Combined exercise training-induced improvements in executive functions are related to baseline plasmatic levels of Brain-Derived Neurotrophic Factor (BDNF) in middle-aged adults and older with Type 2 Diabetes Mellitus

Background: Type 2 Diabetes Mellitus (T2DM) affects many cognitive functions and aerobic plus resistance exercise training, named combined training (CT), by improving metabolic control may mitigate or reverse the cognitive impairment T2DM-related. Brain-derived neurotrophic factor (BDNF) plays a substantial role in cognitive functions. However, the effects of CT on BDNF levels of T2DM subjects are poorly known. Aim: This study analyzed the effects of 8 weeks of CT on circulating BDNF levels and assessed whether plasmatic BDNF levels were related to CT-induced improvements in executive functions and long-term memory of T2DM subjects. Methods: Thirty-ve (63 ± 8 years old) T2DM subjects of both sexes entered into CT (n = 17, thrice-weekly during 8 weeks) or control group (CONT, n = 18). Executive functions (measured through Trail making test, Stroop color task, and Digit Span), long-term memory (measured through the simplied version of Taylor Complex Figure Test), and blood samples were evaluated pre- and post-intervention. Results: Pre-CT plasma BDNF levels were positively related to CT-induced improvements on executive functions composite z-score (r = 0.71), inhibitory control (r = 0.58) and cognitive exibility (r = 0.56), but not to long-term memory. Plasma BDNF levels were not statistically changed (pre-CT: 179 ± 88 pg/ml; post-CT: 148 ± 108 pg/ml; pre-CONT: 163 ± 71 pg/ml; post-CONT: 141 ± 84 pg/ml, p > 0.05). Conclusion: Higher pre-training BDNF levels might be a potentializing factor of the training-induced improvements on executive functions, independently of the training-alterations in resting BDNF levels of T2DM subjects.


Introduction
Type 2 Diabetes Mellitus (T2DM) is a metabolic disorder that impairs several metabolic pathways participating in insulin signaling, hence dysregulating glucose metabolism (DeFronzo et al. 2015).
Impaired glucose metabolism affects several organic tissues, and the brain is one of them. The negative changes in brain functioning and structure by T2DM induced some complications, as the increased risk for mild cognitive impairment (MCI) and dementia in nearly 20% and 50%, respectively (Cheng et al. 2012). Furthermore, these T2DM-related cognitive dysfunctions are progressive and accelerated, as type 2 diabetic subjects with MCI are likely to develop dementia almost three years before those without MCI (Ma et al. 2014).
From a healthcare perspective, adequate levels of cognitive functions are crucial to the T2DM treatment regimen, including blood glucose monitoring, medication adherence, reducing stress, adopting healthy behaviors, caring for both skin and feet. Poor diabetes self-care management (e.g. blood glucose monitoring and self-management of prescribed drug therapy), low diabetes knowledge, and higher help care needs are observed in those with the lowest cognitive screening scores in a sample of older adults with T2DM (Sinclair et al. 2000). Furthermore, older adults with T2DM and MCI are less likely to adhere to exercise and diet (Feil et al. 2012). T2DM subjects have shown impaired performance in tasks assessing some neurocognitive domains, such as executive functions and long-term memory (Palta et al. 2014;Sadanand et al. 2016). Low levels of executive function, somehow, increase the risk of older adults developing frailty and disability (Johnson et al. 2007). Otherwise, higher executive function scores may preserve functionality. T2DM subjects with the highest memory scores are more likely to adhere to prescribed drug therapy (Vedhara et al. 2004), which, in turn, can improve metabolic control and prevent T2DM-related outcomes. Thus, therapeutic strategies focused on improving cognitive functions may modify the course of T2DM, preventing or mitigating the progression of cognitive damages, hence contributing to the maintenance of T2DM subjects' cognitive health and functionality.
The planned exercise training program is a promising and low-cost therapy in preventing and treating many diseases, including cardiovascular diseases, certain types of cancer, and T2DM (Booth et al. 2012).
Moreover, exercise training also has central effects since midlife physical exercise lowers 40% the odd for developing MCI in late-life (Geda et al. 2010). Exercise training improved the long-term memory of cognitively normal older adults and executive functions of healthy and older adults with MCI (Sanders et al. 2019). In adults and olders with T2DM, exercise training also enhances some cognitive functions (Silveira-Rodrigues et al. 2021). However, the underpinning mechanisms involved in exercise traininginduced cognitive improvement in T2DM subjects needs to be better understood.
The Brain-derived Neurotrophic Factor (BDNF), a protein of the neurotrophins family, that when binds to its most abundant receptor, tropomyosin receptor kinase B (TrkB), exerts its central function by triggering neurons survival, growth, and improving their resistance to damage (Rozanska et al. 2020). Due to its actions in modulating synaptic transmission, synaptic plasticity, neurogenesis, learning, and memory processes, BDNF is assumed as a biomarker that re ects mnemonic symptoms in several pathologic conditions (Miranda et al. 2019). Several cross-sectional studies reported lower circulating BDNF levels in T2DM subjects than non-T2DM controls (Geroldi et al. 2006;Krabbe et al. 2007;Fujinami et al. 2008;Zhen et al. 2013;Sun et al. 2018). Indeed, the lower serum BDNF levels in adults and older with T2DM subjects were not only accompanied by poor attentional and long-term memory performance (Zhen et al. 2013) but also increased by at least 50% the risk for developing MCI in this population (Sun et al. 2018).
For these reasons, BDNF appears to play a relevant role in T2DM-related cognitive dysfunctions.
BDNF participates in the muscle-brain crosstalk since is released by skeletal muscle and other tissues after an exercise bout hence increasing their circulating and central levels (Walsh and Tschakovsky 2018). Only single (Rasmussen et al. 2009) or twice aerobic exercise bouts (Neeper et al. 1996) raises the mRNA BDNF expression in the prefrontal cortex and hippocampus, and an accumulating body of evidence suggests that BDNF modulates some of the exercise effects on cognition (Intlekofer et al. 2013;Tang et al. 2017). In humans, due to the inability to assess brain BDNF mRNA expression, current studies often assess the serum or plasma circulating BDNF levels since these levels are strongly correlated to brain BDNF levels (Sartorius et al. 2009). As it crosses the blood-brain barrier through a rapid saturable transporter system, peripheral synthesized BDNF could enter the central nervous systems exerting its neurotrophic role (Pan et al. 1998). In T2DM subjects (Brinkmann et al. 2017) and other populations (Szuhany et al. 2015), circulating BDNF levels increase after a single exercise bout. However, the effects of exercise training (repeated exercise bouts) on circulating BDNF of T2DM subjects need to be better understood. A long-term exercise training program failed to demonstrate signi cant changes in serum BDNF of adults and older with T2DM (Swift et al. 2012). Moreover, to the best of our knowledge, the relationship between BDNF levels and training-induced cognitive improvements remains unknown. Thus, considering that BDNF and its receptors TrkB are expressed in the prefrontal cortex and hippocampus (Neeper et al. 1996;Rasmussen et al. 2009), brain areas that encompass the executive functions and long-term memory, it is tempting to suggest that BDNF levels can mediate the training-induced improvements in these cognitive functions.
Therefore, this study aimed to analyze the effects of 8 weeks of combined (aerobic and resistance) exercise training on circulating BDNF levels of T2DM subjects and whether the pre-training BDNF levels or training-induced changes in BDNF levels were related to training-induced improvements on executive functions and long-term memory. It was hypothesized that BDNF levels would be related to the exercise training improvements in cognitive functions of middle-aged and older adults with T2DM.

Ethical care
The study protocol followed Helsinki's Declaration and was approved by the Research ethics committee and was registered in Brazilian Register of Clinical Trials (register no. RBR86hfz5). Volunteer's recruitment was performed by phone contact between February and May/17.

Subjects
Thirty-ve subjects signed a consent form to participate in this study after an explanation of the experimental procedures. Subjects were aged between 50-79 years, had up to 2 years of T2DM diagnosis, were literate, had normal or corrected visual and auditory functions, were physically inactive during the last three months and were absent of musculoskeletal conditions that would not allow physical training (See sample characteristics in Table 1). Were adopted as exclusion criteria those an attendance rate of less than 60% of the total of 24 CT training sessions. Seventeen subjects entered combined training (CT) and eighteen in control (CONT) groups. Four subjects left the study (one in CT for personal reasons; and three in CONT, one for lower limb fracture in daily living activity, another two for the impossibility to be contacted by phone number). In two participants of CONT, although cognitive tests were performed, it was not possible to assess BDNF levels. Therefore, 31 participants were included in the cognitive analysis, and 29 in BDNF levels analysis (Fig. 1).

Experimental design
Pre-training data were assessed in three moments, interspaced by at least 48h. On day one, the volunteers were cognitively screened and answered a questionnaire with daily living information, health status, occupation, and physical activity level. At the second occasion, blood samples and blood pressure were assessed. On day three, cognitive tasks and the six-minutes walk test were performed. After these procedures, CT-group performed 8 weeks of combined training and CONT-group was guided to maintain their life routine. At least 72 hours after the last training session, post-training blood samples and cognitive tasks were assessed. All the volunteers should avoid intense physical activity, caffeinated or alcoholic beverages ingestion, and to maintain their habitual food and water ingestion in the 24 hours before the experimental procedures.

Combined training program (CT)
The CT-protocol follows recommendations of the American Diabetes Association for exercise in subjects with T2DM (Colberg et al. 2016) and consisted of thrice-weekly non-consecutive sessions for 8 weeks when alternate increments in the exercise volume or exercise intensity conducted after each two-week microcycle. Exercise protocol are detailed in (Silveira-Rodrigues et al. 2021).

Executive functions and data transformation
Executive functions are the suite of processes that underlie goal-directed self-regulatory behaviors (Diamond 2013). The three components of the executive functions' core are cognitive exibility (that refers to creative processes and comprises the ability to alternate the perspective of analysis of a task or situation); inhibitory control (that encompasses the ability of attention, behavior, and emotion control to mitigate the intervention of distractors), and working memory (that includes the ability to hold information in mind and work mentally as math reasonings or reorder a task list, as an example) (Diamond 2013).
We used one different task to assess each executive function component. Cognitive exibility was assessed through the Trail making test (TMT), inhibitory control by the Stroop color task (SCT), and working memory by Digit Span (DS). The interference score (Nagamatsu et al. 2012) was adopted to measure the executive function performance of each domain, subtracting the values of the incongruent trial from the neutral trial. The cognitive exibility score was assessed by TMT-A minus TMT-B. Inhibitory control score through composite z-score of SCT-congruent and SCT-incongruent (total time + (2 * mean time per word * error number). Working memory through DS-reverse minus DS-direct. Executive function composite z-score represents the mean of cognitive exibility, inhibitory control, and working memory zscores. If the volunteer lost or was unable to perform any task, the arithmetic means of the other task parameters were determined as the individual composite z-score.
The cognitive task parameters were z-score transformed as adopted previously (Espeland et al. 2017) through the equation: , where x: individual value, X: sample mean and σ: group standard deviation. Considering that in TMT and SCT the lower time for answering these tasks represents better performance, then, the interference z-scores were multiplied by -1 (Nagamatsu et al. 2012).

Long-term memory
Memory refers to the capacity of retaining and manipulating acquired previous information through neural plasticity (Squire 2004). The simpli ed version of the Taylor Complex Figure Test (Paula et al. 2016), a task involving visuospatial abilities and episodic memory recall was employed to assess long- term memory. Individuals' scores were obtained through the difference between the immediate copy score and the delayed 30min-copy score. In this case, lower values represented better performance.

Blood Samples and biochemical analysis
Samples were collected 72h after the last familiarisation session (pre-training) and last CT-session (posttraining). Blood samples were collected in heparinized tubes in the morning and after 12h of fasting. After collection, samples were stored in styrofoam until centrifugation (10 min, 3500rpm, and 4°C) for plasma separation. After that samples were stored at -80ºC in aliquots of 100 µl until analysis. Plasma BDNF concentrations were quanti ed by sandwich ELISA speci c kit (DY248 DuoSet, R&D Systems™, MN, USA) according to the manufacturer's instructions. A single plate was blocked for 3 h in reagent diluent (1% bovine serum albumin (BSA)/ phosphate-buffered saline (PBS)) and incubated for nearly 12 h with 100 µl of samples at 4° C. Samples were diluted 1:64 to t standard curve range (detection range: 23-1500 pg/ml). The baseline blood glucose, lipids, and fructosamine were quanti ed by the enzymatic colorimetric technique using speci c kits (Gold Analisa™, MG, BR Sample size calculation and a posteriori power (2) power (1-β) = 0,7; (3) allocation rate of 1:1 and (4) aimed Cohen's d = 0.8 (large effect size) in a onetailed test (considering that CT improves executive functions). Thereby, a minimum sample size of 16 subjects was obtained, adjusted for 17 subjects after sample loss consideration. A posteriori power was calculated for BDNF levels using G*Power v3.0.10.

Statistical analysis
Normality and homoscedasticity were tested with Shapiro-Wilk's and Levene's tests, respectively. All variables were normally distributed, allowing parametric analysis. Sample characteristics and plasmatic BDNF levels were presented in mean ± standard deviation (SD). Student's t-test for independent samples (two-tailed) compared the changes in executive function composite z-score and interference scores in both groups after 8-weeks. Two-way ANOVA with repeated measures was used to compare the BDNF levels between time (pre vs. post-training) and group (CT vs. CONT). Three-way ANOVA was performed to compare the BDNF between time (pre vs. post-training), group (CT vs. CONT), and age (CT vs. CONT). Simple linear regression analyses were used to determine the relationship between plasmatic BDNF levels with CT-induced changes in executive functions and long-term memory. Additionally, Cohen's d was calculated: (x 1 -x 2 ) / (σ pooled), where x¹ was the pre-training mean, x² the post-training mean value, and σ the pooled standard deviation of these group. To observe the clinical relevance of the treatment on the studied parameters was classi ed according to the effect size (Cohen's d) as trivial if d < .2, small: d = .2 to .5), medium d = .5 to .8, and large d > .8. Data were analyzed in GraphPad Prism v5.0 (GraphPad Software Inc., CA, USA). The signi cance level was 5%.
Additionally, we tested whether the CT-induced change on BDNF levels (post minus pre-CT) are related to composite executive function z-score (p = 0.20), inhibitory control (p = 0.30), cognitive exibility (p = 0.33), working memory z-score (p = 0.51) or long-term memory z-score (p = 0.77) but no signi cant correlations were found for all.

Discussion
The main nding of our study was that CT-induced improvements in composite z-score of executive function were positively correlated with pre-training plasma BDNF levels in T2DM subjects, even without alterations in plasmatic BDNF levels elicited by the 8-week in those subjects. It suggests that sedentary T2DM subjects with the highest plasmatic BDNF levels may be the best responders in improving some domains of their executive functions after a short-term CT exercise program. Conversely, long-term memory appears not to be related to pre-training plasmatic BDNF concentrations.
Despite recent progress in the T2DM diagnosis and treatment, little attention is given to the chronic complications which impair the quality of life and self-management of T2DM. Growing and novel evidence focuses on the development of course-modifying therapies to counteracts T2DM-related cognitive dysfunction (Biessels and Despa 2018). A recent meta-analysis highlights the tri ing amount of published studies regarding the exercise training effects on cognitive outcomes of T2DM subjects (Cooke et al. 2020). Our results show that CT induced positive effects on executive functions of T2DM adults and older. Executive Functions have been related to general health (see Diamond, 2013) (Diamond 2013), and in T2DM subjects it has not been different (Sadanand et al. 2016). Several health aspects as selfmanagement of T2DM, lower hospitalization number (Sinclair et al. 2000), adherence to diet and exercise (Feil et al. 2012), depending on cognitive health preservation. Furthermore, executive functions were a better predictor of functional decline and mortality of older women than global cognition (Johnson et al. 2007). Enhance the executive functions by exercise training can induce other health bene ts, for example, the 1-year resistance training improvements in executive function contributed to the maintenance of physical activity over the following year (Best et al. 2014). However, the studies exploring concomitantly the mechanisms underlying the training-induced cognitive-bene ts in T2DM subjects are scarce.
BDNF is considered a promising biomarker for early identifying T2DM subjects with a higher risk to develop MCI (Sun et al. 2018). By modulating synaptic plasticity and neurogenesis, BDNF may induce positive cognitive effects (Intlekofer et al. 2013). In this regard, it has been speculated that BDNF can also induce positive effects on the cognition of T2DM subjects (Rozanska et al. 2020). However, according to a recent meta-analysis (Jamali et al. 2020), the effects of exercise training in BDNF levels of adults and older with T2DM are still scarce and inconclusive (Jamali et al. 2020). Tang et al. (2017) investigated the effects of exercise training on the memory of diabetic rats, revealing that resistance training improves not only the learning performance of rats but also upregulates the expression of BDNF mRNA and its receptor TrkB in the hippocampus, suggesting that BDNF also seems to be involved in training-induced cognitive bene ts in the face of metabolic alterations present in diabetes.
The present study failed to demonstrate signi cant changes in BDNF levels between the trained and control group. Evidence reveals that an exercise training program (lasting from a week to months) has a lower capability to increase the circulating BDNF levels compared to a single bout of exercise (Szuhany et al. 2015). In T2DM subjects, similar results are reported, when a single bout of aerobic exercise increases circulating BDNF levels (Brinkmann et al. 2017), whereas nine months of aerobic, resistance, or combined training didn't increase serum BDNF in 30-75 years old T2DM subjects (Swift et al. 2012). Furthermore, older adults with glucose intolerance, aerobically trained for six months also didn't have their plasma BDNF levels altered (Baker et al. 2010). Thus, the present study advances to current knowledge by demonstrating for the rst time, to the best of our knowledge, that an 8-week short-term combined (aerobic plus resistance) does not signi cantly change resting plasmatic BDNF levels in sedentary middle-aged and older T2DM subjects.
In an unprecedented way, we showed that the pre-training plasmatic BDNF levels had a positive correlation with CT-induced improvements in executive function (composite z-score) in middle-aged and older T2DM subjects. Rasmussen et al. (2009) reported that a single bout of prolonged aerobic exercise at low to moderate intensity brie y raised the plasmatic BDNF levels in humans, but one hour after exercise returned to pre-exercise levels. However, when analyzing the brain mRNA BDNF expression in the mice, a long-lasting increasing response was found in this gene expression. Unlikely plasmatic BDNF levels, an evident increase in the BDNF mRNA expressed in the prefrontal cortex and hippocampus occurred immediately after prolonged aerobic exercise in mice (Rasmussen et al. 2009). Interestingly, the peak of BDNF mRNA expression in both brain regions occurred around two hours after exercise cessation and remained upregulated for the next 24h (Rasmussen et al. 2009). In another study, after three weeks of voluntary exercise, brain BDNF mRNA expression increased, which was paralleled to memory improvements in mice (Intlekofer et al. 2013). However, when the BDNF synthesis is blocked, the exerciseinduced improvements in mice's memory didn't occur, suggesting that the cognitive amelioration was BDNF-dependent (Intlekofer et al. 2013). Likewise, higher serum BDNF levels were accompanied by higher increases in functional connectivity in the parahippocampus and middle temporal gyrus of middle-aged and older adults that performed 12 months of a training program focusing on exibility, toning, and balance (Voss et al. 2013). Taken together, all these results endorse that the repeated transitory elevation of circulating BDNF levels after a single bout of physical exercise is a substantial contributor to the exercise-induced cognitive bene ts, even without changing the circulating BDNF levels at rest. Therefore, it is possible that in our study the higher plasmatic BDNF basal levels can potentially favor the CTinduced cognitive improvements of the T2DM subjects.
A possible extrapolation of our ndings deals with the potential existence of an optimal window of opportunity that can boost up the exercise-induced improvements on executive functions in T2DM subjects. As it is known, T2DM subjects show lower basal BDNF levels than non-diabetic controls (Geroldi et al. 2006;Krabbe et al. 2007;Fujinami et al. 2008;Zhen et al. 2013;Sun et al. 2018), and this neurotrophic pro le (i.e. inferred by BDNF levels) seems to be related to the metabolic disruption observed in the subjects (Geroldi et al. 2006;Levinger et al. 2008;Hristova 2013). For example, those with higher risk factors for metabolic syndrome showed higher plasmatic BDNF levels than those with the lowest risk factors (Levinger et al. 2008). Also, higher BDNF levels are reported in the early stages whereas lower BDNF levels are reported in the late stage of T2DM (Geroldi et al. 2006;Hristova 2013). Another study with animals shows that BDNF modulates glucose metabolism. The exogenous intracerebroventricular administration of BDNF inhibited hepatic gluconeogenesis, hence, relieving hyperglycemia in the T2DM murine model, even without explaining the exact neuronal mechanisms involved (Meek et al. 2013). Akin, in our investigation, 75% of the sample studied attained multiple criteria for metabolic syndrome (Expert Panel on Detection, Evaluation 2001). The Neurotrophic Theory proposed by Hristova (2013) postulated that the early raises in BDNF levels observed in subjects with a higher risk factor for metabolic syndrome consist of an organic attempt to counteract the initial T2DM pathophysiology impairments comprising altered glucose metabolism and insulin signaling. Thence, we might suggest that the highest pre-training circulating BDNF levels in the T2DM subjects of our study may have provided an advantage in obtaining the exercise-induced improvements on executive functions.
When dissociated, the subcomponents of executive functions in our study were differently related to pretraining plasma BDNF levels. Higher inhibitory control and cognitive exibility were positively correlated to higher pre-training BDNF levels. However, the exercise-induced improvements in working memory and episodic long-term memory were unrelated to pre-training BDNF levels. Contrasting with our ndings, BDNF was crucial to memory development in a rodent model (Intlekofer et al. 2013). Also, in T2DM subjects, a previous study reported that serum BDNF levels correlated to episodic and semantic memory (Zhen et al. 2013). Indeed, memory is a complex system with some subdivisions, encompassing ultrarapid, short-term, and long-term memories. Long-term memory consists of storing information for late use and are separated into declarative and non-declarative memories. Declarative memory, in turn, is separated into semantics, which comprises ideas and concepts not drawn from personal experience, and episodic memory that is dependent on a speci c event (Squire 2004). Working memory is one of the three core executive functions, composed of verbal and visuospatial working memory (Diamond 2013). The main difference between working and short-term memory is their component of manipulating information in the mind (Diamond 2013). In our study, both visuospatial episodic and verbal working memories were not signi cantly related to circulating pre-training BDNF levels. However, two main aspects should be considered. Firstly, animal studies measure the expression of the brain BDNF mRNA, which is impossible in human in-vivo studies. This might be relevant since, in response to an exercise session, the increases in brain BDNF levels last longer than those observed in plasma (Rasmussen et al. 2009). Secondly, our study assessed long-term memory by the adapted Taylor Complex Figure test, which only assesses episodic visuospatial abilities. Also, the working memory test used only recognizes the verbal component of working memory. Thus, should be included in future investigations the assessment of other memory components, including semantic and visuospatial memories. This fact would explain more properly the relationship between circulating baseline BDNF levels and putative exercise-induced memory improvements.
This study has some limitations. Although the sample size was calculated based on executive function variables, a posteriori power reached between plasmatic BDNF levels pre and post-intervention was 23%, suggesting a considerable likelihood of type II error occurrence. It can be due to the age and cognitive performance heterogeneity of our sample. Also, this study used a commercial enzyme-linked immunosorbent assay (ELISA) kit (R&D Systems™), that does not recognize proBDNF, a molecule that antagonizes mature BDNF in their actions in the central nervous system. Further, most of the study participants underwent medication therapy, which might have modulated plasmatic BDNF levels. Future studies may consider larger sample size and other metabolic pathways to clarify this differential effect of exercise training under distinct baseline BDNF levels.

Conclusion
Page 12/21 The magnitude of exercise-induced improvements in executive functions can be dependent on pretraining BDNF levels in T2DM subjects independently of signi cant training-induced changes in resting plasma BDNF levels.
Declarations Figure 1 CONSORT Flow chart of the study.

Figure 3
Changes (Post minus Pre) in composite score of executive functions (Combined training group, CT, n=16 and Control group, CONT, n=14). Values are mean ± SD. * indicates a signi cant difference (p<0.05) in unpaired Student's T-test.