation, subregions were merged to represent the following ROIs: orbitofrontal (lateral and medial orbitofrontal), anterior cingulate cortex (ACC; rostral and caudal anterior cingulate), superior frontal, middle frontal (rostral and caudal middle frontal), inferior frontal (pars opercularis, pars triangularis, and pars orbitalis), and temporal (superior, middle, and inferior temporal) cortices (Table 2 and Supplementary Table 1).

Table 2.

Test–retest statistics for ¹¹C-raclopride BP_ND measurements performed approximately seven months apart.

	Baseline		Follow-up		VAR (%)	min; max	ICC	r
	Mean (SD)	CV (%)	Mean (SD)	CV (%)	mean (SD)
Caudate	2.276 (0.306)	13.4	2.215 (0.283)	12.8	6.4 (4.6)	0.4; 16.7	0.91	0.82
Putamen	3.384 (0.336)	9.9	3.326 (0.290)	8.7	4.1 (3.4)	0.3; 12.9	0.91	0.82
Hippocampus	0.267 (0.052)	19.5	0.264 (0.048)	18.2	7.7 (5.3)	0.0; 27.4	0.95	0.91
Amygdala	0.393 (0.049)	12.5	0.393 (0.051)	13.0	6.7 (6.1)	0.0; 21.7	0.86	0.74
Pallidus	1.384 (0.149)	10.8	1.352 (0.142)	10.5	6.0 (4.7)	0.7; 18.4	0.85	0.75
Thalamus	0.486 (0.059)	12.1	0.487 (0.054)	11.1	4.7 (3.4)	0.0; 11.5	0.94	0.90
Orbitofrontal	0.249 (0.046)	18.5	0.255 (0.044)	17.3	7.0 (6.0)	0.0; 20.5	0.93	0.88
ACC	0.256 (0.047)	18.4	0.261 (0.038)	14.6	7.0 (4.7)	0.0; 16.0	0.93	0.83
Sup. frontal	0.193 (0.045)	23.3	0.192 (0.036)	18.8	8.2 (6.5)	0.0; 24.5	0.93	0.88
Mid. frontal	0.227 (0.042)	18.5	0.226 (0.037)	16.4	5.3 (3.9)	0.0; 13.4	0.96	0.93
Inf. frontal	0.208 (0.051)	24.5	0.207 (0.050)	24.2	7.9 (8.4)	0.0; 29.7	0.95	0.92
Temporal	0.277 (0.038)	13.7	0.283 (0.038)	13.4	4.9 (3.8)	0.0; 16.3	0.95	0.91

CV: coefficient of variance; ICC: intraclass correlation coefficient; max: maximum; min: minimum; r: Pearson's correlations coefficient (adjusted for gray-matter volume); SD: standard deviation; VAR: absolute variability.

Note: p > 0.05 for comparisons of mean values between sessions.

PET

PET was performed with a Discovery PET/CT 690 (General Electric, WI, US) and ¹¹C-raclopride. Head movements were minimized with individually fitted thermoplastic masks that were attached to the bed surface. A 55 min, 18-frame dynamic PET scan was acquired during resting-state conditions following intravenous bolus injection of approximately 250 MBq ¹¹C-raclopride (baseline: 276.00 ± 12.22 MBq; follow-up: 269.10 ± 10.91 MBq; p > 0.05). Mass of injected raclopride was lower at baseline (0.13 ± 0.06 µg) than at follow-up (0.48 ± 0.15 µg; t(19) = −12.28, p < 0.001), due to differences in specific activity of batches (baseline: 771.78 ± 341.70; follow-up: 190.96 ± 70.64 GBq/µmol; t(19)=7.70, p < 0.001). A CT scan (20 mA, 120 kV, 0.8 s/revolution) preceded ligand injection for attenuation-correction purposes.

Attenuation- and decay-corrected images (47 slices, field of view = 25 cm, 256 × 256-pixel transaxial images, voxel size = 0.977 × 0.977 × 3.27 mm³) were reconstructed with the iterative point-spread function ordered subset maximization (PSF-OSEM) algorithm VUE Point HD-SharpIR (GE⁵⁶; six iterations, 24 subsets, 3.0 mm post filtering), yielding full width at half maximum (FWHM) of 3.2 mm.⁵⁷

The following procedures were performed to determine ¹¹C-raclopride BP_ND in striatal and extrastriatal regions. PET images were motion corrected and co-registered to the structural T1-weighted images from the corresponding session (baseline and follow-up) using the Statistical Parametric Mapping software (SPM8). Within T1-segmented subcortical and cortical structures BP_ND was calculated with orthogonal regression reference Logan analysis⁵⁸ and time–activity curves with the median of ROI voxel values from each time frame. Logan regression was based on frame 10–18 (18–55 min). Cerebellar gray matter served as the reference area, due to its negligible D2DR expression.^5,34

Statistical analyses

Median ¹¹C-raclopride BP_ND was extracted for each ROI and subject, averaged over hemispheres, and entered into analyses. Results for the sample are presented with frequencies, means, standard deviations (SDs), and minimum and maximum values. Differences in mean values for ¹¹C-raclopride BP_ND, gray-matter volumes, injected radioactivity dose (MBq), and specific radioactivity (GBq/µmol) were compared between sessions with paired sample-tests.

Reliability was assessed with the intraclass correlation coefficient (ICC), and also, partial correlations (covariate: regional gray-matter volumes) with the Pearson's correlation coefficient (r). Test–retest absolute variability (VAR, %) between sessions was calculated by entering BP_ND values from the two sessions in the equation below

$$\begin{array}{c}\mathit{VAR}=\frac{2|\mathit{test}-\mathit{retest}|}{\mathit{test}+\mathit{retest}}·100\end{array}$$

Furthermore, the coefficient of variance (CV), defined as the ratio of SD to the mean for the sample (in %), was calculated to compare dispersion of values around the mean for the regions analyzed.

For indications of validity, our regional ¹¹C-raclopride BP_ND values were compared with previously determined regional values of ³H-raclopride B_max²⁸ and ¹⁸F-fallypride BP_ND (available at https://osf.io/h67k4/). Whereas ¹⁸F-fallypride values were found for all 12 brain regions reported in Table 2 of the present work, ³H-raclopride B_max for the superior frontal cortex was an outlier (>3 SD from the other cortical values) and the inferior frontal cortex was not reported by Hall et al.,²⁸ and thus, these data points are missing (Figure 1(a) and (b)). Associations among these values are described with the Spearman's rank correlation coefficient (ρ). As DA constituents decline in aging,^{49 18}F-fallypride data were attained for a subsample aged >60 years (mean: 69.4, SD: 6.8, n = 29) for approximate age-matching with the PHIBRA data. Furthermore, ¹⁸F-fallypride BP_ND values were averaged over subregions and hemispheres to represent the ROIs. Values for B_max based on ³H-raclopride from Hall et al.²⁸ (Figure 1(b) in the original publication) were digitized with the PlotDigitizer software (http://plotdigitizer.sourceforge.net).

Figure 1.

Regional coherence between our ¹¹C-raclopride BP_ND measures and previously determined ³H-raclopride B_max²⁸ and ¹⁸F-fallypride BP_ND⁵¹. Scatter plots illustrate associations when including striatum and globus pallidus (a), and low uptake-regions (b), which are visualized in the gray box in (a). Similarly, high correlations were found between our ¹¹C-raclopride BP_ND measures and ¹⁸F-fallypride BP_ND over regions when including striatum and globus pallidus (d), and only the low uptake-regions (d).

Results

High seven-month test–retest reliability for striatal and extrastriatal ¹¹C-raclopride binding

Test–retest statistics demonstrated high reliability for ¹¹C-raclopride BP_ND measurements in both striatal and extrastriatal regions (Table 2), with minor differences between hemispheres (Supplementary Table 1). No between-scan differences were observed for mean values in any of the regions (p > 0.05). CV indicated that the extent of variability in relation to the means was similar for cortical and subcortical regions, albeit lowest for the putamen (∼9%) and slightly elevated for frontal cortical regions (∼20%). Mean VAR was similar among regions, ranging between approximately 4 and 8%, and with comparable minimum and maximum values. ICCs were all >0.85, and thus in the good-to-excellent range.^59,60 Adjusting for gray-matter volumes, first-order correlations indicated high coherence for within-subject BP_ND values from the two sessions (rs: 0.74–0.93). Notably, similar values were achieved in zero-order correlations, i.e. without adjustment for volumes (rs for putamen: 0.85, caudate: 0.85, hippocampus: 0.91, amygdala: 0.74, globus pallidus: 0.75, thalamus: 0.89, orbitofrontal: 0.87, ACC: 0.89, superior: 0.89, middle: 0.93, inferior frontal: 0.91, and temporal: 0.91 cortices). Aside from the hippocampus,⁵⁰ volumes for all other structures remained similar between test sessions (p > 0.10).

Due to the significant differences between mass of injected raclopride at the baseline and follow-up sessions, we performed zero-order correlations among the percent change in injected raclopride and (1) percent change in ¹¹C-raclopride BP_ND values and (2) VAR in ¹¹C-raclopride BP_ND values. No significant associations were observed for any of the ROIs in Table 2.

Relation between ¹¹C-raclopride BP_ND and previously determined D2-receptor levels

¹¹C-raclopride BP_ND values from our study were compared with previously determined ³H-raclopride B_max²⁸ and ¹⁸F-fallypride BP_ND (https://osf.io/h67k4/) in the corresponding regions. High correlations were found between our ¹¹C-raclopride BP_ND measures and ³H-raclopride B_max over regions when including striatal and pallidal ROIs (ρ = 0.92, p < 0.001; Figure 1(a)), but also, among low D2DR density-regions (ρ = 0.75, p = 0.067; Figure 1(b)). In addition, high correlations were found between our ¹¹C-raclopride BP_ND measures and ¹⁸F-fallypride BP_ND values over regions, with small differences in the strength of the correlation when excluding high-density regions (ρ = 0.97; p < 0.001 and ρ = 0.92; p < 0.01 with and without striatal and pallidal ROIs, respectively; Figure 1(c) and (d)).

Discussion

The use of high-affinity ligands such as ¹¹C-FLB457⁶¹ and fallypride⁶² has been recommended for extrastriatal D2DR assessments. Notably, high-affinity ligands come with their own limitations, including unreliable striatal BP_ND estimations with ¹¹C-FLB457⁶³ and unreliable measurements of DA release with ¹⁸F-fallypride displacement.⁶⁴ Hence, it is well-motivated to further investigate the feasibility of extrastriatal ¹¹C-raclopride assessments. Here, we demonstrate high test–retest reliability for striatal as well as extrastriatal measurements performed seven months apart in older adults. Striatal and extrastriatal values were comparable in terms of degree of reliability, absolute variance, and dispersion of BP_ND values around the mean, with small differences between hemispheres. Importantly, ¹¹C-raclopride BP_ND values throughout frontal, temporal, and subcortical regions were ranked according to previously determined ¹⁸F-fallypride BP_ND,⁵¹ as well as ³H-raclopride B_max from postmortem autoradiography.²⁸ As such, these data add evidence for feasibility and validity of extrastriatal ¹¹C-raclopride assessment.

Test–retest reliability measures the consistency over measurement occasions. Test scores can vary due to variability in true scores and measurement errors. With the estimated striatal D2DR decline being around 5–10% per decade,^3,49 D2DR levels were not expected to change considerably over seven months in this group of healthy, older adults. Any observation of variability would then likely arise from measurement error. In line with previous reports, we noted high test–retest reliability for striatum,^44–46,65 but also for extrastriatal regions.^47,48,66 This stability may relate to factors such as high camera resolution and sensitivity, minimal movement artifacts (achieved via masks and movement correction of images), and small differences in blood flow between sessions.⁶⁷ A recent finding shows, however, that effects from changes in blood flow are negligible.⁶⁸ When it comes to data modeling, BP_ND values estimated with Logan analysis have been shown to be reliable and comparable to those estimated with simplified reference tissue modeling.^69,70 The underlying reasons for the large variability in specific activity between batches at the baseline and follow-up sessions remain unknown but could be a consequence of a service of the system that took place between the two test sessions. At other PET sites, fluctuations in specific activity over time have been related to, e.g. changes in the flow of the carrier gas from the target.⁷¹ That said, the levels of injected mass of raclopride were at tracer doses, not exceeding those of previous reports, e.g. ^5,48,72-73, and thus, the number of available receptors should have remained largely unaffected throughout the experiments. This is supported by the stable mean values for ¹¹C-raclopride BP_ND values between test sessions.

To ensure minimal variability of volumes between sessions, the longitudinal pipeline of the Freesurfer software was used. It is well known that the volume of ROIs may cause under- and overestimations of BP values.⁷⁴ Partial volume effects, caused by limited spatial resolution of the scanner and reconstruction algorithms, give rise to varying recovery of the signal in different regions, where structures having high surface-to-volume ratios are most vulnerable.^75,76 FWHM of the PSF-OSEM algorithm has been estimated to be 3.2 mm on the scanner used in the present study.⁵⁷ Furthermore, higher signal recovery was found for the reconstruction algorithm used here, PSF-OSEM, when compared to filtered back-projection, and did not induce a bias in the estimation of BP depending on the level of radioactivity.⁷⁶ Despite the varying morphology of the structures analyzed, reliability was constant over regions and correlations between ¹¹C-raclopride BP_ND values from the consecutive trials did not change when controlling for gray-matter volume.

When scrutinizing the statistics presented here, ¹¹C-raclopride BP_ND values for subcortical, temporal, and frontal cortical regions were coherent between test sessions at the group level. ICC values ranged from 0.85 to 0.96 and were thus in the range of excellent^59,60 and even higher than in the reports by Alakurtti et al.^47,48 The high ICC values indicate that variance arise from inter- rather than intra-individual variation. Furthermore, the VAR of BP_ND values as well as the dispersion of values around the mean was similar among regions with high and low D2DR expression. The coefficient of variation was, however, slightly elevated for frontal cortical ROIs (∼20%) when comparing with striatal, thalamic, and temporal regions (∼10%), which is on a par with previous reports.⁴⁷ In further comparisons among studies, the values for VAR for subcortical and cortical regions were similar to previous reports^47,65,66 and comparable to test–retest experiments with high-affinity D2DR ligands.^77,78 Consequently, it seems that ¹¹C-raclopride measurements are robust and consistent in regions with varying numbers of D2DRs. Since the present work replicates previous findings in young adults^47,48,66 reliability is not compromised by the well-known age-related DA decline.⁴⁹

It is critical to point out that high reliability does not imply validity of measurements, which in this context would translate to D2DR levels. With its moderate affinity to D2DRs, relatively low signal-to-noise ratio, and only slightly elevated binding when compared with its inactive enantiomer in areas with low D2DR levels,^{29 11}C-raclopride has been considered excellent for striatal and unfit for extrastriatal D2DR measurements. Early work showed minor accumulation of ¹¹C-raclopride in cortical areas^30,79 that was only a few percent higher than binding of its inactive enantiomer.²⁹ Low binding would, then again, be reasonable given that cortical D2DR levels constitute only a few percent of the striatal levels.^28,34,80 Our ¹¹C-raclopride BP_ND measures in striatal as well as extrastriatal regions were ranked according to previously determined ¹⁸F-fallypride BP_ND and ³H-raclopride B_max measures, thereby indicating that extrastriatal ¹¹C-raclopride BP_ND reflects the presence of D2DRs. Noteworthy, the BP_ND values were higher in several subcortical nuclei, but not cortical regions, when determined with fallypride as compared to raclopride. Consequently, the striatal-to-extrastriatal BP_ND ratios differed several-folds between the two D2DR ligands. This may relate to ligand characteristics; however, it remains for future research to understand the source of such observations. Using data from a multiligand study,⁸¹ Egerton et al.⁶⁷ did not observe similar coherence between extrastriatal ¹¹C-raclopride and ¹¹C-FLB457 binding. As pointed out by the authors, the lack of association may relate to the small sample size (n = 10). Small sample sizes are a limitation of most PET studies, and possibly, even more troublesome for the use of ¹¹C-raclopride in extrastriatal regions with low D2DR levels.

Other indications of validity come from studies showing ¹¹C-raclopride displacement during tasks in theoretically expected regions, including in prefrontal regions during working memory,¹⁰ in ACC and substantia nigra during a planning task,⁶⁷ and in motor regions upon motor actions.⁴² Furthermore, in-scanner interventions have resulted in extrastriatal ¹¹C-raclopride displacement during reward-settings,^39,82,83 but also upon pharmacological DA interventions.^{40,41,43,66,84} Findings from a well-powered ¹¹C-raclopride study have shown that both striatal and extrastriatal D2DR correlate and predict cognitive performance,^32,33,36 with observed allelic group differences when stratifying the sample according to a D2DR polymorphism.⁸⁵ Finally, 11C-raclopride binding was found to be organized according to anatomical and functional DA pathways, which further supports the validity of extrastriatal ¹¹C-raclopride measurements.⁸⁶

Concluding remarks

The findings presented here and elsewhere open up the possibility of a reliable extrastriatal ¹¹C-raclopride signal that is displaceable by interventions.⁶⁷ For definite conclusions of the nature of the extrastriatal signal, further observations of extrastriatal ¹¹C-raclopride displacement upon D2DR manipulation are needed in well-powered settings.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was funded by grants from the Swedish Research Council (2012-00530), Västerbotten County Council and Umeå University, the Swedish Research Council for Sport Science and Umeå School of Sport Sciences, the Kamprad Family Foundation, and the Swedish Brain Foundation. The Freesurfer analyses were performed on resources provided by the Swedish National Infrastructure for Computing (SNIC) at HPC2N in Umeå.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Authors’ contribution

NK and LJ performed research. C-JB, KR, LJ, and LN designed and funded the research; NK, LJ, JJ, GP, AS, and MA analyzed data. NK wrote the manuscript, which was edited by all authors.

Supplemental material

Supplemental material for this paper can be found at the journal website: http://journals.sagepub.com/home/jcb