The hippocampal formation as part of the limbic system is located in the medial temporal lobe. It is highly involved in processes of episodic memory formation (Tulving and Markowitsch, 1998; Van Petten, 2004) and spatial navigation (O’Keefe, 1990; Maguire et al., 2000). Both cognitive dimensions are especially vulnerable to performance loss in late adulthood (cf. Hedden and Gabrieli, 2004). Recently, the hippocampus has also been associated with motor sequence consolidation (Albouy et al., 2008). Over the adult lifespan, on average, hippocampal volume shrinks about 0.86% per year, but this development is highly non-linear (Raz et al., 2004b). Whereas for adults below the age of 50 annual hippocampal volume reductions of only 0.51% were observed, adults above the age of 50 revealed a much steeper annual volume decline of 1.18% (Raz et al., 2004b). These volume reductions are mainly attributed to reductions in the neuropil part of the brain structure and – to a smaller extend – to cell body shrinkage and changes in vascularization (Thomas et al., 2012).

The hippocampus seems to be especially sensitive to cardiovascular activity. This has been demonstrated in research with rodents (e.g., van Praag et al., 1999, 2005; Kronenberg et al., 2006; Pereira et al., 2007; Fabel and Kempermann, 2008; Kempermann et al., 2010) and humans (e.g., Pereira et al., 2007; Erickson et al., 2009, 2010, 2011). Rodent studies revealed two main mechanisms to be responsible for structural benefits after cardiovascular training: the enhancement of capillary density (angiogenesis; Black et al., 1990) and the generation of new neurons (neurogenesis; van Praag et al., 1999), whereby the latter is specifically restricted to the dentate gyrus. Typically, old rodents and humans reveal a substantial decline in hippocampal neurogenesis (approx. 10-fold for rodents, cf. Ben Abdallah et al., 2010; and approx. fourfold for humans, cf. Spalding et al., 2013). In old rats the rate of decline in neurogenesis could be halved by 1 month of cardiovascular training and this reduced rate of decline was associated with better spatial learning abilities (van Praag et al., 2005; Kronenberg et al., 2006). These findings indicate a beneficial influence of cardiovascular activity on hippocampal volume and function even in older ages.

In a human sample aged 59–81 years, larger hippocampal volume mediated the relationship between better cardiovascular fitness and higher spatial memory performance (Erickson et al., 2009); and 12 months of cardiovascular training were associated with an increase in hippocampal volume by about 2% in older adults (Erickson et al., 2011). This increase in hippocampal volume was associated with improvement of cardiovascular fitness (VO₂ max) and spatial memory performance, respectively (Erickson et al., 2011).

Physical activity, however, does not only comprise exercises that require metabolic and energetic processes, like cardiovascular training, but also motor-demanding exercises that require perceptual and higher-level cognitive information processing. This is, for example, true for coordinative exercises, like balance, eye–hand coordination or leg–arm coordination tasks (cf. Voelcker-Rehage and Niemann, 2013). Research on animals suggests that effects of motor-demanding activity on brain metabolism differ from those of cardiovascular activity (Black et al., 1990; Anderson et al., 1994). A motor-demanding training based on coordinative skills (e.g., crossing obstacles, balancing on ropes, etc.) facilitated synaptic growth and re-construction (synaptogenesis; Black et al., 1990; Seo et al., 2010). It also led to a higher degree of survival of newly generated cells than cardiovascular exercise (cf. above Curlik et al., 2013). Results from studies with older adults point in a similar direction. They revealed that motor fitness (e.g., action speed, reaction speed, balance) and coordination training facilitate brain function (particularly in those areas related to visual–spatial processing) and cognitive performance in older adults (Voelcker-Rehage et al., 2010, 2011). The associations of motor fitness with and the influences of coordination training on brain volume and specifically on hippocampal volume in older adults have not been investigated so far.

In a first step, analyzing a cross-sectional sample of older adults, we analyzed the relationship between metabolic/cardiovascular fitness and the volume of the hippocampus (Erickson et al., 2009). Moreover, for the first time, we tested the hypothesis that motor fitness is also associated with hippocampal volume in older adults. Based on our own findings revealing a positive association of motor fitness and brain function, we assumed to find a positive relationship between motor fitness and hippocampal volume as well. This seems particularly reasonable as the hippocampus is involved in spatial navigation and motor sequence consolidation (O’Keefe, 1990; Albouy et al., 2008), two processes probably associated with motor fitness.

In the second step, we analyzed interventional data from a subsample of the same dataset to investigate the effects of a 12-month cardiovascular or coordination training on volume of the hippocampus. We expected not only cardiovascular training to positively influence hippocampal volume (as shown by Erickson et al., 2011), but also coordination training. This assumption is based on findings in rodents suggesting that motor activity benefits synaptogenesis in activated brain regions (e.g., Black et al., 1990).

Ninety-two older adults between 62 and 79 years of age were recruited to participate in the Old Age on the Move project. Only participants were included who did not report any history of cardiovascular disease, neurological disorder, or motor or cognitive restriction (e.g., metal implants, a score of less than 27 in the Mini Mental Status Examination, MMSE; Folstein et al., 1975). One participant had to be excluded due to cognitive impairment (MMSE <27, cf. O’Bryant et al., 2008 for cut-off level, remaining participants N = 91). All subjects participated voluntarily in the study and provided written informed consent to the procedures of the study, which was approved by the ethics committee of the German Psychological Society. The study conformed to the Code of Ethics of the World Medical Association (Declaration of Helsinki).

Participants were tested for normal vision (Freiburg Visual Acuity Test; Bach, 2007) and hearing (simultaneous auditory thresholds at multiple frequencies for both ears; Presentation^® software, Neurobehavioral Systems, Albany, Canada; Yund, 2003) and were given a demographics and health questionnaire to determine characteristics of the sample (cf. Table 1).

At baseline (t1), as well as after six (t2) and after 12 months (t3), each participant completed the fitness assessment in one laboratory session of approximately 2 h. All tasks were administered in a fixed order. On another day but within 1 week, participants underwent the MRI scanning at a 3-Tesla Allegra Scanner. Using a MPRAGE sequence with 1 mm × 1 mm × 1 mm isotropic resolution, a TR of 2300 ms, and 160 slices, T1-weighted anatomical scans were collected. Assessment of the structural MR images was part of a larger MRI testing protocol, the results of which are reported elsewhere (cf. Godde and Voelcker-Rehage, 2010; Voelcker-Rehage et al., 2010, 2011).

The T1-weighted MR images were realigned to correct for head tilt, pitch, and rotation passing through the anterior and posterior commissure line (AC-PC alignment) by using Brain Voyager (Brain Innovation B.V., Netherlands). In the next step, hippocampal volumes were measured on coronal sections by manual tracing using Analyze (Analyze Version 10.0, Mayo Clinic, Rochester, MN, USA). Hippocampal volume of the right and left hemisphere as well as intracranial volume (ICV) were defined and traced according to the rules described below (original rules cf. Raz et al., 2003 for the ICV; Raz et al., 2004b for the hippocampus). Then, volumes were computed by multiplying the sum of the areas within each slice by slice thickness. Finally, individual (i) raw volumes were adjusted using the following formula: Volume_adj_i = volume_raw_i - b (ICV_i - ICV_samplemean; Raz et al., 2003) to correct for inter-individual ICV differences, where b is the regression slope of volume_raw_i on ICV. Reliability of volume determination was secured by calculating the intraclass correlation coefficient ICC (2; for more information on the tracing reliability process see Raz et al., 2004a). All ICC(2) values were above the critical level of.90.

Intracranial volume was estimated by tracing its volume on every 10th image slice (approx. 16 slices) between the first slice following the orbits and the last slice on which brain tissue was visible. This method has shown to save a lot of time (in comparison with tracing every slice) by only having negligible reductions in reliability (Eritaia et al., 2000). Tracing started at the right side of the head proceeding clockwise while following the inner table of the skull until the homologous mark on the left side of the head was reached. At this point, tracing was stopped and automatically a straight line back to the origin was drawn by the program (cf. Raz et al., 2003).

Hippocampal volume was estimated on every slice from the first slice where the mammillary bodies appear most bulbous. The posterior boundary was determined by looking for the slice that cuts through the fornices as they rise after leaving the fimbria of the hippocampus and then using the next slice as the last one (cf. Raz et al., 2004b). The volume of the hippocampus was estimated from 22 to 32 coronal sections. Further demarcation lines were identical as described by Raz et al. (2004a; cf. Figure 1 for example images).

Linear regression analysis revealed that fitness dimensions and their interaction as well as age and sex together explained 10.9% of variance in hippocampal volume (see Table 2). Motor fitness uniquely explained 7% of variance in hippocampal volume, whereas the other variables did not further explain variance in hippocampal volume significantly (shared variance 1.5%; cf. additionally Figure 2 for a scatterplot of hippocampal volume against motor fitness (Figure 2A) and metabolic fitness (Figure 2B). The second regression analysis with cardiovascular fitness instead of metabolic fitness revealed comparable results explaining 13.1% of total variance in hippocampal volume (see Table 3 and Figure 2C for a scatterplot of hippocampal volume against cardiovascular fitness). Again, motor fitness uniquely explained 9% of variance in hippocampal volume and none of the other variables did significantly explain additional variance.

Results separately for both hemispheres are similar as for total hippocampal volume (cf. Tables 2 and 3; for the left hippocampus, all variables accounted for 11.5% of variance in volume; for the right hemisphere the explained variance was 7.7%. Motor fitness uniquely explained 5.7% of the variance in left hippocampal volume and 6% in right hemisphere. Neither metabolic fitness, cardiovascular fitness or the respective interaction terms with motor fitness nor the control variables age and sex did explain additional variance.

To compare efficiency of the specific training programs we analyzed training effects on cardiovascular and motor fitness, respectively, for all three study groups (cf. Figure 3A and Table 4). Overall, for cardiovascular fitness we found a significant SESSION × GROUP interaction [F(3.55,81.54) = 3.68, p = 0.01, η² = 0.14], but no SESSION [F(1.77,81.54) = 0.93, p = 0.40, η² = 0.02] or GROUP effect [F(2,46) = 1.15, p = 0.33, η² = 0.05]. Separate analyses for the three groups revealed that cardiovascular fitness in the cardiovascular training group significantly increased by 13.65% across the 12-month intervention [F(2,45) = 5.58, p = 0.01, η² = 0.20] with post hoc pairwise comparisons showing significant improvements only from t2 to t3 (p < 0.01), but not from t1 to t2. No significant changes could be observed in the coordination training group [1.98% increase; F(2,45) = 0.16, p = 0.88, η² = 0.01] and the control group [5.65% decrease; F(2,45) = 0.98, p = 0.38, η² = 0.04].

For the overall change in motor tasks, we found a main effect of SESSION [F(1.70,78.25) = 31.38, p < 0.001, η² = 0.41] and TASK [F(1.37,63.16) = 1538.42, p < 0.001, η² = 0.97] and a significant SESSION × TASK interaction [F(2.40,110.48) = 10.47, p < 0.001, η² = 0.19]. There was no GROUP effect [F(2,46) = 0.07, p = 0.93, η² = 0.00] and no GROUP × SESSION interaction [F(3.40,78.25) = 1.43, p = 0.24, η² = 0.06], no GROUP × TASK interaction [F(2.75,63.16) = 1.78, p = 0.16, η² = 0.07], and no GROUP × SESSION × TASK interaction [F(4.80,110.48) = 1.31, p = 0.24, η² = 0.05]. Calculating separate analyses for the study groups showed that after the 12-month intervention all study groups significantly improved their performance in the three selected motor fitness tasks [cardiovascular group: F(2,45) = 11. 22, p < 0.001, η² = 0.33; coordination group: F[2,45] = 18.44, p < 0.001, η² = 0.45; control group: F(2,45) = 4.03, p = 0.03, η² = 0.15]. Analyses separately for the three tasks action speed, reaction speed and balance revealed a different development for motor fitness indicators within the three groups (cf. Figures 3B–D). The cardiovascular training group improved their action speed performance by 42% [F(2,45) = 12.62, p < 0.001, η² = 0.36; with significant improvements from t1 to t2 (p < 0.001), but not t2 to t3]. Similarly, the coordination training group revealed an increase in action speed performance of 21.1% [F(2,45) = 8.35, p < 0.001, η² = 0.27; with significant improvements from t1 to t2 (p < 0.001), but not from t2 to t3] and in addition an 21.6 % improvement of performance in one-leg-stand with eyes open [F(2,45) = 5.25, p = 0.01, η² = 0.19; with significant improvements from t1 to t2 (p < 0.01), but not from t2 to t3]. Changes in reaction speed did not reach significance level (cardiovascular group: 3.6% improvement [F(2,45) = 1.31, p = 0.28, η² = 0.06]; coordination group: 8.4% improvement [F(2,45) = 0.88, p = 0.42, η² = 0.59)]. The control group did not show any significant changes across the 12-month intervention period in these three tasks.

Using a linear regression analysis with total hippocampal volume at t3 as dependent variable, only change in cardiovascular fitness (VO₂ peak ml/kg) significantly explained variance (9.3%, p = 0.04) in hippocampal volume (cf. Table 7). When analyzing both hemispheres separately, again, only VO₂ peak significantly explained 12.1% (p = 0.02) of variance in volume of the left hippocampus (cf. Table 7) – the side of the hippocampus that revealed a volume increase in the cardiovascular training group. No such effect was found for the right hemisphere.

In line with our hypothesis for the cross-sectional analyses, motor fitness was significantly associated with hippocampal volume accounting for 7% of explained variance. To our knowledge, this is the first study that described a relationship between hippocampal volume and motor fitness, i.e., the component of fitness that requires perceptual and higher-level cognitive processes. Herewith we are adding significantly to previous functional brain data revealing a positive association between motor fitness and functional brain activation in frontal and parietal areas accompanied by better performance in perceptual speed and executive functions (Voelcker-Rehage et al., 2010).

We could not demonstrate a relationship between hippocampal volume and metabolic fitness (comprising cardiovascular fitness and grip strength) or cardiovascular fitness alone. This is in line with cross-sectional studies by Honea et al. (2009), Flöel et al. (2010), and Smith et al. (2011) who also did not find correlations between volume of the medial temporal lobe area, or more specifically the hippocampus, and cardiovascular fitness in different samples of healthy older adults (50+ years of age). However, a relationship between those factors was assumed by Erickson et al. (2009) as well as by Szabo et al. (2011) who both demonstrated a positive association between higher cardiovascular fitness and larger hippocampal volume in samples of healthy American older adults. Whereas our sample did not differ from theirs in age and presumably the level of cardiovascular fitness, differences in data analysis might account for diverging results: it has been shown that using (semi-)automated computer-programs (e.g., VBM, FIRST, cf. Erickson et al., 2009; Szabo et al., 2011) might reveal different results regarding brain volume than tracing predetermined brain regions manually (cf. Kennedy et al., 2009 for a comparison of different analyzing methods). Interestingly, Honea et al. (2009) confirmed a correlation between medial temporal lobe and cardiovascular fitness in patients suffering from early stages of Alzheimer‘s disease (Honea et al., 2009). Overall, although associations between hippocampal volume and cardiovascular fitness might be plausible, results are not such consistent and might depend on reasons mentioned above or on moderating variables still not identified.

Data of the 12-month intervention study revealed a positive effect on total hippocampal volume (sum scores of right and left hemisphere) for both types of exercise, cardiovascular and coordination training. Interestingly, our findings for cardiovascular training mirror the results of Erickson et al. (2011) who also found an increase in hippocampal volume after 12 months of cardiovascular training, although cross-sectional results were different (see above). Possibly, training effects in a controlled trial are more reliable and stronger than cross-sectional associations of hippocampal volume and cardiovascular fitness that are influenced to a higher degree by sample characteristics and further uncontrolled factors.

Whereas Erickson et al. (2011) revealed an increase in left and right hippocampal volume by 2.12% and 1.97%, respectively, we found an even higher (twice as high) volume increase in the left hippocampus of 4.22% and an increase of 2.98% for the right hippocampus. The increase in the right hippocampus was, however, not statistically significant in our cardiovascular training group, probably due to the high variability within our sample. The higher level of volume increase, particularly for the left hippocampus, in our study might be explained by the magnitude of change in VO₂ peak performance. In our sample, participants of the cardiovascular training group improved by 13.65%, whereas Erickson et al. (2011) found increases of cardiovascular fitness only by 7.78%. As Erickson et al. (2011) and our sample support the assumption that there might be an association between change in VO₂ peak and hippocampal volume at post-test, one might suggest that there is a linear relationship between both variables.

In addition, for the first time, we showed that not only changes in cardiovascular fitness seem to be associated with hippocampal volume in older adults, but also a 12-month coordination training led to changes in total hippocampal volume. In contrast to cardiovascular training, these changes were more pronounced in the right hippocampus (right: 3.91%; left: 1.78% non-significant change). Particularly the right hippocampus is known to be involved in spatial memory processes (Smith and Milner, 1981; Nunn et al., 1999; Burgess et al., 2002). As coordinative exercises to a high degree rely on and practice spatial orientation, this might be one explanation for our finding.

Studies in young adults have shown that short-term motor learning (e.g., performing a balance task, learning to write with the left non-dominant hand, etc.) seems to induce very rapid volume changes (increases and decreases within two to six training sessions) in training-related brain regions (Taubert et al., 2010; Gryga et al., 2012; Hamzei et al., 2012). We could not reveal changes in hippocampal volume after 6 months of training, but found a significant increase in volume after the second half of the intervention. These results might indicate that rapid volume changes might fluctuate in the beginning of training (Taubert et al., 2010; Gryga et al., 2012) and, since patterns of plastic changes might differ in regard to training duration (cf. Hamzei et al., 2012), are more stabilized after the continued training of complex motor tasks. In contrast to the cardiovascular training, we could not identify changes in specific motor tasks to cause changes in hippocampal volume. One reason might be that the changes in the motor system due to coordination training were not adequately measured by the selected motor tests.

In addition, in animal research, so far, few studies have compared effects of cardiovascular training and motor-demanding activities (corresponding to coordination training in humans, cf. Voelcker-Rehage and Niemann, 2013 for a further discussion). Greenough and colleagues (Black et al., 1990; Isaacs et al., 1992; Anderson et al., 1994) provided evidence that cardiovascular training increases capillary density in rodents, while acrobatic training in rodents induces the generation of new synapses and glia cells as well as dendritic hypertrophy (cf. Kleim et al., 1996, 1997, 1998; Seo et al., 2010). The motor-based training initiated blood vessel growth only to maintain the diffusion distance (Isaacs et al., 1992). Diverging underlying mechanisms might also explain the different time courses of volume change that we observed after the two types of exercise training. While the cardiovascular training group showed almost linear increase in hippocampal volume, hippocampal volume of the coordination training group increased from month 6 to 12 of the intervention only (cf. Figure 4). A recent review (Thomas et al., 2012) provided time courses for the different components of brain plasticity, i.e., angiogenesis and capillary density, neurogenesis, astrocyte volume, and neuropil volume. Following them, angiogenesis as well as neurogenesis show a rapid time course. For example, in mice capillary density is already enhanced after three days of increased aerobic activity, but also returns to baseline within 24 h of sedentary behavior (Van der Borght et al., 2009). Similarly, in rats astrocyte changes occur rapidly but return to baseline volume as soon as activity is stopped (Kleim et al., 2007). In contrast, again in rats changes in the neuropil are more stable and might last for several weeks (Kleim et al., 1997, 4 weeks after removed from enriched environment; see also a review by Markham and Greenough, 2004). Although to date, it is not terminally resolved whether the time course in brain plasticity is transferable from studies in rodents to humans, these findings from the animal literature let us suggest that also different underlying mechanisms might be responsible for the volume changes accompanied by different time courses in our sample of healthy older adults.

In addition, animal studies focusing on neurogenesis in the hippocampus support our diverging findings for cardiovascular and coordination training on hippocampal volume. They indicate that different types of activity stimulate the generation of new neurons in different ways (van Praag et al., 1999). In rodents, both types go in line with an enhancement of cell numbers: cardiovascular activity induces the proliferation of newborn cells in the hippocampus to a large extent. In contrast, an “enriched environment” condition consisting of a complex combination of physical, motor and social stimulation also promoted the proliferation of new born cells (approx. 50% of cardiovascular training effects) leading, in the long run, to a higher degree of integration and survival of the neurons (van Praag et al., 1999; Curlik et al., 2013). This finding could also add to our results in humans after 6 months of training, where we found changes in the cardiovascular group but not the coordination group and might explain why we found comparable volume increases after 12 months of intervention for both types of training. However, as we do not have follow up data after cessation of the exercise training, we, unfortunately, do not have any information about how long exercise changes last.