Exercise response variability: Random error or true differences in exercise response?

Connections link a sequence of three related research papers. The central article which links the other two papers has been published in Experimental Physiology. In a Connections article, an author (or authors) of the central article outlines its principal novel findings, tracing how they were influenced by the first article and how the central article has contributed to the developments made in the third article. The author(s) may also speculate on the direction of future research in the field. Connections articles aim to set the research in a wide context. Advances in molecular biology have provided invaluable insight into the cellular and molecular underpinning of exercise adaptation. This insight, coupled with a growing awareness of exercise response variability, ignited interest in identifying the predictors of exercise-induced adaptations for the purpose of personalized exercise prescription (Ross et al., 2019). Accordingly, attempts to explain exercise response variability using molecular modulators of the adaptive response are increasing across many facets of exercise physiology. Although molecular analytical techniques may potentially elucidate key regulatory steps that contribute to exercise-induced adaptations, the confounding influence of random error - noise attributable to technical and/or biological sources (Atkinson & Batterham, 2015) - may limit our ability to accurately elucidate predictors of individual response. In this paper, we make connections between several recent studies examining acute and chronic responses to exercise in an attempt to bring awareness to the question: When we observe exercise response variability, what exactly are we observing? Congruent with the central dogma of molecular biology (e.g. DNA to RNA to protein), adaptation to exercise is widely accepted as being initiated at the transcriptional level. The work of Perry et al. (2010) was pivotal in cementing this paradigm. Specifically, their research demonstrated that acute increases in muscle mRNAs encoding transcriptional and metabolic proteins precede chronic increases in the expression of these proteins. Consequently, the field of exercise physiology as a whole has accepted the paradigm that larger increases in mRNA expression after acute exercise indicate a greater activation of the molecular pathways that underpin adaptation. Thus, the work of Perry et al. (2010) provided foundational evidence supporting the hypothesis that the magnitude of change in mRNA expression after a given exercise stimulus predicts an individual's adaptive potential to that same stimulus. Despite the compelling nature of this hypothesis for researchers interested in personalized exercise prescription, existing studies have largely failed to establish direct relationships between acute gene expression and chronic phenotypic changes mediated by training. In an attempt to understand the failure of previous work to identify molecular markers of adaptive potential, we recently tested the repeatability of acute mRNA responses in exercised human muscle (Islam et al., 2019). On the one hand, if the observed changes in mRNA expression represent an inherent response unique to a given individual, then changes in mRNA expression after identical exercise stimuli should be repeatable. On the other hand, if two identical exercise bouts fail to elicit repeatable changes in mRNA expression, either the transcriptional response to exercise is variable within an individual, or the observed response is substantially influenced by sources other than the exercise stimulus itself (i.e. random error). In the latter case of non-repeatable responses, the utility of mRNA as a biomarker of adaptive potential would be severely limited. We exposed 11 active young men to two identical bouts of continuous endurance cycling (30 min at ∼65% of peak work rate) separated by a minimum of 2 weeks (Islam et al., 2019). Skeletal muscle biopsies were obtained from the quadriceps after each bout, and changes in a variety of mRNAs encoding transcriptional and metabolic proteins were quantified. Despite highly repeatable exercise bout characteristics (e.g. blood lactate, work rate and heart rate), we found that changes in muscle mRNA expression were not repeatable (i.e. individuals appeared to respond differently to the same stimulus). Moreover, this intra-individual variability in mRNA expression could not be explained by technical error arising from major analytical steps involved in gene analysis, pointing to sources of random error originating from within the muscle (e.g. morphological differences between samples, random shifts in gene expression, diurnal fluctuations in metabolism and/or transcription, etc.). Importantly, our observation that the observed responses to exercise differ under identical experimental conditions question the utility of mRNA as a biomarker of adaptive potential for personalized exercise prescription. An important caveat to our work is the absence of a non-exercising control group, which is a prerequisite for quantification of the amount of random error present in the observed response (Atkinson & Batterham, 2015). Owing to our failure to include a non-exercising control group, we were unable to comment on the cause of our observed variability. In other words, we were unable to discern variability attributable to exercise from variability attributable to non-exercise sources. To address this issue, Dankel et al. (2020) compared changes in muscle size and strength after two different resistance training programmes (both involving 18 sessions of elbow flexion over 6 weeks) with a time-matched no-exercise control group. Although both resistance training protocols expectedly increased muscle strength over the course of the intervention, comparisons with the no-exercise control group revealed that much of the variability in the observed responses was attributable to random error as opposed to the exercise stimulus itself. Importantly for the classification of individual responses for personalized exercise prescription, only ∼21% of the individuals in one of the training groups could be confidently classified as high or low responders for muscle strength gains after accounting for random error. The demonstration that chronic exercise response variability appears to be a consequence of random error rather than individual inherent trainability extends our work on acute exercise responses (Islam et al., 2019). Of note, we have recently demonstrated poor repeatability of training-induced changes in cardiorespiratory fitness after high-intensity interval training (Del Giudice et al., 2020), suggesting that the influence of random error on training adaptation is not unique to resistance training. Perry, C. G. R., Lally, J., Holloway, G. P., Heigenhauser, G. J. F., Bonen, A., & Spriet, L. L. (2010). Repeated transient mRNA bursts precede increases in transcriptional and mitochondrial proteins during training in human skeletal muscle. The Journal of Physiology, 588, 4795–4810. Islam, H., Edgett, B. A., Bonafiglia, J. T., Shulman, T., Ma, A., Quadrilatero, J., … Gurd, B. J. (2019). Repeatability of exercise-induced changes inmRNA expression and technical considerations for qPCR analysis in human skeletal muscle. Experimental Physiology, 104, 407–420. Dankel, S. J., Bell, Z. W., Spitz, R. W., Wong, V., Viana, R. B., Chatakondi, R. N., – Loenneke, J. P. (2020). Assessing differential responders and mean changes in muscle size, strength, and the crossover effect to 2 distinct resistance training protocols. Applied Physiology, Nutrition, and Metabolism, 45, 463–470. Despite the attractiveness of molecular biomarkers of adaptive potential for personalized exercise prescription (Perry et al., 2010), our work clearly highlights repeatability issues associated with such biomarkers (Islam et al., 2019) and, consequently, questions the utility of biomarkers for predicting chronic changes in phenotype. Furthermore, the demonstration by Dankel et al. (2020) that the majority of the observed response variability following training is attributable to random error questions our ability to accurately quantify an individual's true phenotypic response to exercise. In combination, these studies seem to suggest that the answer to the question, ‘When we observe exercise response variability, what exactly are we observing?’, appears to be ‘random error’ rather than ‘true differences in exercise response’. This being the case, two further questions arise. How can we improve interpretations of our findings? And how can the confounding influence of random error be minimized? A straightforward answer to both questions is to evaluate the repeatability of desired biomarkers of adaptive potential before using those biomarkers for predictive purposes. Likewise, the inclusion of time-matched no-exercise control groups in studies of exercise training interventions allows for the discrimination of random error from an individual's inherent response. Although these two approaches might seem trivial in the age of ‘omics’, there is little value in characterizing entire ‘-omes’ if we misinterpret what we are observing. Ultimately, reprioritizing the quality of data (as opposed to the quantity) may be the most critical factor for advancing the field of personalized exercise prescription. None declared. Both authors contributed to the drafting and revision of the manuscript. Both authors approved the final version of the manuscript and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. All persons designated as authors qualify for authorship, and all those who qualify for authorship are listed.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.