Acute increases in brain-derived neurotrophic factor in plasma following physical exercise relates to subsequent learning in older adults

Multidomain lifestyle interventions represents a promising strategy to counteract cognitive decline in older age. Brain-derived neurotrophic factor (BDNF) is essential for experience-dependent plasticity and increases following physical exercise, suggesting that physical exercise may facilitate subsequent learning. In a randomized-controlled trial, healthy older adults (65–75 years) completed a 12-week behavioral intervention that involved either physical exercise immediately before cognitive training (n = 25; 13 females), physical exercise immediately after cognitive training (n = 24; 11 females), physical exercise only (n = 27; 15 females), or cognitive training only (n = 21; 12 females). We hypothesized that cognition would benefit more from cognitive training when preceded as opposed to followed by physical exercise and that the relationship between exercise-induced increases in peripheral BDNF and cognitive training outcome would be greater when cognitive training is preceded by physical exercise. Greater increases of plasma BDNF were associated with greater cognitive training gains on trained task paradigms, but only when such increases preceded cognitive training (ß = 0.14, 95% CI [0.04, 0.25]). Average cognitive training outcome did not differ depending on intervention order (ß = 0.05, 95% CI [−0.10, 0.20]). The study provides the first empirical support for a time-critical but advantageous role for post-exercise increases in peripheral BDNF for learning at an interindividual level in older adults, with implications for future multidomain lifestyle interventions.

S2. Detailed descriptions of working memory training tasks. Task Task description Difficulty Level Level description Criterion (max) N-back Stimuli were presented on the screen, one by one, and the task was to press a button whenever the currently shown stimulus was the same as the one presented N steps back in the series. Difficulty was varied by the number of steps (2-back, 3-back) and whether the stimuli were evaluated based on one or two aspects of the stimuli (single, dual). For example, a dual 2-back task required participants to base their decision on the word and the color of the stimulus (targets marked with *): sofa, football, chest, football*, tree*, sofa, football, sofa*. The four stimulus sets included words and colours (as in the example), shape and patterns, shape and position in a 3x3 grid, and, number and position in a 3x3 grid. With the exception of position, which included 9 items, stimulus sets consisted of six items. Each stimulus was presented for 2250ms with an interstimulus interval of 250ms. 24 stimuli were presented for each of the eight 1-minute runs with an average of 30% of the stimuli in a run being targets. A fixed criterion, expressed as number correct in a run (hits + correct rejections), had to be reached to move on to the next difficulty level. In each training session, 12 runs were completed. Participants received figural feedback on their performance on each run. Stimuli were presented on the screen, one by one, and the task was to remember the N last stimuli. The difficulty was varied by how many stimuli that had to be remembered (2 last, 3 last) and by presenting one or two series of stimuli simultaneously (single, dual). For example, a 2-last dual task required participants to indicate the two last pairs of stimuli whenever the stimuli stream stopped (targets marked with *): moneychair, money-car, ball-chair, elephant*-plane*, elephant*-car* (stimuli presented as images). The four stimulus sets included two image sets (as in the example), letter and shape, word and color, and, number and position in a 3x3 grid. With the exception of position, which included 9 items, stimulus sets consisted of three items. Each stimulus was presented for 2000 ms with an inter-stimulus interval of 250 ms. In each run, a response was required 7 times. At response, all stimuli items were presented and participants responded by clicking the items in the correct order within 8 seconds.
A fixed criterion, expressed as number correctly identified stimuli items for each run, had to be reached to move on to the next difficulty level. In each training session, 4 runs were completed. Participants received figural feedback on their performance on each run. The daily n-back task was non-adaptive and was administered after each cognitive training session. Stimuli were presented on the screen, one by one, and the task was to press a button whenever the currently shown stimulus was the same as the one presented 3 steps back in the series. Stimuli were evaluated based on a single aspect of the stimuli (identity). The stimuli set consisted of six symbolic pictures that were not part of the cognitive training. Each stimulus was presented for 2250ms with an interstimulus interval of 250ms. A total of 24 stimuli (8 targets) were presented for each of the four 1-minute runs, with 10 seconds break between each run. Subjects received feedback on their performance after each run. Both hits and correct rejections contributing to the score, resulting in a maximum score of 24. Responses were required at 10 occasions at the 2-last dual level (max correct/trial=4), at 10 occasions at the 3-last single level (max correct/trial=3) and at 10 occasions at the 2-last dual level (max correct/trial=6). The stimuli set consisted of letters and shapes, which was used during training. In the dual task version, participants had to remember the letter as well as the shape. Completion time was approximately 20 minutes.

Average proportion correct per trial
Trained WM (untrained stimuli) N-back, untrained stimuli The task was identical to the N-back (trained stimuli) with the exception of the stimuli, which was not used during the cognitive training. The stimuli set consisted of six photographic images (e.g. fire, water) that were presented at different counts (1-6 images). In the dual task version, participants were required to base their decision on the count as well as the identity of the image.

Number correct (Number hits + number correct rejections)
Running span, untrained stimuli The task was identical to the Running span task (trained stimuli) with the exception of the stimuli, which was not used during the cognitive training. The stimuli set consisted of two dice presented in different colors showing 1-3 pips, which was not used during training. In the dual task version, participants had to remember the color as well as the number of pips on the dice. Spatial updating A display of three 3x3 was shown for 4000 ms in each of which one blue dot was present in one of the nine locations. Those three locations had to be memorized and updated according to shifting operations, which were indicated by arrows appearing below the corresponding field. Presentation time of the arrows was 3000 ms with an inter-stimulus interval of 250 ms. After six updating operations, the three grids reappeared and the resulting three end positions had to be clicked on (maximum three correct responses per trial). Ten trials were completed. Completion time was approximately 10 minutes.

Average proportion correct per trial
Numerical updating Four single digits (ranging from 1 to 9) were presented simultaneously in four cells situated horizontally for 3000 ms. After an inter-stimulus interval of 250 ms, a sequence of eight updating operations were presented in a second row of four cells below the first one. These updating operations were additions and subtractions within a range -9 to +9. Each updating operation was applied to a different cell from the one a step earlier in the sequence, so that no two updating operations had to be applied to one cell in a sequence. Those updating operations had to be applied to the digits memorized from the corresponding cells above and the updated results had to be memorized and types in at response (maximum four correct responses per trial). Ten trials were completed. Completion time was approximately 10 minutes.

Average proportion correct per trial
Spatial reasoning

Ravens progressive matrices
The task involved completing as many of the 18 trials within a 12-minute time limit. Each trial consisted of increasingly complex visual geometric designs with a missing piece. The task was to select the missing piece from eight alternatives.

WASI-II matrices
The task involved completing as many of the 30 trials within a 15-minute time limit. Each trial consisted of increasingly complex visuospatial objects with a missing piece. The task was to select the missing piece from five alternatives.

Number correct within time limit
BETA-III matrices The task involved completing as many of the 25 trials within an 8-minute time limit. Each trial consisted of increasingly complex visual geometric designs with a missing piece. The task was to select the missing piece from five alternatives.

Number correct within time limit
Verbal reasoning

ETS kit verbal inference
The task involved completing as many of the 20 verbal inference trials as possible within a 12minute time limit. The task involved selecting Number correct within time limit the valid conclusion given the information given in a verbal statement. Example trial: One year a particular farmer's stand of wheat yielded 40 bushels per acre. 1) The farmer's land is extremely fertile 2) The farmer has raised wheat on his land 3) The weather that year was unfavourable for growing wheat 4) Forty bushels per acre is a high yield 5) The field would be more suitable for some other crop

BIS analogies
The task involved completing as many of the 14 analogies within a 5-minute time limit. The task involved finding completing analogies. Example trial: Newspaper is to text as speaker is to 1) sound 2) song 3) news 4) listening 5) stereo Number correct within time limit

Syllogisms
The task involved completing as many of the 30 syllogism trials within a 20-minute time limit. The task on each trial was to decide whether the conclusion that followed the premises was valid or not. Example trial: All trees are fish. All fish are horses. Therefore all trees are horses. 1) valid conclusion 2) invalid conclusion

Number correct within time limit
Episodic memory Spatial memory Sequences of 12 colored photographs of real-world objects were displayed at different locations in a 6x6 grid. Each photograph was presented for 8000ms with an inter-stimulus interval of 1000ms. After presentation, objects appeared at the right side of the screen and had to be moved in the correct order to the correct locations by clicking on the objects and the locations with the computer mouse. Two trials were included in the test (maximum number correct=24).

Number correct
Verbal memory Sequences of 16 nouns were presented in the middle of the screen. Each noun was presented for 6000ms with an inter-stimulus interval of 1000ms.After presentation, the nouns had to be recalled by typing in the word into a blank field. Two trials were included in the test (maximum number correct=32.

Perceptual speed
Perceptual matching 1 Pairs of 4-item letter strings were presented and the task was to decide whether they were identical or not as quickly as possible. The number strings could differ by 1-4 letters. Once a response was made the next stimulus was presented. There were two runs with 50 trials in each run, of which the letter strings differed in 50% of the trials. Participants were instructed to respond as quickly and as accurately as possible. Completion time was approximately 5 minutes.
Average response time for correct trials Perceptual matching 2 The task was identical to Perceptual matching 1, with the exception of the stimuli set that consisted of 4-item number strings.
Average response time for correct trials S5. Descriptive statistics for all cognitive composites at pretest and posttest, separately for each intervention group. Task performance is expressed as T-scores with a mean of 50 and a standard deviation of 10. COG=cognitive training, PE=physical exercise, WM=working memory.
S6. Drop-out analysis comparing age, reasoning ability and self-reported physical activity at pretest for the randomized and the final sample, separately for the four intervention groups. Standardized mean differences (SMD) were calculated as (Mrand -Mfinal)/SDrand. In absolute terms and considering all intervention groups, SMDs were all small and ranged from 0.06 to 0.15 for age, from 0.004 to 0.05 for reasoning ability, and from 0.03 to 0.14, from 0 to 0.

COG+PE PE+COG
Completers (n=21) Completers (n=27) S7. Descriptive statistics on un-transformed BDNF concentrations from pretest, separately for each intervention group. Sample 1 was taken before the first intervention, sample 2 after the first intervention and sample 3 after the second intervention or after rest, all at pretest. COG=cognitive training, PE=physical exercise.
n mean sd n mean sd n mean sd n mean sd Serum (pg/mL) Sample 1 20 25011 4958 26 23964 4979 23 25696 7077 25 24198 5336 Sample 2  S8. Unstandardized parameter estimates from the linear mixed effects models and F-tests of main and interaction effects used to test whether cognition in older adults benefits more from repeated sessions of working memory training when each training session is directly preceded as opposed to followed by physical exercise (H1). Linear mixed models with random intercept were employed for all factors, with the exception for cognitive training progress, for which a random linear slope model was employed. Task performance is expressed as T-scores with a mean of 50 and a standard deviation of 10 for all factors, with the exception of cognitive training progress, which is expressed as raw scores (maximum=24). Intervention = COG+PE < PE+COG; Time = pretest < posttest; Training visit = cognitive training visits 1-32; WM=working memory. Note that the corrected a-level was 0.025 for trained WM (trained stimuli, untrained stimuli) and for untrained WM (updating, switching) and 0.0125 for untrained domains (processing speed, episodic memory, spatial reasoning, verbal reasoning). S9. Unstandardized parameter estimates from the linear mixed effects models and F-tests of main and interaction effects used to test whether cognition in older adults benefits more from repeated sessions of working memory training when each training session is combined with physical exercise, irrespective of order, compared to working memory training alone (H3). Linear mixed models with random intercept were employed for all factors, with the exception for cognitive training progress, for which a random linear slope model was employed. Task performance is expressed as T-scores with a mean of 50 and a standard deviation of 10 for all factors, with the exception of cognitive training progress, which is expressed as raw scores (maximum=24). Intervention = COG < COMB; COMB=PE+COG & COG+PE; Time = pretest < posttest; Training visit = cognitive training visits 1-32; WM=working memory. Note that the corrected a-level was 0.025 for trained WM (trained stimuli, untrained stimuli) and for untrained WM (updating, switching) and 0.0125 for untrained domains (processing speed, episodic memory, spatial reasoning, verbal reasoning). S10. Unstandardized parameter estimates from the linear mixed effects models and F-tests of main and interaction effects used to test whether cognition in older adults benefits more from repeated sessions of physical exercise when each exercise session is combined with working memory training, irrespective of order, compared to physical exercise alone (H4). Linear mixed models with random intercept were employed for all factors. Task performance is expressed as T-scores with a mean of 50 and a standard deviation of 10. Intervention = PE < COMB; COMB=PE+COG & COG+PE; Time = pretest < posttest; Training visit = cognitive training visits 1-32; WM=working memory. Note that the corrected a-level was 0.025 for trained WM (trained stimuli, untrained stimuli) and for untrained WM (updating, switching) and 0.0125 for untrained domains (processing speed, episodic memory, spatial reasoning, verbal reasoning). S11. Unstandardized parameter estimates from the linear mixed effects models and F-tests of main and interaction effects used to test whether the relationship between change in peripheral BDNF levels in response to physical exercise at pretest and the outcome of repeated sessions of working memory training is greater when each training session is directly preceded as opposed to followed by physical exercise (H2). Linear mixed models with random intercept were employed for all factors, with the exception for cognitive training progress, for which a random linear slope model was employed. Task performance is expressed as T-scores with a mean of 50 and a standard deviation of 10 for all factors, with the exception of cognitive training progress, which is expressed as raw scores (maximum=24). Separate analyses were performed for serum and plasma. Intervention = COG+PE < PE+COG; Time = pretest < posttest; Training visit = cognitive training visits 1-32; BDNF=acute concentration change following physical exercise at pretest; WM=working memory. Note that the corrected alphalevel was 0.0125 for trained WM (trained stimuli, untrained stimuli) and for untrained WM (updating, switching), 0.00625 for untrained domains (processing speed, episodic memory, spatial reasoning, verbal reasoning) and 0.025 for cognitive training progress. S12. Scatterplots visualizing individual data points and zero-order correlations between training gains (pre-post change in performance on trained tasks with untrained stimuli) and BDNF concentrations in sample 1 (A), sample 2 (B) and sample 3 (C) at pretest, separate for the group that received physical exercise only during the intervention (PE, in red) and the group that received cognitive training only during the intervention (COG, in blue).