Effect of brain-derived neurotrophic factor gene polymorphisms on motor performance and motor learning: A systematic review and meta-analysis

Brain-derived neurotrophic factor (BDNF) gene polymorphisms may modulate neurotransmitter efficiency, thereby influencing motor performance and motor learning. However, studies to date have provided no consensus regarding the genetic influence of BDNF genotypes (i.e., Val/Val, Val/Met, or Met/Met type). This study aimed to investigate the effect of BDNF genotype on motor performance and motor learning in healthy human adults via a systematic review and meta-analysis. A total of 19 relevant studies were identified using PubMed and Web of Science search for articles published between 2000 and 2021 with motor performance or motor learning as the primary outcome measures. The results of our systematic review suggest that the BDNF genotype is unlikely to contribute to motor performance and motor learning abilities because only 2/32 datasets (6.3%) from 16 studies on motor performance and 3/19 datasets (17.6%) from 13 studies on motor learning indicated a significant genetic effect. Moreover, a meta-analysis of motor learning publications involving 17 datasets from 11 studies revealed that there was no significant difference in the learning score normalized using baseline data between Val/Val and Met carriers (Val/Met + Met/Met or Val/Met; standardized mean differences = 0.08, P = 0.37) with zero heterogeneity (I2 = 0) and a relatively low risk of publication bias. Taken together, the BDNF genotype may have only a minor impact on individual motor performance and motor learning abilities.


Introduction
Brain-derived neurotrophic factor (BDNF), which is widely distributed in the human brain, plays a role in the central nervous system and has been investigated both in vivo and in vitro. The BDNF protein binds and activates tropomyosin-related kinase B receptors, which promotes neural survival and synaptic function in the brain [1,2]. BDNF expression can enhance glutamatergic and gamma aminobutyric acid (GABA) ergic synaptic efficacies [3][4][5][6][7], which contribute to neural plasticity associated with learning and memory processes [1,8].
Several single-nucleotide polymorphisms (SNPs) have been identified in the human BDNF gene. These SNPs have been observed at the frequencies of approximately 0-70%, depending on the geographic region covered by the study [9]. Val is replaced by Met at codon 66 (Val66Met, rs6265) in one of these SNPs in the BDNF Met allele. As a result, BDNF secretion in the hippocampus of Val/Met and Met/Met mice is significantly reduced by 18% and 29%, respectively, resulting in reduced synaptic transmission efficacies via glutamate and GABA [10,11]. As these are the main brain neurotransmitters, this genetic difference may influence various brain functions. For example, stroke patients with BDNF polymorphisms showed delayed recovery from motor deficits compared with those with Val/Val polymorphism [12]. In addition, healthy participants with BDNF polymorphisms had poorer episodic memory than those with Val/Val polymorphism [13]. In terms of pathology, the risk of Alzheimer's disease and depression increases in people with BDNF polymorphisms [14,15]. Although these findings remain controversial, they suggest that decreased BDNF secretion in people with genetic polymorphisms influences synaptic efficacy, Abbreviations: BDNF, Brain-derived neurotrophic factor; GABA, Gamma aminobutyric acid; Non-Val/Val, Val/Met + Met/Met; Met carriers, Non-Val/Val or Val/ Met; PRISMA, Preferred reporting items for systematic review and meta-analysis; SD, Standard deviation; SMD, Standardized mean differences; SNP, Singlenucleotide polymorphisms. resulting in impaired behavior and cognitive skills.
Although BDNF-related genetic effects have been widely investigated in the fields of cognition and psychiatry via meta-analyses [15][16][17][18][19], the genetic effects of BDNF polymorphisms on motor function are not adequately understood. Motor performance and motor learning are vital to the activities of daily living; however, considerable interindividual variability in motor ability exists. In general, the ability to acquire new motor skills has been shown to be strongly associated with neural plastic changes in both the structural and functional organization of the primary motor cortex (M1) [20,21]. Furthermore, transcranial magnetic stimulation research in which M1 excitability was evaluated using motor-evoked potentials revealed that motor-skill learning induces short-term changes in M1 excitability [22][23][24], suggesting that M1 plays an important role in the acquisition of new motor skills. However, other brain regions, including the cerebellum, frontal cortex, posterior parietal cortex, and striatum, can also contribute to the acquisition in a complex manner via cortico-cortical connections depending on the learning stage [25]. For example, the premotor cortex integrates visual and sensorimotor information and prepares motor output for M1 [25]. The parietal cortex integrates somatosensory and visual information and prepares multimodal sensorimotor input for the premotor cortex [25]. The BDNF genotype may be one of the factors that determine motor ability via impaired synaptic changes in motor-related brain networks. However, previous studies that investigated the effect of this genotype on motor performance and motor learning failed to produce consistent results [26][27][28]. In our previous study, we used magnetic resonance spectroscopy as a reliable method to measure the concentrations of neurometabolites; accordingly, we reported lower glutamate + glutamine concentrations (as an index of the excitatory neurometabolites in M1) in Non-Val/Val (Val/Met + Met/Met) subjects compared with those in Val/Val subjects [29]. Furthermore, reduced BDNF expression in the layer II/III and layer V neurons of M1 was shown to impair motor learning in 3-month-old mice [30]. These findings suggested that the BDNF genotype is likely to influence motor-related brain networks at the physiological level; thus, BDNF polymorphisms may eventually affect individual motor ability.
Given that there are high interindividual differences in motor function, the specific BDNF genotype may partially contribute to these differences. Here, we investigated the effects of the BDNF genotype on motor function via a systematic review and meta-analysis of data from healthy adult humans to summarize the results from previously published articles. This review aimed to contribute to the further understanding of the physiological processes involved in motor performance and motor learning associated with the BDNF genotype.

Protocol
This systematic review and meta-analysis was conducted according to the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-analyses for Protocols 2015 (PRISMA-P 2015) [31].

Search strategy and study selection
A search of the literature published between January 1, 2000, and July 31, 2021, was performed in the PubMed and Web of Science databases on July 31, 2021. The following search terms were included in these combinations: "motor learning" OR "motor performance" OR "motor skill" OR "motor acquisition" AND "brain-derived neurotrophic factor" OR "BDNF" OR "Val66Met." A manual search of the reference sections in the retrieved studies was performed.

Eligibility criteria
The selected studies were required to meet the following eligibility criteria: 1) Peer-reviewed published studies in the English language 2) Healthy adult humans (aged > 18 years but not the elderly group, i. e., > 60 years old) 3) Formal ethical approval 4) Motor performance or motor learning tasks classified as per the BDNF genotype groups 5) Motor tasks involving the upper limbs The titles and abstracts of the articles were initially screened by one reviewer (R.S.). Eligibility for inclusion was determined independently by two reviewers (R.S. and S.M.), who assessed the full text against the eligibility criteria.

Data extraction and assessment
The selected articles classified subjects into three genotype groups (i. e., Val/Val, Val/Met, and Met/Met) or two genotype groups [i.e., Val/ Val and Non-Val/Val (Val/Met + Met/Met) or Val/Val and Val/Met]. Some selected articles had merged the Val/Met and Met/Met groups into one group as the Non-Val/Val group. Motor performance or motor learning was investigated as the main outcome measure. In this review, motor tasks at the baseline or motor task without blocks that did not include an element of practice and learning were defined as motor performance; by contrast, motor tasks with multiple blocks including an element of practice and learning were defined as motor learning. However, the retention phase (> 24 h) of motor learning was removed from the data extraction process because some of the selected articles did not include data collected at the retention phase. Therefore, motor learning that included the acquisition and consolidation phases (< 24 h) was extracted for the meta-analysis. As the variables for motor performance were largely different among studies, motor performance was not assessed via meta-analysis. In the systematic review, we evaluated the genetic differences in motor performance among the genotype groups based on the statistical data presented in each study to determine the percentage of articles that showed significant differences by genotype. If a study contained baseline data on a motor learning task, we included the article to collect the motor performance data. The majority of articles compared the genetic differences between motor performances using the analysis of variance (ANOVA) or t-test. P values of < 0.05 among the genotype groups were determined to indicate a significant difference. Several articles included the results of a motor learning task with timeline data (i.e., a two-way ANOVA with "GROUP" and "TIME" factors), and thus the motor learning task variable at baseline was not compared purely among the genotypes. In this case, we determined that there was a significant difference among the groups if the significant main effect of the group was observed without a significant effect from the "GROUP" × "TIME" interaction or based on a post-hoc test. Finally, we classified the results into "Not significant" or "Significant difference," after which we calculated the percentage of significant differences in the overall results.
Motor learning was assessed via systematic review and metaanalysis. First, we assessed differences in the motor learning ability and motor performance among the genotype groups based on the statistical data presented in each study. Significant difference among the groups was indicated if the P values were < 0.05 for the main effect of the "GROUP" and "GROUP" × "TIME" interaction. We assessed the detailed genetic difference based on a post-hoc analysis. A few articles simply compared the motor learning ability among groups without multiple block data (i.e., pre and post) using t-test, according to the Pvalue. Finally, the percentage of significant differences according to the genotype was calculated in addition to the motor performance. Subsequently, the variables for motor learning were extracted from the main text, figures, tables, or supplementary data to perform the meta-analysis. If graphs were used instead of reporting the original data, the data were extracted using GetData Graph Digitizer (v. 2.26.0.20), which can digitize scanned graphs and obtain the original data. Previous metaanalyses were also included [32,33]. Standard deviation (SD) was calculated when the standard of error of the mean was used with mean data according to Eq. (1) [34]: where N indicates the number of subjects in each genotype group. The motor learning score and SD of the first and last blocks or percentage change mean and SD measured within 24 h were extracted for each group. In the meta-analysis, when the BDNF genotype was classified into three groups in the selected articles, the data for the Val/Met and Met/ Met groups were merged to produce a Non-Val/Val group. However, some articles did not recruit Met/Met participants. Therefore, the groups tested against the Val/Val genotype in the meta-analysis were Non-Val/Val (Val/Met + Met/Met) or Val/Met only, namely, the Met carrier groups (i.e., Non-Val/Val or Val/Met). The percentage change in motor learning was expressed using Eq. (2) or (3) as follows: Eq.
(2) was used when a decreased learning score represented an improved motor performance, whereas Eq. (3) was used when an increased learning score represented an improved motor performance. A positive value for motor learning expressed an improved motor performance, whereas a negative value for motor learning expressed a worse motor performance across the blocks.

Meta-analysis
Comprehensive Meta-Analysis Software Version 3.0 was used to calculate the effect size and 95% confidence interval (CI) using a random-effect model. The meta-analysis compared the motor learning ability between the Val/Val and Met carrier groups. The mean, SD, and number of subjects were used as inputs to compute the standardized mean differences (SMD) using Hedges' adjusted g, which is similar to Cohen's d but includes an adjustment for small sample bias [35]. The SMD value was interpreted as small (0.2 ≤ 0.49), medium (0.5 ≤ 0.79), or large (≥ 0.8), as per the convention suggested by Cohen [36]. The sample size of each study influenced the weight given to its mean and SD in the analysis. A Z-statistic and P values were calculated to determine whether the mean effect size was significantly different among the genotype groups. The value for statistical significance was set to P < 0.05.

Test for heterogeneity
We utilized the Cochran's Q, P-value, and I 2 test to determine any differences underlying the included study results. The power of Cochran's Q is low when fewer studies are used. By contrast, the I 2 value is a more reliable measure for analyzing heterogeneity among the results of the studies included in the analysis [37,38]. The I 2 statistics quantify heterogeneity in the range of 0-100%, with 0% indicating no heterogeneity and > 50% representing substantial heterogeneity [37].

Risk of bias
The risk of bias was assessed using the Cochrane Collaboration tool [39], which is used to estimate (a) random sequence generation (selection bias), (b) allocation concealment (selection bias), (c) blinding of participants and researchers (performance bias), (d) blinding of outcome assessment (detection bias), (e) incomplete outcome data (attrition bias), (f) selective reporting (reporting bias), and (g) other sources of bias in the individual studies under assessment. The risk of bias was categorized as low, unclear, or high according to the assessment method [39]. If the bias appeared to severely affect the result, we judged it as high risk, e.g., when blinding was lacking for both participants and researchers. However, in cases where relevant information was not provided in the article, we judged the risk as unclear. Furthermore, if the bias appeared to slightly affect the results or the outcome was uncertain, we regarded the risk as unclear. An example included the use of a single-blind design, a couple of subjects being removed from the analysis, or lack of clarity on whether sequences were in a randomized order in a study. Low risk was attributed if the risk of bias was not observed, or when the study did not involve the risk of bias owing to the experimental design. An example included all prespecified outcomes having been reported or the use of a double-blind design in a study (for detailed information, see Higgins et al. [39]). Finally, the percentage of bias in the selected articles was calculated for each category using a specific website (https://mcguinlu.shinyapps.io/robvis/).

Publication bias
Possible publication bias was assessed using a funnel plot with the trim-and-fill method [40,41] and by performing Begg's adjusted rank correlation test [42] and Egger's regression test [43].

Results
Fig . 1 shows the PRISMA flow chart. Electronic literature searches identified a total of 3771 studies that matched the search terms. After removing duplicates, 1174 studies remained. An initial screening of titles and abstracts was performed against the eligibility criteria, and in case of insufficient information, the full-text articles were assessed. In total, full-text versions of 32 articles were screened for eligibility, and 19 studies were ultimately included in the systematic review. Selected articles were then categorized into two groups (i.e., motor performance or motor learning) based on the motor task. A total of 32 datasets were obtained from 16 studies on motor performance, whereas 19 datasets were obtained from 13 studies on motor learning. Each dataset indicates that one motor task and genotype group were included. The metaanalysis included a total of 11 studies (17 datasets) associated with a motor learning task. Although van der Vliet et al. [44] used four learning tasks (eyeblink conditioning, vestibulo-ocular reflex, saccade adaptation, and visuomotor adaptation tasks), only the visuomotor adaptation task was assessed in the present study because the other tasks included an element of reflex due to involuntary motion.

Risk of bias
The risk of bias, as judged by the authors, is presented in Fig. 2A and B. The risk was low to unclear in both motor performance and motor learning. The experimenter was blinded to the BDNF genotype in 4 of 16 studies on performance and 4 of 13 studies on motor learning [45][46][47][48]. However, the remainder (and the majority) of the studies did not provide this information. These studies did not provide adequate information about blinding during the outcome assessment. One study in both categories blinded the researchers to the participant's BDNF genotype when they analyzed the motor task data [47], but these data were not added to this review owing to the eligibility criteria. Table 1 summarizes the participant and methodological characteristics of the selected motor performance studies. Only 2 of 32 datasets (6.3%) from 16 studies revealed a significant difference based on the individual study results [27,28]. González-Giraldo et al. [28] reported that Non-Val/Val participants showed better performance compared with Val/Val homozygous participants. By contrast, McHughen et al. [27] reported that Val/Val participants performed better as compared with Val/Met participants. Table 2 presents a summary of the participant and methodological characteristics of the selected studies on motor learning. Percentage changes in motor learning identified from selected articles are displayed in Fig. 3. Only 3 of 19 datasets (17.6%) from 13 studies showed significant differences based on the results of each study [47][48][49]. Two studies revealed that Val/Val homozygous participants performed better as compared with Val/Met or Non-Val/Val participants [47,48]. By contrast, the third study reported that Met/Met participants performed better as compared with Val/Val and Val/Met participants [49]. Although McHughen et al. [27] reported that motor learning in the Val/Val participants was better than that in the Val/Met participants, we judged there to be no significant difference in the motor learning ability, because the main effects of the genotype group and time on motor learning, including the baseline data, were observed, but no significant interaction was included in the assessment.

Effect of the BDNF genotype on motor learning from the perspective of meta-analysis
A funnel plot was generated to examine the results of the metaanalysis for the evidence of publication bias (Fig. 2C). The trim-andfill method estimated the overall effect size as 0.00 with two imputed data points, which was approximately the same as the original value (0.08). Furthermore, Begg's adjusted rank correlation test (Tau = 0.08, P = 0.97) and Egger's regression test did not indicate any evidence of publication bias (t = 0.28, P = 0.78). Overall, the potential risk of publication bias was not observed in either analysis.

Discussion
To the best of our knowledge, this is the first study that systematically evaluated the effects of the BDNF genotype on motor performance Fig. 1. PRISMA flow chart of the analysis. One data point indicates that one motor performance or motor learning task and the associated genotype groups were included. and motor learning. Surprisingly, the systematic analysis of the results of the individual studies revealed that there was no significant difference in motor performance and motor learning among the genotype groups in approximately > 85% of the data. Moreover, the meta-analysis revealed that motor learning did not vary between the Val/Val and Met carrier groups, with zero heterogeneity and relatively low publication bias. Taken together, we found the impact of BDNF genotype to be more limited than expected at the behavioral level.
The present systematic review indicated that 6.3% of the motor performance data and 17.6% of the motor learning data varied significantly among the genotype groups according to the statistical results of each study. In terms of motor performance, González-Giraldo et al. [28] reported that the Non-Val/Val group performed better than the Val/Val group using the pursuit rotor task. By contrast, McHughen et al. [27] indicated that there was a better performance from the Val/Val group than that from the Val/Met group. The participants performed three  motor tasks, including ballistic movement, pegboard, and pinch tasks, in two experiments, respectively [27]. However, a genetic difference was found only for the ballistic movement task in experiment 2. Considering these data, the genetic effect is unlikely to contribute to motor performance. Only 3 of 19 selected datasets for motor learning reported a significant genetic effect in the systematic review. Motor learning in the Val/Val group was superior to that in the Met carriers in two datasets [47,50], but Met/Met groups showed better performance than the other groups in one dataset [49]. In addition, the meta-analysis of motor learning revealed no significant difference among the BDNF genotype groups. Therefore, we suggest that the BDNF genotype is unlikely to contribute to both motor performance and motor learning abilities in healthy adults.
We initially expected that motor function may be influenced by the BDNF genotype based on the brain function differences related to genotype. For example, 18% and 29% lower BDNF expression was observed in Val/Met and Met/Met mice, respectively, in the hippocampus relative to the expression in Val/Val mice [51]. Furthermore, N-methyl-D-aspartate and GABA receptor-dependent synaptic transmissions were impaired in the hippocampal and prefrontal areas of Met/Met mice [10,11]. These findings implied that the BDNF genotype affects brain function, including synaptic transmission. Various brain areas, including M1, cerebellum, frontal cortex, posterior parietal cortex, and striatum, contribute to motor performance and motor learning via cortico-cortical connections, [25]. Our recent study on the genetic effects on the M1 area indicated that the Non-Val/Val group had lower levels of excitatory neurometabolites than the Val/Val group; however, this was not the case in S1 and cerebellum [29]. A functional magnetic resonance imaging study reported a lower activation volume of M1 in the Val/Met group than that in the Val/Val group during hand movement [27]. However, some motor-related cortical areas have been shown to functionally and structurally differ among the genotype groups [52][53][54][55] but not only in M1. Therefore, although genetic differences affect multiple motor-related networks associated with motor performance and motor learning, the functional changes may not become apparent at the motor behavior level. By contrast, our recent study reported that learning score in the Met/Met group was better than that in the Val/Val and Val/Met groups [49]. The learning speed across 10 blocks in the Met/Met group was significantly faster than that in the other groups; however, no significant differences among the groups were noted based on two-factor mixed model ANOVA (fixed effects: learning block and genotype group) [49]. In addition, our subsequent analysis and meta-analysis demonstrated that the motor learning score did not vary among the groups in the last block. By contrast, Joundi et al. [47] reported motor task-related genetic effects on motor learning. They found that the genetic difference was observed more prominently when undertaking a more difficult motor learning task than an easier one. As another possibility, the task type may also influence genetic differences. Motor performance was assessed between genotypes based on statistical data [i.e., S.D., N.S., and Not available (− )]. Motor performance was extracted from motor learning tasks at baseline in some articles. Abbreviations: BDNF, brain-derived neurotrophic factor; DZ, dizygotic twins; Non-Val/Val, Val/Met + Met/Met; N.S., not significant; MZ, monozygotic twins; S.D., significant difference. For example, some motor tasks, such as visuomotor tracking and pegboard tasks, might demand the use of more cognitive skills than needed in other tasks, such as the simple ballistic movement task; thus, the required brain networks could differ among tasks. A functional magnetic resonance imaging study reported that simple finger tapping activated motor-related cortical areas including M1, primary somatosensory cortex, and cerebellum, but did not activate high-order cortical areas [56]. By contrast, a visuomotor adaptation task was shown to activate the dorsolateral prefrontal cortex, which contributes to spatial working memory, suggesting that the task type influences the absence or presence of cognitive skills during recruitment [57]. Because cognitive ability may also be impaired in the Non-Val/Val group [18,53,58], BDNF polymorphisms might affect some motor tasks that demand cognitive skill. The present review summarized a variety of motor tasks; therefore, we did not investigate the effects of task-related genetic differences, including task difficulty and task type, on motor performance and motor learning. Nevertheless, genetic differences in motor function could potentially be pronounced depending on such factors. Moreover, compensation for impaired motor-related networks associated with motor performance and motor learning, may occur. For example, in a structural magnetic resonance imaging study, larger cortical volumes, including in the lateral occipital/superior parietal, precentral, and postcentral gyri, were observed in the Non-Val/Val children group relative to those in the Val/Val children group [55]; this effect may compensate for reduced synaptic transmission in the motor-related cortical areas.
We evaluated the risk of bias with regard to motor performance and motor learning groups based on the Cochrane Collaboration tool [39]; however, none of the categories were found to have a high risk of bias. Nonetheless, many studies lacked clarity on whether the researchers were blinded to the BDNF genotype when collecting and analyzing data. One epidemiology study reported that a lack of blinding may exaggerate the obtained effect size by approximately 25% [59]. However, despite this lack of information, the majority of the studies on motor performance and motor learning did not report a genetic difference, suggesting that any exaggeration effect is likely to have been small. By contrast, our meta-analysis included some overlapping subjects in the same task or in different tasks, which may have inflated the false positive rate of the meta-analysis [60]. However, because no effect was observed between the genotype groups, the overlapping subjects apparently did not induce a type 1 error in the results.
According to previous study, an asymmetric funnel plot reveals publication bias and identifies low-quality studies [40]. Furthermore, the trim-and-fill method estimates the missing studies to maintain a symmetrical funnel plot [41]. In our study, the effect size was 0.08, whereas the corrected effect size was 0, and thus, a relatively small difference existed between the values, suggesting that the effect of publication bias was small. However, a funnel plot finalized using the trim-and-fill-method is not the ideal method for assessing publication bias because this method may underestimate the positive effect if high heterogeneity exists among the studies [61]. To avoid this effect, we used additional methods to evaluate publication bias, including Begg's adjusted rank correlation test [42] and Egger's regression test [43]. Neither of these tests showed a significant difference, suggesting that the possibility of publication bias existing was low and corroborating the funnel plot. The I 2 value indicates the degree of inconsistency across studies in a meta-analysis and I 2 statistics are preferably used to test for heterogeneity when assessing the consistency of evidence [37]. In the present study, the I 2 statistics indicated zero heterogeneity, i.e., there was low variability among the results from the included studies.
There are several limitations to this study. First, because we summarized a variety of motor tasks, the task-dependent effect of the BDNF genotype was not investigated. Motor learning task-dependent M1 plasticity has previously been reported [62,63], but a further study  [64]. Third, although the effects of genotype on motor performance and learning were not observed in this review, an interaction between genotype and a specific factor (such as sex) might strengthen the genetic effect. A meta-analysis showed that there was a sex difference in the allelic association; the Met66 allele confers susceptibility to Alzheimer's disease in women but not in men [15]. Furthermore, the interaction of the BDNF genotype and sex in relation to motor performance was also observed; a more pronounced genotype effect on motor performance was shown in Val/Val women compared with this effect in Val/Val and Non-Val/Val men and Non-Val/Val women [65], and thus, the genotype effect, which was not investigated here, may become more apparent for changes in the motor system according to the interaction. Fourth, our analysis methods might have influenced our interpretations of genetic effects. When a significant interaction of group and time was observed in an article, we deduced that there was a significant difference among the BDNF genotypes on motor learning to express how many articles revealed significant differences by percentage. Thus, a significant interaction of group and time was observed in an article to express how many articles showed significant differences by a percentage; thus, our method may underestimate the true effect. However, we believe that this effect is relatively small because the meta-analysis similarly showed no effect. Finally, because motor learning was assessed within the first 24 h in this review, long-term effects including the retention phase, were not investigated. Different brain areas are involved in different learning stages, such as acquisition, consolidation, and retention phases [25], which may have led to learning stage-dependent changes in motor learning in the BDNF genotype group. Deveci, et al. [66] reported that a Val/Val group produced a higher basketball score at day 5 than that achieved by a Non-Val/Val group. Perhaps, therefore, the BDNF genotype may have a greater influence on motor learning ability during the retention phase rather than in earlier phases such as acquisition and consolidation. However, further research will be required to reveal the existence of learning stage-dependent effects related to the BDNF genotype.

Conclusions
There is growing evidence that the BDNF genotype affects cognition and influences the onset of brain disorders [16][17][18][19]. Surprisingly, the present review showed that genetic effects are unlikely to contribute to motor performance and motor learning in healthy human adults, although motor-related cortical networks may differ among the genotype groups at the physiological level [26,27,29]. This mismatch at the neural and behavioral levels suggests that the genetic-related neural changes do not affect behavior or are compensated for by brain function in healthy adults. However, delayed motor recovery from stroke is reported in patients with BDNF polymorphisms [12]. Thus, further clinical research on motor function related to the genetic effects of the BDNF genotype is required.