Hippocampal Subfield Volumes in Amateur Marathon Runners

ABSTRACT

Purpose

Numerous studies have implicated the involvement of structure and function of the hippocampus in physical exercise, and the larger hippocampal volume is one of the relevant benefits reported in exercise. It remains to be determined how the different subfields of hippocampus respond to physical exercise.

Methods

A 3D T1-weighted magnetic resonance imaging was acquired in 73 amateur marathon runners (AMR) and 52 healthy controls (HC) matched with age, sex, and education. The Montreal Cognitive Assessment, the Pittsburgh Sleep Quality Index (PSQI), and the Fatigue Severity Scale were assessed in all participants. We obtained hippocampal subfield volumes using FreeSurfer 6.0. We compared the volumes of the hippocampal subfield between the two groups and ascertained correlation between the significant subfield metrics and the significant behavioral measure in AMR group.

Results

The AMR had significantly better sleep than HC, manifested as with lower score of PSQI. Sleep duration in AMR and HC was not significantly different from each other. In the AMR group, the left and right hippocampus, cornu ammonis 1 (CA1), CA4, granule cell and molecular layers of the dentate gyrus, molecular layer, left CA2–3, and left hippocampal–amygdaloid transition area volumes were significantly larger compared with those in the HC group. In AMR group, the correlations between the PSQI and the hippocampal subfield volumes were not significant. No correlations were found between hippocampal subfield volumes and sleep duration in AMR group.

Conclusions

We reported larger volumes of specific hippocampal subfields in AMR, which may provide a hippocampal volumetric reserve that protects against age-related hippocampal deterioration. These findings should be further investigated in longitudinal studies.

Physical exercise (PE) has beneficial effects on brain function and cognitive enhancement (1). Running, a vigorous-intensity aerobic exercise in nature (2), has greatly increased over the last decade. In recent years, participation in a “marathon run” has become increasingly attractive for millions of amateur runners worldwide. This trend might in part reflect the increasing awareness among the general population that PE provides health benefit and reduces all-cause mortality (3). Compared with the few professional athletes, amateur marathon runners (AMR) are large in number, and their goals are primarily to improve their physical fitness, enrich their leisure life, maintain emotional well-being, or challenge themselves (4,5). Several studies have reported changes of cardiac and renal function (6), electrocardiography (7), and iron metabolism (8) in AMR. However, relatively few studies with a large sample size have examined brain health benefits in this specific population (9). Therefore, understanding the alterations in brain physiology that associated with the running population is crucial for promoting the development of public health.

The hippocampus is essential for the mnemonic functions and spatial navigation (10). It is involved in cognitive functions and other complex behaviors, including sensorimotor integrations, stress responses, emotions, and sleep, all of which could be improved by PE (2,3). Among specific brain regions, hippocampus seems to be the key region responsive to PE (11,12), probably because of its plasticity and susceptibility to age-related atrophy (13). Noticeably, randomized controlled trials (RCT) investigating the effects of PE in healthy adults on hippocampal volumes have produced equivocal results. Two RCT found that aerobic exercise was associated with increased hippocampal volumes compared with stretching exercise in older adults followed for 1 yr (14,15). Recently, similar evidence for young to middle-age healthy adults has been reported that PE increases the volume of the left hippocampal head (16,17). Conversely, Tarumi et al. (18) found no evidence of larger hippocampal volumes in cognitively normal older adults after a 1-yr aerobic exercise or stretching program. Nonetheless, recent meta-analyses have indicated that PE is associated with larger total (19), left (11), and right (20) hippocampal volumes. These inconsistent results from RCT may be attributable to variations in sample sizes, type of PE (aerobic, resistance), intensity, or duration of the interventions.

PE can be undertaken in various domains, including leisure, occupation, education, household, and/or transport (3). One population-based study examined whether domain-specific physical activity is associated with brain volumes and found no association between domain-specific physical activity and hippocampal volume in volumetric analysis and voxel-based morphometry (21). In addition, PE provides a complex stimulus for adaptation in the body, and its effect can be modulated by various parameters, including intensity, frequency, duration, and the type or mode of exercise (22). Some cross-sectional studies have reported that greater time spent in moderate- to vigorous-intensity but not low-intensity physical activity was associated with greater hippocampal volume (23–25); other study has found the amount of low-intensity but not moderate- to vigorous-intensity physical activity to be predictive of hippocampal volume (26). Despite this discrepancy related to the intensity of the activity, these cross-sectional studies suggest that greater amounts of PE tend to be associated with greater hippocampal volume.

Although most aforementioned human studies investigating effect of PE on hippocampus have focused on its gross volume, evidence reveals the hippocampus to contain anatomical subfields with corresponding functional specialization (27,28). Basic preclinical research often investigates hippocampal function at individual subfields (29) and circuit-based level (30). It is conceivable that subfield changes may influence the total hippocampal volume, and subfield variations may be obscured when the hippocampus is examined as a whole. An approach examining individual hippocampal subfields may help bridge the elusive gap between clinical and preclinical hippocampal research (17,31). This approach would help to identify specific hippocampal neurobiology associated with hippocampal neuroimaging characteristics in running population.

Recent evidence suggests that PE improve sleep quality in healthy older adults (32). However, it remains unknown whether sleep quality is better in AMR. In addition, it has been reported that worse sleep or fatigue was associated with lower hippocampal volume in healthy older adults (33,34). Nonetheless, two studies tested the association between subjective sleep quality and longitudinal changes in hippocampal volume and did not detect significant effects (35,36). Worse sleep quality has been associated with reduced cornu ammonis (CA) 1 volume in patients with primary insomnia (37). It is possible that analysis of the hippocampal subfields volume may provide more sensitive measures related to sleep quality.

Advances in automated tissue segmentation methods have enabled examination of hippocampal subfields with improved accuracy (38). In the current study, hippocampal subfields volume in AMR and HC subjects were evaluated using this latest technology. Participants were studied using automated hippocampal subfield segmentation of high-definition T1-weighted magnetic resonance images. Our hypotheses were that AMR would have larger hippocampal subfield volumes than healthy controls (HC) and that significant subfield metrics would be related to the improved function.

MATERIALS AND METHODS

Participants

Participants were recruited through Wechat advertisements and running clubs in local community from March 2021 to July 2022. Written informed consent was obtained from each participant. The Ethics Committee of Renmin Hospital of Wuhan University approved the study protocol. The inclusion criteria for AMR were as follows: 1) they were not registered athletes and 2) they participated in regular training during the past 2 yr. The AMR participated in the official half-marathon at least three times, and a majority of them participated in the official full-marathon at least once. The AMR completed questionnaires about their exercise habits, including the duration of training and the number of full-marathon and half-marathon they participated. Age, sex, and education duration-matched HC were primarily recruited rather sedentary people showing no regular engagement in exercise. All participants were excluded if they had head trauma with residual effects, neurological or psychiatric disorders, any contraindications for MRI scan, uncontrolled major medical conditions based on self-reports, and scored <26/30 on the Montreal Cognitive Assessment (MoCA). For this study, we recruited a total of 75 AMR and 56 HC. Two participants in the AMR group and four participants in the HC group were excluded because they did not complete MRI protocol because of excessive head movement or claustrophobia. Therefore, the final participants included 73 AMR (AMR group, 42 males, 31 females; mean ± SD age, 43.59 ± 8.63 yr) and 52 HC (HC group, 30 males, 22 females; mean ± SD age, 41.69 ± 9.05 yr). The demographic and clinical characteristics are presented in Table 1.

TABLE 1.

Demographic and clinical characteristics of AMR and HC.

	AMR (n = 73)	HC (n = 52)	P
Age, mean ± SD (yr)	43.59 ± 8.63	41.69 ± 9.05	0.197a
Sex, n (% male)	42 (57.53%)	30 (57.69%)	0.476b
BMI, mean ± SD	22.23 ± 2.06	23.26 ± 2.53	0.014c
Education, mean ± SD (yr)	15.44 ± 2.46	16.06 ± 3.12	0.212a
Duration of training, mean ± SD (yr)	4.82 ± 2.55	n.a.	n.a.
Half-marathon time (n)	18.84 (26.22)	n.a.	n.a.
Full-marathon time (n)	11.78 (27.20)	n.a.	n.a.
MoCA, mean ± SD	27.71 ± 1.30	27.87 ± 1.44	0.622a
PSQI, mean ± SD	5.92 ± 3.76	9.34 ± 4.37	<0.0001a
FSS, mean ± SD	29.21 ± 12.69	33.40 ± 11.83	0.062a

^aMann–Whitney U-test.

^bChi-square test.

^cStudent’s t-test for unpaired samples.

n.a., not applicable.

Behavioral measures

Emerging evidence suggests that exercise is a promising strategy for combating cognitive decline (22,39). The MoCA was developed as a screening tool for mild cognitive impairment (MCI) and mild Alzheimer’s disease. It has been shown to have high sensitivity and specificity for differentiating individuals with MCI from healthy individuals in several developed countries and areas (40,41). The most widely used version of MoCA in mainland China is the Beijing version (MoCA-BJ); thus, we used MoCA-BJ in this study. MoCA-BJ has good internal consistency and criterion-related validity in general, and it is fairly reliable to differentiate MCI from normal aging and dementia (42). The sleep quality was assessed using the Pittsburgh Sleep Quality Index (PSQI). The PSQI is a frequently used validated 19-item questionnaire to assess quality of sleep in the last month. The PSQI consists of seven subscales: subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disorders, use of sleeping medication, and daytime dysfunction. By summing the subscale scores, a global sleep score reflecting overall sleep quality can be calculated, with higher score means the worse sleep quality (43). In general population, lower levels of physical activity (44) and poorer sleep quality (44,45) have been shown to be related to general feelings of fatigue. Fatigue Severity Scale (FSS) was designed for use with the general population and demonstrated good psychometric properties (46). The FSS is a 9-item self-report scale, with higher scores indicating greater fatigue. It has good internal consistency as indicated by a Cronbach α coefficient of 0.8; good test–retest reliability, as indicated by a correlation coefficient of 0.84; and sensitivity to change in clinical state (47). Thus, we used FSS to evaluate levels of fatigue in this study.

MRI data acquisition

All participants were instructed to avoid the ingestion of caffeinated substances or alcoholic drinks at least 3 d before scanning. MRI data were acquired at the Renmin Hospital of Wuhan University in a whole-body 3.0-T MRI scanner using a 16-channel head coil (Discovery MR 750; GE-Healthcare, Milwaukee, WI). High-resolution anatomical images were acquired using a sagittal three-dimensional T1-weighted BRAVO sequence with the following parameters: repetition time/echo time = 7.2/2.7 ms, inversion time = 450 ms, flip angle = 12°, number of slices = 160, slice thickness = 1.0 mm, field of view = 25.6 × 25.6 cm, and in-plane matrix = 256 × 256. All scans were visually inspected to rule out gross artifacts.

Image analysis

The volumes of hippocampal subfields were obtained using FreeSurfer software (version 6.0, http://surfer.nmr.mgh.harvard.edu/, RRID:SCR_001847), which is widely used for healthy adults (17,48) and patients with neuropsychiatric disorder (49,50). FreeSurfer interrogates contrast differences between substructures using previously defined in vivo and ex vivo hippocampal atlases to determine substructure characteristics (50). First, “recon-all” processing was conducted in FreeSurfer. The subcortical stat files were created using the “recon-all” built feature of the FreeSurfer software. The estimated total intracranial volume (ICV), which is required when evaluating the hippocampal subfields volume, was included in the stat file. Second, the hippocampal subfields segmentation protocol was used (38). The 12 hippocampal subfields volume and the whole hippocampus volume in each hemisphere were segmented and calculated by the protocol (Fig. 1). The reliability and validity of the FreeSurfer 6.0 hippocampal segmentation protocol has been previously demonstrated. The regional specificity of the PE was investigated by examining thalamus regions that served as control, which has previously been found to be unrelated to PE (14). As a control region, the volume of the whole thalamus and the thalamic nuclei group were also examined. Based on functional and anatomical overlap (anterior, lateral, ventral, intralaminar, medial, and posterior), we divided the thalamus into six groups (Fig. 2) as previously described (51). We derived the volume of each nuclei group by computing the sum of the nucleus belonging to a group (see Table S1, Supplemental Digital Content, which shows the structural divisions of the thalamus, http://links.lww.com/MSS/C805). All segmented scans were visually inspected by two coauthors independently following standard procedures to exclude segmentations with poor registration or with wrong assignment of the subfields. The outliers were checked within the data set and excluded if outside the five standard deviations.

FIGURE 1.

Hippocampal subfield segmentation. Figure displays the axial, coronal, and sagittal views of the hippocampal subfields in an example subject.

FIGURE 2.

Thalamic nuclei group segmentation. Figure displays the axial, coronal, and sagittal views of the thalamic nuclei group in an example subject.

Statistical analysis

Statistical analyses were performed using IBM SPSS Statistics (Version 22; IBM Corp., Armonk, NY), Graphpad Prism 7.0 (https://www.graphpad.com/, RRID: SCR_002798), and the R 4.2.2 (https://www.r-project.org/). All data were presented as mean ± SD. As a preliminary analysis, we used the Shapiro–Wilk test with statistical significance set at P < 0.05 to analyze the normality of value of age, body mass index (BMI), education years, PSQI scores, MoCA scores, and FSS scores. Demographic and clinical characteristics were compared between groups using Mann–Whitney test or two-tailed independent samples t-tests, and sex difference was evaluated using chi-square test. For each hippocampal subfield and thalamic nuclei group volume, we used multivariate ANCOVA to investigate the effect of diagnosis. Diagnosis group (between AMR and HC) was included as an independent variable, and total hippocampal volume and subfield volumes were included as a dependent variable, with sex, age, education years, BMI, and ICV as covariates. Partial eta squared (η²) was calculated to estimate effect sizes. The P values of the comparisons between the diagnosis groups were corrected using the Benjamini–Hochberg false discovery rate. The statistical significance metrics derived from the above volumetric analysis were included in the correlation analysis to observe the relationship between these metrics and PSQI as well as sleep duration. The partial correlation analyses were performed between the hippocampal subfield volumes and the PSQI score as well as sleep duration in AMR group controlling for sex, age, education years, BMI, and ICV. Statistical significance was set to P < 0.05.

RESULTS

Demographic and clinical characteristics of the groups

A total of 125 subjects were recruited (73 AMR and 52 HC). Table 1 presents the demographic and clinical characteristics of the study cohort. AMR and HC did not differ significantly regarding age (P = 0.197), education (P = 0.212), and sex (P = 0.476). The mean BMI was significantly different between AMR and HC (22.23 ± 2.06 vs 23.26 ± 2.53, P = 0.014). AMR had significantly better sleep than HC, manifested as with lower score of PSQI (5.92 ± 3.76 vs 9.34 ± 4.37, P < 0.0001). However, sleep duration in AMR and HC was not significantly different from each other (397.5 ± 58.02 min vs 400.2 ± 58.96 min, P = 0.346). In addition, AMR and HC had no significant difference in the MoCA (P = 0.622) and FFS scores (P = 0.062).

Volumetric analysis

In the present study, we found that AMR had larger volumes of hippocampal subfields, specifically in the left and right hippocampus, CA1, CA4, granule cell and molecular layers of the dentate gyrus (GC–DG), molecular layer, left CA2–3, and left hippocampal–amygdaloid transition area (HATA) compared with HC (Table 2, Fig. 3). In AMR, the correlations between the PSQI and the hippocampal subfield volumes were not significant (see Fig. S1, Supplemental Digital Content, Correlations between hippocampal volumes and the PSQI scores in amateur marathon runners, http://links.lww.com/MSS/C805). Also, no correlations were found between hippocampal subfield volumes and sleep duration in AMR (see Fig. S2, Supplemental Digital Content, Correlations between hippocampal volumes and the sleep duration in amateur marathon runners, http://links.lww.com/MSS/C805). The regional specificity of the PE was investigated by examining thalamus regions that served as control, which has previously been found to be unrelated to PE (14). The volume of the whole thalamus and the six thalamic nuclei group were larger in AMR compared with those in HC, but these trends were not significant (Table 3, Fig. 4).

TABLE 2.

Whole hippocampus and hippocampal subfield volumes (mm³).

	HC	AMR	P	Partial η2
Left hippocampus
Total volume	3509.71 ± 32.68	3633.24 ± 27.29	0.001**	0.939
CA1	625.88 ± 7.95	656.65 ± 6.64	0.001**	0.934
CA2–3	187.98 ± 2.52	195.08 ± 2.10	0.007**	0.913
CA4	234.85 ± 2.66	244.19 ± 2.23	0.001**	0.940
GC–DG	277.86 ± 3.08	289.94 ± 2.66	0.001**	0.944
Subiculum	476.31 ± 6.34	486.20 ± 5.29	0.133	0.742
Presubiculum	347.59 ± 5.60	357.45 ± 4.67	0.109	0.774
Parasubiculum	68.08 ± 2.06	70.15 ± 1.72	0.339	0.494
Fimbria	93.37 ± 2.40	99.16 ± 2.01	0.060	0.829
HATA	56.90 ± 0.88	58.87 ± 0.74	0.033a	0.862
Molecular layer	559.57 ± 5.88	581.71 ± 4.91	0.001a	0.935
Hippocampal fissure	154.28 ± 3.14	159.39 ± 2.63	0.146	0.726
Hippocampal tail	581.33 ± 9.40	593.83 ± 7.85	0.185	0.686
Right hippocampus
Total volume	3608.44 ± 33.02	3678.34 ± 27.57	0.016a	0.891
CA1	674.15 ± 7.92	687.64 ± 6.61	0.039a	0.852
CA2–3	206.66 ± 3.07	206.72 ± 2.56	0.221	0.629
CA4	242.84 ± 2.66	247.53 ± 2.22	0.016a	0.886
GC–DG	286.83 ± 3.10	292.86 ± 2.59	0.016a	0.895
Subiculum	460.93 ± 5.48	469.94 ± 4.57	0.114	0.764
Presubiculum	328.10 ± 4.63	338.09 ± 3.86	0.060	0.824
Parasubiculum	63.62 ± 1.57	65.06 ± 1.31	0.567	0.248
Fimbria	92.22 ± 2.34	97.70 ± 1.96	0.074	0.807
HATA	59.00 ± 0.91	59.35 ± 0.76	0.237	0.605
Molecular layer	579.16 ± 5.45	590.50 ± 4.55	0.016a	0.887
Hippocampal fissure	160.87 ± 3.27	165.52 ± 2.73	0.221	0.634
Hippocampal tail	614.94 ± 9.18	622.96 ± 7.66	0.207	0.659

Data are presented as mean ± SD.

^aSignificant after correction for multiple testing with the Benjamini–Hochberg false discovery rate. P values are presented after Benjamini–Hochberg correction.

FIGURE 3.

Volumes of the hippocampal subfields in AMR and control subjects. Violin plot of volumes of hippocampal subfields in AMR compared with HC, adjusted for age, sex, education, BMI, and ICV. *Significance after the Benjamini–Hochberg false discovery rate correction. *P < 0.05, **P < 0.01.

TABLE 3.

Whole thalamic and thalamic nuclei group volumes (mm³).

	HC	AMR	P	Partial η2
Left thalamic nuclei groups
Total volume	6986.80 ± 85.06	7238.96 ± 71.04	0.085	0.849
Anterior	133.42 ± 2.45	136.10 ± 2.05	0.316	0.613
Lateral	155.06 ± 2.83	156.52 ± 2.36	0.547	0.294
Ventral	2054.02 ± 25.11	2119.13 ± 20.97	0.074	0.875
Intralaminar	414.91 ± 5.57	425.48 ± 4.65	0.233	0.714
Medial	1129.34 ± 15.52	1144.04 ± 12.96	0.547	0.306
Posterior	2312.40 ± 26.24	2377.83 ± 21.91	0.074	0.840
Right thalamic nuclei groups
Total volume	6823.50 ± 85.96	7052.49 ± 71.79	0.085	0.849
Anterior	133.12 ± 2.34	134.15 ± 1.95	0.232	0.729
Lateral	145.99 ± 3.12	148.86 ± 2.60	0.414	0.509
Ventral	1987.90 ± 23.59	2022.09 ± 19.70	0.074	0.852
Intralaminar	403.78 ± 5.08	407.68 ± 4.24	0.316	0.618
Medial	1178.76 ± 15.69	1192.41 ± 13.10	0.547	0.267
Posterior	2250.75 ± 25.11	2293.43 ± 20.97	0.074	0.838

Data are presented as mean ± SD.

P values are presented after Benjamini–Hochberg correction.

FIGURE 4.

Volumes of the thalamic nuclei group in AMR and control subjects. Violin plot of volumes of thalamic nuclei group in AMR compared with HC, adjusted for age, sex, education, BMI, and ICV.

DISCUSSION

In this study, with the novel automated hippocampal segmentation approach, we compared the hippocampal subfield volumes in the AMR and HC groups. The results showed that the volumes of the left and right hippocampus, CA1, CA4, GC–DG, molecular layer, left CA2–3, and left HATA in the AMR group were larger than those in the HC group. In addition, AMR and HC had no significant difference in their thalamic subfield volumes. In the AMR group, the PSQI score was significantly lower than that in the HC group, and correlations between the hippocampal subfield volume and the PSQI score and sleep duration were not significant.

The AMR group had larger volumes of the left and right hippocampus, CA1, CA4, GC–DG, molecular layer, left CA2–3, and left HATA than the HC group. The CA1 subregion of the hippocampus is particularly vulnerable to various insults (52) and is involved in self-awareness and contextual memory retrieval (53). In addition, CA1 pyramidal neurons integrate multimodal sensory information about environmental stimuli during locomotion (54). Evidence suggests that exercise can affect the structural plasticity of the CA1 region (55). The present study also found that the left and right CA1 were larger in AMR group compared with those in the HC group. The larger CA regions in AMR group were consistent with the results of the previous studies (17,56). The CA2 plays a crucial role in the formation of social memory (57). The CA3 primarily contributes to encoding of spatial information and memory retrieval (58). In line with previous study (17), our results showed that the AMR group had larger CA2–3 volumes. By contrast, a previous study suggests that exercise intervention decreases hippocampal subfields CA2/3, subiculum, and DG (59). However, the participants were young male students, and the sample size was relatively small in that study. These factors might have an impact on the results.

The CA4 is the hilar region and is a polymorphic layer that contains different types of interneurons within the DG (60). Cells in GC–DG are important to functional neurogenesis during brain development and adulthood (61). In rodent studies, the reduced neurogenesis in the adulthood can be recovered by certain interventions, such as running (62). The larger GC–DG volume in the AMR group observed in our study was in line with this theory. These findings are in line with previous imaging studies, identifying the DG as an area where exercise-mediated neurogenesis is most pronounced (16). Lining the fissure just below the cerebrospinal fluid, molecular layer contains interneurons, connecting axons, and glial cells (63). The left and the right molecular layers were larger in AMR compared with those in HC. This was a novel finding, because molecular layer was not labeled using the previous method (17). HATA anatomically connects the hippocampus to the amygdala. Atrophy of the HATA may damage the hippocampus–amygdala pathway, affect information processing, and promote cognitive dysfunction. Although several previous studies did not assess the HATA within hippocampal subfields (17,64), our AMR group showed a larger volume in the HATA. Therefore, further research focusing on the HATA is required. Further research is needed to elucidate the detailed anatomical and functional roles in the hippocampal subfield.

A growing number of studies have established that PE promotes hippocampal neurogenesis and plays prominent role in hippocampus-dependent brain functions (65,66). Unfortunately, the adult hippocampal neurogenesis cannot be directly tracked in vivo in humans; however, angiogenesis has contributed important indirect evidence that possibly accompanied by neurogenesis in humans and rodents (67,68). In fact, exercise-related changes in hippocampal perfusion are closely linked to changes in hippocampal volume (69). In our work, simply detecting brain volume change did not reveal the nature of the underlying biological processes of PE. In the future, using multiple imaging techniques will provide a more comprehensive account of the hippocampal changes related to PE, with integrating resting-state fMRI to provide further insights into the hippocampal circuitry, diffusion tensor imaging to track changes in human hippocampal integrity, and multi-inversion time pulsed arterial spin labeling to measure hippocampal blood flow. In addition, some studies in healthy participants reported partially positive interrelations between PE, hippocampal subfield volumes, and improved cognitive functioning (16,70), but associations tended to be weak. On the contrary, one study found a negative relationship, with exercise-induced increases in the left hippocampus significantly associated with worse cognitive performance after a 6-month intervention (71). In our study, the cognitive performance did not differ significantly between AMR and HC. Although we observed larger hippocampal subfield volumes in AMR, it cannot be concluded that structural increases are causative or necessary for behavioral measure increase (72). Future longitudinal studies with larger samples are warranted to examine the direct link between exercise-induced increases in brain volume and actual cognitive benefits.

Previous evidence suggests that there is no relationship between PE and thalamus (14,73). Thus, we used thalamus as a control region. Consistent with previous findings, the total thalamus volume in the AMR group did not differ significantly from those in the HC group. Our results also showed no differences in the thalamic nuclei group volumes between the AMR and the HC groups. The effects of exercise on the brain appear to involve a number of key molecular and cellular mechanisms, including adult neurogenesis (1), brain-derived neurotrophic factor (74), insulin-like growth factor 1 (75), vascular endothelial growth factor (76), synaptogenesis and dendritic branching (77), increased cerebral blood flow (78), and increased glia cell proliferation (29).

PE is a promising nonpharmacological treatment to improve sleep and reduce fatigue in adults (79–81). Partly in line with this, our results suggest that the AMR group sleeps better compared with the HC group, but there were no significant differences in the FSS score and sleep duration between the two groups. The present study also suggests that there may be no relationship between hippocampal subfield volume and subjective sleep quality in running population. This finding is partly consistent with previous research (35,36) in which no relationship between subjective sleep quality assessed by PSQI and hippocampal volume was found in middle-age to older adults. Conversely, lower hippocampal volume is reportedly correlated with worse sleep or fatigue in healthy older adults (33,34). Given these mixed observations, the relationship between subjective sleep quality and hippocampal volume remains to be determined. Various aspects of hippocampus, beyond hippocampal subfield volume, such as structural network, functional connectivity, and cerebral blood flow, need to be investigated in the future to enrich our knowledge about the relationship between sleep and hippocampus.

Strengths and limitations

One of the major strengths of the present study was a relatively large sample of AMR and HC (n = 125). The present study also used advanced tissue segmentation protocols to detect hippocampal subfield volume. In addition, AMR show better sleep quality than HC. Our results add to the current body of knowledge by suggesting that exercise may influence the hippocampal substructure. There are several limitations to this study that warrant caution when interpreting the results. First, cross-sectional design of the study limits interpreting the directionality of the current results. Further longitudinal and interventional studies will be necessary to clarify the role of exercise in modulating the hippocampal substructures. Second, there are limitations of using 3-T scanners for segmentation of the hippocampus compared with 7-T scanners, which are far more accurate. Future studies using 7-T scanners to obtain high-resolution anatomical delineation of hippocampal subfield are needed. Lastly, we only focused on AMR, and there is a lack of investigation on how exercise (i.e., frequency, intensity, type) influences the effects on the hippocampal structure. Future research is needed to determine which kind of PE principle will best relate to changes of hippocampal structure. Furthermore, other unknown or unmeasured factors that may have influenced both PE and brain volumes could have led to residual confounding.

CONCLUSIONS

In this study, the volumes of the left and right hippocampus, CA1, CA4, GC–DG, molecular layer, left CA2–3, and left HATA in the AMR were larger than those in the HC. The AMR had significantly better sleep than HC, manifested as with lower score of PSQI. AMR and HC had no significant difference in their sleep duration. Besides, AMR group showed no significant correlation between the hippocampal subfield volume and the PSQI as well as sleep duration. These findings should be further investigated in longitudinal studies. Furthermore, despite the need for cautionary interpretation of our cross-sectional results, we believe that they hold promise for a deeper understanding of the effects of PE and the potential application of PE as a nonpharmacological approach for enhancing the structure and function of the hippocampus.