Being active might be your best health ally. Current scientific literature states that an active lifestyle might have a positive impact on most hallmarks of aging, especially for brain and cognitive health (Ji et al., 2021). Behavioral interventions such as aerobic exercise (AE) and cognitive training programs are promising approaches to maintain and enhance cognition (Phillips, 2017; Sprague et al., 2019). Systematic reviews generally agree on the small-to-moderate effect size and consistent effects of AE on executive function, attention and speed (Colcombe and Kramer, 2003; Barha et al., 2017; Northey et al., 2018; Stillman et al., 2020) and the benefits of cognitive training in the trained domains but limited transfer to non-trained domains (Sprague et al., 2019). Combining both interventions showed promising results about the potential for greater effects on cognition (ten Brinke et al., 2020) yet recent systematic reviews reported that the extent of the combined effects were not significantly different from the individual programs (Gavelin et al., 2020; Guo et al., 2020). Studies including these approaches are examining biological mechanisms that could inform associations in prior studies. Identifying the molecular/cellular and brain changes related to exercise and cognitive training benefits, applied individually or in combination, is key to better understanding and prescribing these interventions as well as elucidating which combination of strategies would be more successful.

AE-related cognitive benefits are the result of multiple biological and psychosocial changes which were organized in a three-level model by Stillman et al. (2016). Molecular and cellular changes are described at Level 1 and include key factors involved in neuroplasticity, angiogenesis, metabolism and inflammation. For example, brain derived neurotrophic factor (BDNF) seems sensitive to AE (Knaepen et al., 2010). While available evidence consistently reported increased BDNF levels after an acute bout of exercise, the effect of regular exercise on resting BDNF remains unclear (Walsh and Tschakovsky, 2018). Initial systematic reviews and meta-analyses suggested that exercise training led to elevated peripheral BDNF (Knaepen et al., 2010; Coelho et al., 2013; Szuhany et al., 2015; Dinoff et al., 2016), but these results were driven by trials including clinical populations (Walsh and Tschakovsky, 2018). Recent reviews including healthy samples have reported non-significant effects of low to moderate AE on BDNF levels (Marinus et al., 2019) and conclude that evidence in human studies is still insufficient for making strong conclusions (Loprinzi, 2019). Other growth factors such as insulin growth factor (IGF) and vascular endothelial growth factor (VEGF) have consistently been related to endothelial cell proliferation and vessel growth as well as enhanced neurogenesis in the hippocampus after exercise due to their interaction with BDNF (Cotman et al., 2007; Stimpson et al., 2018). Molecular markers modulated by IGF and VEGF, such as stromal cell derived factor 1 (SDF1) or intercellular adhesion molecule 1 (ICAM-1) were suggested to be involved in the promotion of angiogenesis in the brain after exercise based on animal studies (Stimpson et al., 2018). Significantly decreased levels of ICAM-1 related to physical activity were found in patients with cardiovascular risk factors (Palmefors et al., 2014) but evidence in healthy samples is still inconclusive with some (Abd El-Kader et al., 2019) showing significant and others (Mohammadi et al., 2018) non-significant changes in post-test vs. pre-test comparisons. In healthy young men, SDF1 levels increased in response to acute exercise with an increase of endothelial cells in a 24 h follow-up (Chang et al., 2015). However, the effect of exercise programs on these markers in humans is still insufficient for drawing clear conclusions (Palmefors et al., 2014). Another growth factor potentially involved in this cascade of changes is the hepatocyte growth factor (HGF) which is a hepatokine primarily linked to liver regeneration. HGF has an anti-inflammatory role in adipose tissue and is linked to insulin resistance and diabetes (Oliveira et al., 2018). However, the effect of exercise on HGF levels has been scarcely addressed. Yasuda et al. (2004) found increased HGF levels after acute exercise which were negatively related to cardiovascular fitness (CRF) in patients after acute myocardial infarction. Exercise is known to also induce antioxidant and anti-inflammatory actions by modulating cytokines and oxidative stress factors on adipose tissue, body muscles and immune system (Sallam and Laher, 2016). Recent meta-analyses reported significant benefits on the pro-inflammatory markers in healthy populations after AE interventions (Monteiro-Junior et al., 2018; Zheng et al., 2019). However, evidence is still unclear about how exercise might influence each one of these markers. For example, Monteiro-Junior et al. (2018) and Zheng et al. (2019) agree on reduced levels of C-reactive protein and interleukin 6, but only Zheng et al. (2019) found significant effects for decreased TNF-α.

Changes at the molecular level are involved in the structural and functional brain benefits related to exercise as described at Level 2 by Stillman et al. (2016). In relation to brain volume, systematic reviews reported that AE training could be effective to prevent brain volume loss, increase gray matter volume (Erickson et al., 2014) and have a positive but small impact on global white matter volume (Sexton et al., 2016). Most of the research on this topic has focused on the impact of exercise on the hippocampus showing significant positive effects specifically on left hippocampal volume (Firth et al., 2018), specifically in older populations (>65 years) and in interventions that lasted over 24 weeks (Wilckens et al., 2020). However, there are still inconsistencies among trials including healthy samples. For example, increased local gray matter volumes in prefrontal (Colcombe et al., 2006; Ruscheweyh et al., 2011), cingulate (Ruscheweyh et al., 2011) and temporal cortices (Colcombe et al., 2006) and hippocampus (Erickson et al., 2011; Rosano et al., 2017) have been observed after long-term AE interventions, whereas non-significant changes in these same areas have been observed in trials lasting only 12 weeks (Maass et al., 2015; Matura et al., 2017; Sexton et al., 2020). This fact is in accordance with evidence highlighting how FITT-VP (Frequency, Intensity, Time, Type, Volume and Progression) parameters of exercise programs might be modulating the effect of the intervention on biomarkers at either at Level 1 or 2 (Cabral et al., 2019; Chen et al., 2020). Moreover, current literature suggests that individual characteristics of participants such as age, sex or genetics (Barha et al., 2019; Stillman et al., 2020) interplay with the intervention modulating the cognitive benefits related to exercise.

The Hebb (1949) principle has been used to generate hypotheses about the mechanisms involved in cognitive benefits related to cognitive training. Many authors have suggested that the regular practice of a cognitive task would repeatedly and simultaneously activate a group of neurons and, in turn, strengthen synaptic connections (Taya et al., 2015; ten Brinke et al., 2017; Gates et al., 2020). Animal studies using an “enriched environment” revealed increased long-term potentiation and enhanced BDNF levels in the hippocampus which led many authors to hypothesize a similar pattern in humans (Valenzuela and Sachdev, 2009). While cognitive training induces a significant increase of BDNF in patients with risk of dementia (Damirchi et al., 2017), evidence is still unclear if this effect occurs in healthy populations. There is evidence showing significant increases of BDNF after cognitive training programs lasting 5 (Ledreux et al., 2019) and 7 weeks (Rahe et al., 2015) and, at the same time, other trials did not find significant changes in BDNF after 10 (Küster et al., 2017) and 12 weeks (Tarassova et al., 2020). To our knowledge, other molecular and cellular mechanisms related to cognitive training effects have been scarcely addressed in healthy or clinical human populations. However, based on the overlapping effects of BDNF, IGF, and VEGF in the promotion of neuronal growth in the brain (Cotman et al., 2007), it could be suggested that these growth factors and other related molecular factors such SDF1 might also mediate cognitive training effects. For example, given the negative effect of pro-inflammatory cytokines on IGF signaling, which may alter protein synthesis in the brain (Cotman et al., 2007), we might expect a complex interaction between multiple factors at this level. Szabó et al. (2014) found that the number of stressful events and nuclear factor-jB, which is related to pro-inflammatory cytokines, predicted the increase of the right hippocampus after a videogame-based cognitive training approach. Therefore, understanding cognitive training related changes occurring at the molecular and cellular level could help to interpret any observed benefits in the structure and function of the brain. Few studies have published structural brain changes associated with cognitive interventions in healthy older adults (Belleville and Bherer, 2012). A systematic review (ten Brinke et al., 2017) reported mixed results from only three trials reporting changes in brain volume in healthy older adults. Lampit et al. (2015) found an increase in grey matter density in the right post-central gyrus in the cognitive training group compared to an active control group. However, Heinzel et al. (2014) and Antonenko et al. (2016) found no significant changes in grey matter and hippocampal volume, respectively. As suggested above, the heterogeneity of the findings at a microscopic and macroscopic level could be related to parameters of the training program and sample characteristics (Taya et al., 2015; Gates et al., 2020). The cascade of molecular mechanisms of cognitive training-induced neuroplasticity in healthy older adults remains understudied and the research on structural brain changes is still limited (ten Brinke et al., 2017; Gates et al., 2020).

Current research has focused on whether the positive impact of exercise on growth factors, inflammatory profile and brain structure might facilitate the neuroprotective effect of cognitive training when they are combined (Joubert and Chainay, 2018). In relation with BDNF, Rahe et al. (2015) found increased peripheral BDNF from pre to post-testing when cognitive training was applied alone or in combination with physical training in a 7-week program. Anderson-Hanley et al. (2012) reported that 3 months of cybercycling that included a cognitive component induced greater changes in BDNF levels compared to traditional cycling. The effects of a combined training (COMB) program on markers of inflammation have been studied in mice showing decreased TNF-α in the hippocampus. Related to effects on brain structure, Lövdén et al. (2012) reported that healthy older men engaging in a spatial navigation task (navigation task + walking) for 4 months maintained stable hippocampal volumes after the intervention compared to a control group which showed volume decrements consistent with age-related decline. However, a 7-month program of combined cognitive and AE training in a sample with mild cognitive impairment (MCI) found no effect on grey matter loss (Train the Brain Consortium, 2017). Therefore, despite the hypothesis of the potential additive effects of both interventions on neurobiological measures, there is a great degree of mixed evidence and more research in humans is needed.

Projecte Moviment is a randomized controlled trial about the effect of a high-frequency (5 days per week) short-term (12 weeks) program of AE, computerized cognitive training (CCT) and their combination in healthy physically inactive older adults (Castells-Sánchez et al., 2019). The observed changes on cognition, psychological status and physical activity outcomes have been published in Roig-Coll et al. (2020). In this study, we aim to examine the effect of the interventions on BDNF levels, markers of inflammation and volume changes in relevant brain areas compared to healthy controls. Secondly, we aim to test whether significant changes in physical activity outcomes are related to changes in molecular markers and brain volume. Finally, we aim to assess the moderating role of sex and age on molecular and brain volume changes and the possibility that changes in molecular and brain volume outcomes mediate the relationship between the intervention and cognitive benefits.

Correspondingly, we hypothesize that: (1) Intervention of physical activity, alone or combined with computerized cognitive training, would result in lower levels of inflammatory markers, increased BDNF, and greater brain volume. Computerized cognitive training, alone or combined with physical activity, would result in increased BDNF levels, and greater brain volume. (2) Sex and age would have a significant moderating role in the effects of the intervention on Level 1 and 2 biomarkers. (3) Levels 1 and 2 biomarkers would mediate significant intervention- related changes on cognition.

In this paper we report molecular and brain volume changes in the Projecte Moviment trial, which studied the effects of AE, CCT and their combination in healthy inactive late-middle-aged adults compared to a waitlist control group. Cognitive changes in Executive Function and Attention-Speed in the AE group and in Attention-Speed in the COMB group were previously published (Roig-Coll et al., 2020) and led us to address other micro and macroscopic correlates that might be involved in these interventions. We previously found that regular exercise was related to the inflammatory profile, brain volume and cognition in a cross-sectional sample (Castells-Sánchez et al., 2020, 2021).

In the present trial, baseline-to-follow-up changes in Level 1 biomarkers, including BDNF, were not significant within either group or in comparison with those in the control group, only changes in Level 1 biomarkers, such as ICAM-1 and SDF1-α, correlated with physical activity outcomes in groups including physical exercise (AE and COMB) indicating their relevance. At Level 2, brain volume biomarkers, the CCT group showed a significant intervention-related increase in the volume of the precuneus.

Participants in the AE group did not show significant within-group pre to post-test changes nor significant changes in comparison with the control group for any of the molecular correlates or targeted brain volume areas after 12 weeks in a 5 days per week, 45 min a day brisk walking program. The intervention-related results are in accordance with recent evidence stating that short-term trials typically do not report significant changes in neurotrophic factors or structural brain changes although they do in brain connectivity markers (Cabral et al., 2019; Wilckens et al., 2020). These results suggest that despite an increased frequency of activity during a 12-week program, the duration of the program might be more relevant for these molecular and structural brain changes since training programs lasting on the order of 6–12 months reported structural brain benefits (Colcombe et al., 2006; Erickson et al., 2011; Ruscheweyh et al., 2011). Therefore, our results suggest that the parameters of the activity, such as duration and intensity, are critical aspects of an exercise intervention. Another issue that might explain our results is that the sample was healthy and aged 50–70 years. As suggested by Erickson et al. (2014), the positive effects of exercise on gray matter volume might be significant in older age ranges in which there are greater brain volume losses. This suggestion could be extended to samples with pathology in which evidence tends to be more consistently significant (Wilckens et al., 2020).

Participants in the CCT group did not show significant changes on BDNF and, as expected, for any of the other targeted biomarkers at Level 1. The magnitude of the changes in BDNF might not be detectable yet over this duration and might require more prolonged exposure to cognitive training to induce a significant change. In addition, the parameters of the CCT program, the lack of pathology and age of our participants might also be playing a role as described for the AE interventions. Regarding markers at Level 2, the CCT group showed a significant intervention-related increase in the volume of the precuneus and compared to the change observed in the control group. These findings suggest that engaging in a multimodal CCT for 45 min a day, 5 days a week for 12 weeks might increase the volume of the precuneus which has a fundamental role in cognition: it is highly involved in the integration of tasks, visuo-spatial imagery, episodic memory retrieval as well as self-referential processing (Cavanna and Trimble, 2006). This result highlights current research challenges in this field since participants in the CCT group did not show significant changes in the cognitive assessments compared to the control group (Roig-Coll et al., 2020) in accordance with other trials of cognitive training (Boujut et al., 2020). The possible neuroprotective effects of a cognitive training program should be addressed by examining molecular, brain and cognitive markers given the challenge of detecting far transfer effects. We also found that change in the frontal lobe was also significant in the CCT group compared to the control group although neither participants in the CCT group nor the control group showed significant increase or loss in this area. In this case, we found significant results in the regression model but not when we performed the pre-to post t-test to see where the difference was and in which direction. These results indicate that there were changes with opposite directions in the two groups (i.e., CCT and control groups). Although these changes themselves were not significantly different from zero (i.e., non-significant pre-to-post change), they were significantly different across the two groups. The sufficiently large sample size provides considerable safeguard against selection bias. One can speculate, however, that our results indicate a slight, potentially age-related, change in the control group, against which the intervention provides protection.

Our study also shows that combining brisk walking and multimodal CCT for 12 weeks, 5 days per week in bouts of 45 min for each did not lead to greater benefits in the molecular biomarkers and brain volume areas analyzed pre to post-test and compared to the control group. As suggested for each intervention individually, parameters of the activity or characteristics of the sample might be responsible, though the scarcity of studies and high heterogeneity in the intervention characteristics (Anderson-Hanley et al., 2012; Lövdén et al., 2012; Rahe et al., 2015) make a challenge to draw firm conclusions. Moreover, determining whether a sequential or simultaneous strategy is best for cognitive benefits is still a controversy (Guo et al., 2020). Our findings might relate to the importance of the scheduled order of the interventions, specifically when interventions are applied simultaneously (Gheysen et al., 2018; Guo et al., 2020). Enhanced effects have been reported when exercising before the cognitive training suggesting that exercise facilitates an environment that enhances the effect of cognitive training.

We also found that significant changes in physical activity outcomes (amount of PA and CRF), found in the AE and COMB groups (Roig-Coll et al., 2020), were related to significant changes in immune, cardiovascular and brain health biomarkers. Increases in the amount of PA observed in the AE group and higher CRF in the COMB group were related to reduced levels of ICAM-1, which is a pro-inflammatory molecule promoted by IGF and TNF-α and highly expressed in endothelial cells and related to atherosclerosis (Che et al., 2002; Frank and Lisanti, 2008). Increase in CRF in the AE and COMB groups was also related to reduced levels of SDF1-α which is the most common variant of SDF, a chemokine protein involved in angiogenesis and brain tissue repair, that contributes to neuroinflammation which in higher concentrations can lead to toxic environments and clinical consequences (Guyon, 2014). These results suggest that a 12-week program of brisk walking, in combination with cognitive training or alone, might enhance markers associated with cardiovascular and inflammatory health.

Further analyses examined the potential moderating role of individual characteristics on the interventions given their important role when designing personalized programs. We did not find a significant moderation effect of age on intervention-related changes in biomarkers at Level 1 and Level 2 in any group. Despite evidence is scarce for CCT programs, previous papers reported interactions of age with the neuroprotective effect of exercise (Colcombe and Kramer, 2003; Erickson et al., 2014) suggesting that the effect could be more apparent at older ages due to greater brain volume losses. Moreover, the mediating role of CRF in exercise-related cognitive benefits has been found in participants aged 70 or older but not in younger (Bherer et al., 2021). Therefore, the lack of a significant moderating effect of age in our study could be related to the relatively younger age of our participants as well as the limited age range of our participants. In relation to sex, we did not detect any significant moderation effects of sex on intervention-related changes in molecular biomarkers (Level 1) in any group despite previous evidence suggesting that sex moderates the effect of exercise on BDNF levels (Szuhany et al., 2015). However, we found that sex moderated the effects of the intervention on brain volumes on AE (total WM, parietal, temporal, and dorsolateral prefrontal cortex) and COMB groups (ventricles, temporal and dorsolateral prefrontal cortex) suggesting that men may benefit more than women. These results are consistent with our previously published results (Castells-Sánchez et al., 2019, 2020, 2021) and the larger adaptations that men might experience after exercise, as suggested in previous research (Al-Mallah et al., 2016). Moreover, these findings add support to the literature highlighting biological sex as a relevant moderator of the relationship between exercise and brain health (Barha and Liu-Ambrose, 2018; Barha et al., 2019). Sex differences in the physical activity-induced cognitive and brain changes may be associated with specific physiological adaptations in women and men, such as changes in neuroplastic processes, hormones, neurotransmitter systems and neurotrophic factors (Barha et al., 2019). Projecte Moviment adds support to the recent and growing evidence suggesting that sex matters. However, larger samples are necessary to run more highly powered intra-group analyses and further research should address the mechanisms underlying cognitive and physical training effects stratified by sex.

Finally, we examined the potential mediation effect of changes in molecular biomarkers and brain volume of targeted areas in the intervention-cognitive benefits through testable models in order to add evidence about the mechanisms underlying the neuroprotective effect of behavioral interventions (Stillman et al., 2020). The results of these mediation analyses showed that changes in biomarkers at Levels 1 and 2 did not significantly mediate the observed cognitive benefits in AE (Executive Functions and Attention-Speed) and in COMB groups (Attention-Speed). Previous evidence including a 1-year AE intervention reported that change in BDNF levels mediated the effect of intervention-related cognitive (task-switch) benefits in older adults (>71 years) (Leckie et al., 2014) and that intervention-related increases in hippocampal volume were significantly correlated to improvements in memory performance (Erickson et al., 2011). Since our intervention programs lasted 12 weeks, the duration of the activity as well as other potential methodological issues might be key parameters to find consistent effects of these interventions on these molecular and structural brain outcomes and their effect on cognition. The length of the intervention could be an important factor not only due to the persistence of the activity but also for the time needed to observe molecular and structural changes in healthy subjects. We might also hypothesize that early intervention-related cognitive benefits might be mediated by changes in other molecular markers such as other growth factor, cytokines or hormones which might interact differently for women and men and in relation to brain outcomes such as increased white matter integrity and better functional connectivity (Cabral et al., 2019). The sample size is also an important issue for trials examining the mechanisms of interventions since large samples are needed to detect mediation effects when the outcome measure is highly variable (neuroimaging data, molecular biomarkers) (Stillman et al., 2016). Those questions should be addressed for COMB interventions including AE and CCT in order to better understand the role of each single intervention. New omics technology might shed light on the complexity of biological networks underlying the cognitive benefits related to behavioral interventions (Hoffman, 2017).

The cascade of changes initiated by behavioral interventions that lead to cognitive enhancements is complex and requires a combined multi-level approach that addresses moderators such as sex, age, pathology and genetics. Further research should overcome our limited intra-group sample and include a wider range of participants with age and sex balanced across groups in order to be able to make specific intra-group analyses. We also acknowledge potential desirability bias in our sample since we obtained self-reported adherence and participants were highly motivated, the interventions were individual and home-based, and the use of a waitlist control group.

The significant relationship found between changes in physical activity outcomes and changes in immune, cardiovascular and brain health biomarkers highlight the role of the aerobic exercise in the promotion of a molecular cascade involving immunity and cardiovascular related markers that might enhance a neuroprotective effect in the aerobic exercise and combined groups. Our findings also highlight the importance of sex as a relevant variable for brain volume in these interventions. Our results provide further support for the effectiveness of CCT and reveals some underlying neurobiological mechanisms. CCT showed promising results enhancing the brain volume of the precuneus, despite previously published results that did not show significant changes in cognitive assessments. These results highlight the importance of a multi-level approach when assessing and describing the neuroprotective effect of behavioral interventions.

Moreover, parameters of the activity such as duration or intensity might be relevant to detect changes in the targeted molecular and structural brain outcomes. Other molecular and brain mechanisms should be addressed in order to better understand the cognitive benefits observed at early stages.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by the Bioethics Commission of the University of Barcelona (IRB00003099) and Clinical Research Ethics Committee of IDIAP Jordi Gol (P16/181). The patients/participants provided their written informed consent to participate in this study.

MM conceptualized the study, contributed to the study design and implementation as Principal Investigator, and guided and supervised all the statistical analysis and writing of this manuscript. PT-M and KE made substantial contributions to the design and implementation of the trial. AC-S and FR-C contributed to the design and implementation of this trial, collaborated with recruitment and evaluated participants before and after interventions, analyzed the data, and wrote this manuscript. NL-V collaborated with recruitment and did the follow-up of the intervention groups. AG-M and JT assisted in the use of GNPT program for computerized cognitive training. GP guided and supervised the statistical analysis. RD-A made the revision of the present manuscript and answered to the reviewer s questions and contributed processing the neuroimaging data and MV with the management of molecular markers. PM-A, AH-T, JS-R, CC, and SD contributed to the implementation of the trial from their area of expertise. All authors reviewed the manuscript and provided final approval for publication of the content.

The 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.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.