Long-term exercise adaptation. Physical aging phenomena in biological networks

1 Introduction

Long-term adaptation to exercise, consisting of the cumulative, delayed, and residual effects (Issurin, 2009), is a consequence of task-dependent functionally integrated physiological networks and their interaction with the environment. However, principles of dynamic integration of these networks remain partially unknown (Ivanov and Bartsch, 2014; Balagué et al., 2020; Ivanov et al., 2021a; Ivanov et al., 2021b; Ivanov, 2021; Rogers, Mourot, Doucende, and Gronwald, 2021; Romero-Ortuño et al., 2021; Garcia-Retortillo and Ivanov, 2022). Particularly, the long-term effects of exercise on network dynamics are a much less researched area (e.g., Balague et al., 2016; Vázquez et al., 2016). A long-recognized general characteristic to the dynamics of the long-term effects is the tendency toward the slowing down of fitness¹ record value² occurrence, on average, on a large variety of time scales (e.g., Platonov, 1988). Moreover, the long-term rate of decay of the achieved fitness level, after temporary or permanent cessation of exercise, is inversely proportional to the accumulated training time (e.g., Zatsiorsky and Kraemer, 2006). This highly reproducible effect has been known since Hettinger’s early works on isometric strength (Hettinger, 1966). “Soon ripe, soon rotten” is a common catchphrase that depicts this effect. The causes of this phenomenon, however, are still unknown (Gavanda, Geisler, et al., 2020). In this opinion paper, our aim is to suggest that the previously mentioned phenomena may be successfully, theoretically, and methodologically approached by conceptualizing them as multilevel-physiological-network aging phenomena, known from condensed matter physics (see, e.g.,; Bouchaud, 2000)³. Before explicitly discussing this possibility, we briefly build an argument for our case.

Supercompensation, the process of increasing the substrate and functional possibilities of biological systems above the levels that precede an acute exercise perturbation, is the most fundamental multilevel process that underpins exercise-induced long-term adaptation. Recently, it has been conceptualized in terms of a strong anticipation phenomenon (Hristovski and Balagué, 2020). With this respect, important developments suggest that this kind of adaptive anticipatory process can be understood in terms of multiplicative cascades synchronized to the statistical structure of the environmental dynamics (Stephen and Dixon, 2011). Multiplicative cascades were recently discussed as formal characterization of anomalous diffusion processes (Mangalam, Metzler, and Kelty-Stephen, 2023), which are ubiquitous in hierarchical energy/entropy landscapes, characteristic for aging complex systems with many metastable states (Bouchaud, 2000). Complex, and by extension biological, networks may contain a very large number of metastable states. Metastable states are states in which the network may reside for a prolonged time but which eventually are being abandoned. Metastability has already been shown to exist in many types of biological networks, e.g., neural, transcriptomic, and endocrine networks (see Gili, Ciullo, Spalletta, 2018; Helmling, C., Klötzner, et al., 2018; Avila-Varela, Hidalgo-Lopez, et al., 2023). The metastable states in biological systems are constrained by the interactions between a vast number of multiscale co-adaptive networks, such as neuromuscular, fascial, cardiorespiratory, hormonal, immune, and the -omics networks, in cells, blood, etc. (Balagué et al., 2020). One of the main consequences of the heterogeneity of multiscale networks is the wide difference of time-scales of different variables that influence the biological network dynamics and, hence, the nested landscape of network-wide metastable states (i.e., minima), in which the fitness level may be temporarily trapped (Bouchaud, 2000). The metastable states are separated by energy/entropy barriers⁴. The depth of the trapping minima is proportional to the height of the barriers that surround them. On the other hand, the height of the barriers is roughly proportional to the quantity of stabilized reconfigurations that occurred in the network during its evolution. In this view, exercise-induced fitness can be considered a strong anticipation-based collective variable (Hristovski et al., 2010) generated by heterogenous network-wide hierarchical multiplicative cascade processes each time it is measured.

2 A heuristic model of the exercise-induced aging process

Aging is manifested in two basic facts: 1) The longer the fitness evolves (i.e., ages) under a certain sequence of perturbations (i.e., exercise stimuli), the more it slows down; and 2) the longer the fitness variables evolve, the longer the achieved fitness effects last after the cessation of the perturbations. Aged variables become ‘inert’ and resist change either for further enhancement or decay. The aging of the fitness variables can be measured by some macroscopic variables, that is, global fitness measure (e.g., max-power, strength, endurance, skill-related variables etc.), and/or some network-specific (e.g., network excitability, connectivity, efficiency, or energy flux) fitness variables. Aging is reflected in the increasingly longer waiting times for the record values of the fitness variable (Jensen and Sibani, 2013), or more technically as $\dot{\mu }\propto a{t}^{-1}$, where $\dot{\mu }$ represents the rate of decay of the average number of record values of the variable per unit time t and $a$ represents a constant.

Hence, in our heuristic model, aging is an exercise-activated process of relaxation (equilibration) of metastable multilevel-network configurations (see Bouchaud, 2000 for details). The core of this model includes the mutual parametrizing interactions of processes that relax on different characteristic time-scales. This means that, after some exercise time, the network’s faster evolving variables (e.g., enzymatic or signaling network activity, neural excitability, etc.) may have already relaxed and stabilized, the more slowly evolving variables (e.g., muscle; tendon; bone and heart remodeling; intra-, inter muscle, and interlimb coordination; capillarization, etc.) have not yet. Then, the stabilized network processes parametrize (i.e., constrain) the dynamics of the slower processes. As a result, hierarchical, network-wide, adaptive processes are being trapped, i.e., stabilized, between increasingly higher energy/entropy barriers (Bouchaud, 2000). Consequently, the response to exercise perturbations often decelerates and plateaus (Gorostiaga, Izquierdo, et al., 1999).

Plateauing (Gorostiaga et al., 1999) is captured phenomenologically by general, experimentally found, decelerating (e.g., logarithmic) laws of learning (Harris, 2022). Hence, the decelerating scenario of fitness growth as relaxation over hierarchical energy and/or entropy landscape (Bouchaud, 2000) directly predicts empirically detected laws. In order to enhance the system further, one must exert a change in the perturbations, e.g., directed variation in the load. The ultimate plateauing phases of the aging process may be detected on different time scales, from novices who practice their first weeks of exercise to individuals who exercise for decades (Platonov, 1988).

2.1 Rejuvenation and memory effects

Increasing the specificity (directed reduction of the variability⁵ of exercise perturbations) (Gamble, 2006) can initiate a fitness increase (on scales of months to decades). This is exactly what the phenomenon of rejuvenation means in physically aging systems. By reducing the variability of perturbations, i.e., increasing their specificity, the previously acquired coarser organization network modes stay stable, but more specific detailed configurational refinements are being activated (see Bouchaud, 2000). The finer structure of the network energy/entropy landscape is being revealed. The system restarts the aging process. Here, the measure of susceptibility (receptivity) χ may be used, which is defined as a change in some fitness function $\left(f\right)$ per constant unit change in the accumulated exercise perturbations $\left(p\right)$, or $\mathrm{\chi }=\frac{df\left(p\right)}{dp}$. As a consequence of the increase in the specificity, the susceptibility to the new specific perturbations again increases, which further improves the values of the fitness function $\left(f\right)$. This is usually an effect of exploring the more detailed structure of the landscape of task-specific physiological/biomechanical metastable configurations (Stone et al., 2022). When variability of perturbations is returned to the previous (less specific and more variable) level, the susceptibility to exercise perturbations $\mathrm{\chi }$ returns to the values characteristic for the states before the increment of the exercise specificity. The network energy/entropy landscape recovers the pre-rejuvenation coarser form. As a consequence, when measured, the system is still able to organize the previously stabilized coarser, less specific, structure (as in the bicycling skill). This is called a memory effect in physically aging systems (Bouchaud, 2000). Likely, different modalities of transfer may show different influences on network rejuvenation and memory effects (Brearley and Bishop, 2019). It is also possible that rejuvenation and memory effects exist only in some fitness variables, and detecting these variables would provide information about the structure of the biological networks.

2.2 The detraining effects of aging

Detraining effects of aging are measured starting from the highest fitness value achieved after switching-off of perturbations that acted on the system while it aged for time t_e. The main effect of aging detraining dynamics can be written down in both a stretched exponential form $f\left(t_{e,}{t}_e+t\right)\propto e^{{-\left(\gamma \left(t_e\right)t\right)}^{p\left(t_e\right)}}$ and power law form $f\left(t_{e,}{t}_e+t\right)\propto a{t}^{\left(-p{t}_e\right)}$, where f(t_e, t_e+t) represents the fitness function, t represents detraining time, $a$ is a constant which may depend on t_e, and constants γ and p depend on t_e and control the rate of decay of the fitness function.

The longer the aging process, i.e., the accumulated exercise time (t_e) of the system, the longer it takes the system to lose its memory of the past states, acquired during aging and vice versa. This very effect is typical in long-term detraining effects (e.g., Hettinger, 1966; Zatsiorsky and Kraemer, 2006). In the theoretical framework of physical aging, this results from the network being trapped in a deep minimum (i.e., the exceedingly stable state) due to the accumulated aging of the slow variables (Jensen and Sibani, 2013). As a consequence, escaping from that state would require much more time, proportional to the quantity of the required reverse (detraining) reconfigurations of the network, compared to the time required to escape from much shallower minima (i.e., less stable states), characteristic for networks with less accumulated exercise time (t_e). However, the biological reasons for this stabilization are yet unknown (Gavanda, Geisler, et al., 2020).

For some fitness components (e.g., strength), one may attribute this effect, at least partially, to the phenomenon of muscle memory (see Sharples, and Turner, 2023). Skeletal muscle memory, which is hypothesized to comprise the synergistic action of cellular (myonuclear) and epigenetic network processes responsible for hypertrophy, would enable the quick recovery of fitness function (e.g., strength) during retraining, after longer periods of detraining. However, in Hettinger (1966), the smaller the maximal acquired strength and the longer it has been acquired (larger accumulated exercise time), the slower was the rate of its decay during the detraining period and vice versa. In other words, it was not the maximal level of acquired strength and hypertrophy but the accumulated exercise time, which had a positive effect on the strength stability during the detraining period. Certainly, longer accumulated exercise time often brings about a higher level of strength as well. However, the said work shows that, when partialized with respect to the acquired maximal strength level, the accumulate exercise time, alone, is the generator of the strength stability. Thus, the following question is raised: How does the accumulated exercise time (t_e) stabilize the multilevel-network processes, including possible myonuclear and epigenetic mechanisms, which positively contribute to the fitness stability during detraining?

Moreover, the “soon ripe soon rotten” phenomenon is also present in the domain of skill acquisition (e.g., Yin et al., 2009), where the early acquired skill is easily subject to decay, but after prolonged use, it may last a lifetime. Hence the said phenomenon cannot be explained by the same cellular processes/mechanisms as muscular hypertrophy. In this respect, different aging phenomena may consist of different local processes/mechanisms but may also have common mechanisms. Hence, we hypothesize that it is the mutually parametrizing (constraining) role of multi-network fast and slow processes, which is the general principle that generates the diverse physical aging phenomena.

3 Research program

The heuristic model described previously remains largely qualitative, and to become quantitatively predictive, a research program focused (though not exclusively) on the following directions is required: 1. Establishing the essential network variables and their characteristic time-scales at each organism level (e.g., rate constants, connectivity measures, reciprocally compensating, i.e., synergic couplings, etc.), which potentially age and rejuvenate. In our opinion, the best approach would be the research on the long-term evolution of a macroscopic fitness variable (e.g., strength) and one or more nested networks associated with the macroscopic variable (see, e.g., for muscle-network connectivity measures during acute fatigue Garcia-Retortillo et al., 2023); 2. determination of networks’ energy/entropy landscapes; for example, the landscape of connectivity measures of intermuscular or endocrine interactions; 3. research on the dynamic mechanisms of aging of synergies between multilevel-network processes; 4. empirical and modeling work on the heterochronicity of multilevel network variables that age and rejuvenate and, hence, constrain the reconfiguration process of the physiological network fitness level at different time-scales. It is highly likely that heterochronicity has large interindividual variability and different organic systems, so the networks may have a strong individual imprint; 5. how the change in the variability of exercises (less-more specificity) affects the multilevel aging and rejuvenation of network configurations; 6. defining the multilevel network bio-markers that can explain and assess the long-term exercise-induced fitness effects of aging, rejuvenation, and memory. This research program requires effort and time, but in the long run, it may prove beneficial for science and the society.

Author contributions

RH conceived the idea and wrote the draft of the texts. NB and MS reviewed the draft and approved the final version. All authors contributed to the article and approved the submitted version.

Conflict of interest

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.

Publisher’s note

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.

Footnotes

¹Under “fitness,” here, we understand any exercise-induced adaptation in components such as strength, power, types of endurance, speed, complex perceptual-motor skills, and flexibility, as well as the generalized adaptations which emerge from the interactions among these components.

²The record value is the highest value measured in some fitness variables in some period of time for a certain individual.

³The concept of “aging,” in this paper, refers to the phenomenon of “physical aging” present in, e.g., complex disordered systems. It does not refer to human biological aging. The reader should not conflate both phenomena.

⁴In networks, the term energy barrier may or may not refer to physical energy. It may refer simply to the difficulty of overcoming some state. Entropy barriers, on the other hand, correspond to the possibilities (paths) that the system can use to escape the current metastable state. The more paths are afforded, the smaller is the entropy barrier and the higher is the probability of escaping and switching to another metastable state (see, e.g., Jensen and Sibani, 2013).

⁵In the general formulation of the Fokker–Planck equation, the variance corresponds to the temperature term in physical systems.

References

Avila-Varela, D. S., Hidalgo-Lopez, E., Dagnino, P. C., Acero-Pousa, I., del Agua, E., Deco, G., et al. (2023). Whole-brain dynamics and hormonal fluctuations across the menstrual cycle: the role of progesterone and age in healthy women. bioRxiv, Avaialable at: https://doi.org/10.1101/2023.07.23.550200.

Google Scholar

Balague, N., González, J., Javierre, C., Hristovski, R., Aragonés, D., Álamo, J., et al. (2016). Cardiorespiratory coordination after training and detraining. A principal component analysis approach. Front. Physiol. 7, 35. doi:10.3389/fphys.2016.00035

PubMed Abstract | CrossRef Full Text | Google Scholar

Balagué, N., Hristovski, R., Almarcha, M., Garcia-Retortillo, S., and Ivanov, P. C. (2020). Network physiology of exercise: vision and perspectives. Front. Physiol. 11, 611550. doi:10.3389/fphys.2020.611550

PubMed Abstract | CrossRef Full Text | Google Scholar

Bouchaud, J. P. (2000). “Aging in glassy systems: new experiments, simple models, and open questions,” in Soft and fragile matter: Nonequilibrium dynamics, metastability and flow. Editors M. E. Cates, and M. R. Evans (Bristol and Philadelphia: IOP Publishing), 285–304.

CrossRef Full Text | Google Scholar

Brearley, S., and Bishop, C. (2019). Transfer of training: how specific should we Be? Strength Cond. J. 41 (3), 97–109. doi:10.1519/SSC.0000000000000450

CrossRef Full Text | Google Scholar

Gamble, P. (2006). Implications and applications of training specificity for coaches and athletes. Strength & Cond. J. 28 (3), 54–58. doi:10.1519/00126548-200606000-00009

CrossRef Full Text | Google Scholar

Garcia-Retortillo, S., and Ivanov, P. C. (2022). Inter-muscular networks of synchronous muscle fiber activation. Front. Netw. Physiol. 2, 1059793. doi:10.3389/fnetp.2022.1059793

PubMed Abstract | CrossRef Full Text | Google Scholar

Garcia-Retortillo, S., Romero-Gómez, C., and Ivanov, P. C. (2023). Network of muscle fibers activation facilitates inter-muscular coordination, adapts to fatigue and reflects muscle function. Commun. Biol. 6, 891. doi:10.1038/s42003-023-05204-3

PubMed Abstract | CrossRef Full Text | Google Scholar

Gavanda, S., Geisler, S., Quitmann, O. J., Bauhaus, H., and Schiffer, T. (2020). Three weeks of detraining does not decrease muscle thickness, strength or sport performance in adolescent athletes. Int. J. Exerc Sci. 13 (6), 633–644. PMID: 32509134; PMCID: PMC7241623.

PubMed Abstract | Google Scholar

Gili, T., Ciullo, V., and Spalletta, G. (2018). Metastable states of multiscale brain networks are keys to crack the timing problem. Front. Comput. Neurosci. 12, 75. PMID: 30254581; PMCID: PMC6141745. doi:10.3389/fncom.2018.00075

PubMed Abstract | CrossRef Full Text | Google Scholar

Gorostiaga, E. M., Izquierdo, M., Iturralde, P., Ruesta, M., and Ibáñez, J. (1999). Effects of heavy resistance training on maximal and explosive force production, endurance and serum hormones in adolescent handball players. Eur. J. Appl. Physiol. Occup. Physiol. 80 (5), 485–493. PMID: 10502084. doi:10.1007/s004210050622

PubMed Abstract | CrossRef Full Text | Google Scholar

Harris, J. A. (2022). The learning curve, revisited. J. Exp. Psychol. Animal Learn. Cognition 48 (4), 265–280. doi:10.1037/xan0000331

CrossRef Full Text | Google Scholar

Helmling, C., Klötzner, D. P., Sochor, F., Mooney, R. A., Wacker, A., Landick, R., et al. (2018). Life times of metastable states guide regulatory signaling in transcriptional riboswitches. Nat. Commun. 9, 944. doi:10.1038/s41467-018-03375-w

PubMed Abstract | CrossRef Full Text | Google Scholar

Hettinger, T. (1966). Isometrische muskeltraining. Stuttgard, Germany: Fischer-Verlag.

Google Scholar

Hristovski, R., and Balagué, N. (2020). Theory of cooperative-competitive intelligence: principles, research directions, and applications. Front. Psychol. 11, 2220. doi:10.3389/fpsyg.2020.02220

PubMed Abstract | CrossRef Full Text | Google Scholar

Hristovski, R., Venskaitytė, E., Vainoras, A., Balagué, N., and Vazquez, P. (2010). Constraints-controlled metastable dynamics of exercise-induced psychobiological adaptation. Medicina 46 (7), 447–453. doi:10.3390/medicina46070064

PubMed Abstract | CrossRef Full Text | Google Scholar

Issurin, V. B. (2009). Generalized training effects induced by athletic preparation. A review. J. Sports Med. Phys. Fit. 49 (4), 333–345. PMID: 20087292.

PubMed Abstract | Google Scholar

Ivanov, P. C. (2021). The new field of network physiology: building the human physiolome. Front. Netw. Physiol. 1, 711778. doi:10.3389/fnetp.2021.711778

PubMed Abstract | CrossRef Full Text | Google Scholar

Ivanov, P. C., and Bartsch, R. P. (2014). “Network physiology: mapping interactions between networks of physiologic networks,” in Networks of networks: the last frontier of complexity (Cham: Springer International Publishing), 203–222.

CrossRef Full Text | Google Scholar

Ivanov, P. C., Wang, J. W., Zhang, X., and Chen, B. (2021b). “The new frontier of network physiology: emerging physiologic states in health and disease from integrated organ network interactions,” in 2019-20 MATRIX annals (Cham: Springer), 237–254.

CrossRef Full Text | Google Scholar

Ivanov, P. C., Wang, J. W., and Zhang, X. (2021a). “Signal processing in network physiology: quantifying network dynamics of organ interactions,” in 2020 28th European Signal Processing Conference (EUSIPCO), Amsterdam, Netherlands, 18-21 January 2021, 945–949.

CrossRef Full Text | Google Scholar

Jensen, H. J., and Sibani, P. (2013). Stochastic dynamics of complex systems: From glasses to evolution (Vol. 2). Singapore: World Scientific Publishing Company.

Google Scholar

Mangalam, M., Metzler, R., and Kelty-Stephen, D. G. (2023). Ergodic characterization of nonergodic anomalous diffusion processes. Phys. Rev. Res. 5 (2), 023144. doi:10.1103/PhysRevResearch.5.023144

CrossRef Full Text | Google Scholar

Platonov, V. I. (1988). Adaptation in sport. Kiev: Zdorovje.

Google Scholar

Rogers, B., Mourot, L., Doucende, G., and Gronwald, T. (2021). Fractal correlation properties of heart rate variability as a biomarker of endurance exercise fatigue in ultramarathon runners. Physiol. Rep. 9 (14), e14956. doi:10.14814/phy2.14956

PubMed Abstract | CrossRef Full Text | Google Scholar

Romero-Ortuño, R., Martínez-Velilla, N., Sutton, R., Ungar, A., Fedorowski, A., Galvin, R., et al. (2021). Network physiology in aging and frailty: the grand challenge of physiological reserve in older adults. Front. Netw. Physiology 1, 712430. doi:10.3389/fnetp.2021.712430

CrossRef Full Text | Google Scholar

Sharples, A. P., and Turner, D. C. (2023). Skeletal muscle memory. Am. J. Physiology-Cell Physiology 324 (6), C1274–C1294. doi:10.1152/ajpcell.00099.2023

CrossRef Full Text | Google Scholar

Stephen, D. G., and Dixon, J. A. (2011). Strong anticipation: multifractal cascade dynamics modulate scaling in synchronization behaviors. Chaos, Solit. Fractals 44 (1-3), 160–168. doi:10.1016/j.chaos.2011.01.005

CrossRef Full Text | Google Scholar

Stone, M. H., Hornsby, W. G., Suarez, D. G., Duca, M., and Pierce, K. C. (2022). Training specificity for athletes: emphasis on strength-power training: A narrative review. J. Funct. Morphol. Kinesiol 7 (4), 102. doi:10.3390/jfmk7040102

PubMed Abstract | CrossRef Full Text | Google Scholar

Vázquez, P., Hristovski, R., and Balagué, N. (2016). The path to exhaustion: time-variability properties of coordinative variables during continuous exercise. Front. Physiol. 7, 37. doi:10.3389/fphys.2016.00037

PubMed Abstract | CrossRef Full Text | Google Scholar

Yin, H., Mulcare, S., Hilário, M., Clouse, E., Holloway, T., Davis, M. I., et al. (2009). Dynamic reorganization of striatal circuits during the acquisition and consolidation of a skill. Nat. Neurosci. 12, 333–341. doi:10.1038/nn.2261

PubMed Abstract | CrossRef Full Text | Google Scholar

Zatsiorsky, V. M., and Kraemer, W. J. (2006). Science and practice of strength training. Champaign: Human Kinetics.

Google Scholar