A deep phenotyping approach to assess the association of handedness, early life factors and mental health

The development of handedness and other form of functional asymmetries is not yet understood in its critical determinants. Early life factors (e.g., birth weight, birth order) have been discussed to contribute to individual manifestations of functional asymmetries. However, large-scale data such as the UK Biobank suggest that the variance in handedness that is explained by early life factors is minimal. Additionally, atypical handedness has been linked to clinical outcomes such as neurodevelopmental and psychiatric disorders. Against the background of this triad, the current study investigated associations between different forms of functional asymmetries and (a) early life factors as well as (b) clinical outcomes. Functional asymmetries were determined by means of a deep phenotyping approach which notably extends previous work. In our final sample of N = 598 healthy participants, the different variables were tested for associations by means of linear regression models and group comparisons (i.e., ANOVAs and Chi-squared tests). Confirming previous findings from larger cohorts with shallow phenotyping, we found that birth factors do not explain a substantial amount of variance in functional asymmetries. Likewise, functional asymmetries did not seem to have comprehensive predictive power concerning clinical outcomes in our healthy participants. Future studies may further investigate postulated relations in healthy and clinical samples while acknowledging deep phenotyping of laterality.

A deep phenotyping approach to assess the association of handedness, early life factors and mental health Lena Sophie Pfeifer 1,6* , Judith Schmitz 2,6 , Maike Schwalvenberg 3 , Onur Güntürkün 3 & Sebastian Ocklenburg 3,4,5 The development of handedness and other form of functional asymmetries is not yet understood in its critical determinants.Early life factors (e.g., birth weight, birth order) have been discussed to contribute to individual manifestations of functional asymmetries.However, large-scale data such as the UK Biobank suggest that the variance in handedness that is explained by early life factors is minimal.Additionally, atypical handedness has been linked to clinical outcomes such as neurodevelopmental and psychiatric disorders.Against the background of this triad, the current study investigated associations between different forms of functional asymmetries and (a) early life factors as well as (b) clinical outcomes.Functional asymmetries were determined by means of a deep phenotyping approach which notably extends previous work.In our final sample of N = 598 healthy participants, the different variables were tested for associations by means of linear regression models and group comparisons (i.e., ANOVAs and Chi-squared tests).Confirming previous findings from larger cohorts with shallow phenotyping, we found that birth factors do not explain a substantial amount of variance in functional asymmetries.Likewise, functional asymmetries did not seem to have comprehensive predictive power concerning clinical outcomes in our healthy participants.Future studies may further investigate postulated relations in healthy and clinical samples while acknowledging deep phenotyping of laterality.
Functional asymmetries are widespread across species 1,2 and can be found in simple motor tasks 2 , complex socio-behavioral patterns 3,4 , and in cognitive information processing 5,6 .For humans, the most obvious form of lateralization is handedness.Importantly, human handedness is not only asymmetric on an individual level but also on a population level.More precisely, a recent meta-analysis estimated that 10.6% of the population is left-handed 7 .Despite such clear evidence on the phenotypic level, it is still largely unknown in how far genetic and environmental factors contribute to the development of individual handedness and other forms of functional asymmetries 8 .Similarly, it is an open question how far the ontogenesis of a lateralized brain may overlap with developmental pathways of psychopathology.In this regard, several neurodevelopmental and psychiatric disorders have been associated with atypical lateralization 9 .Unraveling causal mechanisms in the development of structural and functional asymmetries may hence have clinical relevance.
Early (mono-)genetic theories on the development of handedness 10,11 have been refuted as being too simplistic 12 .Likewise, candidate genes initially found to show associations with handedness could often not be replicated 13 .Along these lines, twin studies confirm that genetic factors explain about a quarter of the variance in human handedness 14,15 .Twin studies estimate the additive genetic heritability of a trait by comparing phenotypic concordance between monozygotic and dizygotic twins.Genome-wide association studies (GWAS) take a molecular approach in that millions of single nucleotide polymorphisms (SNPs) are tested for an association with the phenotype of interest.The largest GWAS on handedness so far (N = 1,766,671) suggested common SNPs www.nature.com/scientificreports/allow to capture different evolutionary pressures and may be better suited for research across species 3 .From a dimension of measurement, it may be claimed that handedness, the most common proxy for hemispheric asymmetries, is commonly not assessed accurately.In contrast, handedness is often only deduced from a unidimensional measure.In its most extreme, this rationale can be found in the assessment of handedness in terms of only one item that typically asks for writing hand.Thereby handedness is treated as a binary concept and only refers to one manual task (i.e., writing).Of note, many studies in the field of laterality have recognized this issue and satisfy an assessment of handedness using several items.For instance, many researchers use the Edinburgh Handedness Inventory (EHI 44 ), a questionnaire that queries hand preference for various manual tasks (e.g., handling a knife, brushing teeth).Such an approach ultimately allows the calculation of a laterality quotient (LQ) which satisfies handedness as a continuous variable.Still, for the EHI as well as for the simple assessment of writing hand, research broadly relies on self-reported preference measures.Only a few studies further integrate performance measures that assess hand skill.A promising example of such a performance measure is the Pegboard task 45 , which requires participants to place several pegs initially stuck in a straight row of holes on a board in a second parallel row of other holes as quickly as possible.As this is done with both hands consecutively while reaction time is taken, it is possible to compare performance of the left and the right hand.Using the Pegboard task rather than a shallow handedness phenotype has resulted in the identification of the first genetic variants associated with handedness in GWAS 46 .Ideally, to come to a preferably differentiated picture of an individual's lateralization, studies may combine different measures of self-reported hand preference and measures of hand skill as well as measures of other forms of functional asymmetries (e.g., language lateralization).This notion may be especially relevant considering that different performance measures of handedness (the Pegboard task amongst them) have been revealed to show only small correlations among each other and may reflect distinct dimensions of asymmetries 47 .Undoubtedly, a comprehensive assessment of functional asymmetries-which we refer to as deep phenotyping-may not be applicable for larger-scale studies.However, even though the rigid focus on handedness as a sole proxy for functional asymmetries alongside its unidimensional measurement may have become some kind of common minimal standard, a deeper phenotyping of functional lateralization may be indicated to achieve further progress in laterality research 3 .
Capturing functional lateralization phenotypes by means of deep phenotyping, this study aims to further accumulate knowledge on how environmental factors (i.e., birth factors) play a role in the ontogenesis of handedness and other forms of functional lateralization.Therefore, the first part of the study may be considered a replication approach of the findings by de Kovel et al. 24 .Second, we aimed at understanding associations between handedness and other forms of functional lateralization with subclinical tendencies of several mental disorders in a healthy sample.

Sample.
We recruited healthy participants between 18 and 35 years with German language skills sufficient for understanding questionnaires and instructions given in our study.Moreover, all participants were of Central European ancestry.Ancestry was assessed by means of self-report inquiring the country of descent of participants as well as all parents and grandparents.Individuals reporting Central European ancestry for all three generations were eligible to participate.In our definition, Central European ancestry covered all Northern, Western and Southern Europe, including Spain, while Portuguese descent was excluded.We included Polish and Russian ancestry, but excluded individuals of Southeast European descent (i.e., Turkey and Greece).With respect to handedness, we had no specific inclusion and exclusion criteria but we aimed for a balanced ratio of all handedness categories so that we specifically enrolled left-and mixed-handed individuals.Thus, we over-selected participants with atypical handedness in order to improve statistical power and approach variance homogeneity in statistical analyses.Study advertisement only indicated that the study investigated handedness and did not reference handedness and mental health.Since we also excluded participants reporting psychopathological conditions, we do not believe the results of the current study to be biased by participant recruitment and advertising.In total, we tested N = 631 participants.
This study was approved by the local ethics committee of the Faculty of Psychology at Ruhr University Bochum, Bochum, Germany.All participants gave written informed consent and were treated in accordance with the declaration of Helsinki.
Procedure.Data collection took place between 11/04/2018 and 14/10/2022.Having given informed consent, participants completed an online survey asking for above-mentioned inclusion and exclusion criteria as well as for several factors surrounding their birth (e.g., birth weight, mother's health, breastfeeding).Eligible participants were then invited for testing at Ruhr University Bochum.Testing sessions started with a second online survey including the Edinburgh Handedness Questionnaire (EHI 44 ) and the Waterloo Footedness Questionnaire (WFQ 48 ) as self-report asymmetry measures of handedness and footedness, respectively.Moreover, participants completed validated German versions of the following clinical questionnaires: the Beck's Depression Inventory (BDI; English original 49 ; German version 50 ), the Adult ADHD Self-Report Scale Symptom Checklist (ASRS -v1.1;English original 51 ; German version 52 ), the State-Trait Anxiety Inventory-Trait (STAI-T; English original 53 ; German version 54 ), the Childhood Trauma Questionnaire (CTQ; English original 55 ; German version 56 ) and the Schizotypal Personality Questionnaire (SPQ; English original 57 ; German version 58 ).Finally, participants performed various hand skill tasks including the Pegboard task 45 , the Alphabet test 59 , and the Tapley-Bryden test 60 .
Moreover, language lateralization was assessed using a Dichotic listening task (DLT 61 ) and lateralization for visual attention/visuo-spatial perception was assessed using a Line bisection task 62  Laterality quotients (LQs) were calculated by means of the following formula: LQ = [(right − left)/ (right + left)] × 100.Using the EHI and the WFQ to create categories of left-, mixed-, and right-handedness/footedness, we defined scores of < = − 60 as left-handed/left-footed and scores of > = + 60 as right-handed/rightfooted.Participants scoring between these cutoffs were classified as mixed-handed/mixed-footed. Behavioral asymmetry tasks (e.g., Alphabet test, Line bisection) as well as clinical questionnaires were analyzed according to corresponding manuals.

Statistical analysis.
Statistical analysis was conducted in R version 4.1.2(2021-11-01) and RStudio.The manuscript was prepared using the papaja package 63 .R scripts used for analysis can be retrieved from the Open Science Framework (https:// osf.io/ nkem6/).
We grouped measured variables in three conceptual categories: (1) birth factors, (2) asymmetry measures, and (3) clinical questionnaires.Quantitative asymmetry measures, birth factors, and clinical questionnaire scores were transformed to normality using the bestNormalize() function 64 , which tests different normalizing procedures and applies the one with the best outcome.For details, see supplementary material ("Methods" section, Figs.S1 to S6).Due to intercorrelations of the variables (shown in the supplementary material, Figs.S7 to S9), the number of effective tests was estimated using the meff() function from the poolr package 65 for each set of variables.Table 1 summarizes variables included in our statistical analysis after data transformation including number of effective tests.
We applied different statistical models to analyze hypothesized relations between these variables.After descriptive statistics (Part 1), we modeled asymmetry measures as a function of birth factors (Parts 2-5).Subsequently, we modeled clinical questionnaires as a function of asymmetry measures (Parts 6 and 7).Therefore, different variables (binary/categorical vs. quantitative) served either as predictor or outcome variables (Table 2).Table 1.Overview over asymmetry measures, birth factors, and clinical questionnaires included as predictors or outcomes in our statistical analysis alongside their scale level (binary/categorical vs. quantitative).EHI Edinburgh Handedness Questionnaire, LQ Laterality quotient, WFQ Waterloo Footedness Questionnaire, DLT Dichotic listening task, L-M-R left-mixed-right, BDI Beck's Depression Inventory, ASRS Adult ADHD Self-Report Scale Symptom Checklist, STAI-T State-Trait Anxiety Inventory-Trait, CTQ Childhood Trauma Questionnaire, SPQ Schizotypal Personality Questionnaire, M number of measured variables, M eff Since measured variables within one set are highly intercorrelated (Figs.S7 to S9), the number of effective tests was estimated using the meff() function from the poolr package.www.nature.com/scientificreports/FDR correction was applied to adjust for multiple comparisons using the product of the number of effective tests for predictors and outcomes (e.g., in Part 2, we applied FDR correction for 5 (quantitative asymmetry measures) × 4 (quantitative birth factors) independent tests).For significant effects of ANOVAs concerning non-binary outcome measures, we continued with Bonferroni-corrected pairwise post-hoc tests.In Part 5b, we specifically aimed at replicating the results by de Kovel et al. 24 .We did so by modeling writing hand measured by the first EHI item (left vs. right, excluding mixed-handers) as a function of significant predictor variables in the study by de Kovel et al. 24 , specifically birth weight, birth size, breastfeeding, twin status, and the presence of any birth complication.We checked statistical assumptions of nominally significant (p < 0.05) linear regression models (Parts 2 and 6) by means of visual inspection of the following residual plots: correct specification of the model (residuals vs. fitted values), normality of residuals (normal Q-Q), homoscedasticity of residuals (scale location) and existence of outliers or influential data points (residuals vs. leverage).
For nominally significant (p < 0.05) ANOVAs (Parts 3, 4, and 7), we checked required assumptions by means of visual inspection of the following residual plots: correct specification of the model (residuals vs. fitted values) and normality of residuals (normal Q-Q).Moreover, we performed Levene's tests in order to test for homogeneity of variance.
Models in which at least one of the required assumptions seemed to be violated were excluded so that in the following sections, we only discuss regression models producing significant results while fulfilling all required assumptions.Table 3 shows the descriptive statistics of the quantitative laterality indices, quantitative birth factors, and clinical questionnaires for the final sample.Descriptive statistics for the categorical asymmetry measures and birth factors can be found in Table 4.
Part 2: asymmetry ~ quantitative birth factors.We ran linear regression models for seven outcomes (quantitative asymmetry measures) and five predictors (quantitative birth factors), applying FDR correction for www.nature.com/scientificreports/20 independent tests (Table 2).Figure 1 and Table S1 show the regression results.None of the models showed a significant association (all p > 0.05).
Part 4: asymmetry ~ quantitative birth factors.We ran ANOVAs for two outcomes (categorical asymmetry measures) and seven predictors (quantitative birth factors), applying FDR correction for 4 independent tests (Table 2).Figure 3 and Table S3 show the ANOVA results.www.nature.com/scientificreports/Left-handed participants reported higher maternal age at birth compared to mixed-(p = 0.017) and righthanded participants (p = 0.049).Right-and mixed-handed participants did not differ from each other regarding maternal age at birth (Table S3, model 5).This model did not remain significant after FDR correction (p = 0.056).
To sum up, effects for ANOVAs on clinical questionnaires as a function of categorial asymmetry measures seem to be driven by the (mixed-handed/mixed-footed) category mainly.www.nature.com/scientificreports/

Discussion
In this study, we considered associations between functional hemispheric asymmetries and early life factors as well as subclinical symptoms of psychopathology in healthy participants.This triad is interesting from a conceptual point of view: it relates to the idea that the ontogenesis of functional hemispheric asymmetries may be disturbed by certain birth factors resulting in atypical lateralization that sets a greater vulnerability or risk for psychopathological outcomes, or actually mediates their development 43 .Previous studies such as de Kovel et al. 24 relied on handedness as a single, categorical index of functional asymmetries which was inquired with only one item asking for hand preference.In this regard, the current study is unique since different forms of functional hemispheric asymmetries (i.e., handedness, footedness, language lateralization, visuo-spatial perception) were approached by means of deep phenotyping with different measures (i.e., self-report questionnaires, tests measuring dexterity, DLT and Line bisection task).Birth factors and clinical questionnaires were assessed by means of self-report.Thereby, our approach allowed for more nuanced statistical tests and also acknowledged the fact that different laterality phenotypes may differ in their strength of association with our predictors (e.g., early life factors and mental health outcomes) as well as in their sensitivity for showing these associations in our statistical analyses.
Few associations reached statistical significance and most did not survive correction for multiple testing.In Parts 2 to 5, we modelled birth factors as a function of functional asymmetries but did not find any significant effects in Part 2. For Part 3, we found the occurrence of birth complications and higher birth order position to model the LQ as calculated for the Pegboard task in that birth complications and higher birth order position were associated with lower Pegboard LQs (i.e., more leftward lateralization).For Part 4, handedness categories (left-handed/mixed-handed/right-handed) were identified as a function of maternal age at birth in that lefthanded participants reported higher maternal age at birth as the other two handedness categories (which did not differ amongst each other).For Part 5a, prevalence of handedness categories (left-handed/mixed-handed/ right-handed) significantly differed between participants of different birth order positions.In Parts 6 and 7, we modelled functional asymmetries as a function of clinical questionnaires.For Part 6, scores on clinical questionnaires were identified as a function of LQs as calculated for different tasks.The EHI LQ as well as the Alphabet LQ were shown to model the SPQ score in that higher SPQ scores were associated with lower LQs in these tasks (i.e., more leftward lateralization).Similarly, higher BDI scores predicted lower Alphabet LQs while higher CTQ www.nature.com/scientificreports/scores predicted lower DLT LQs, and higher ASRS scores predicted lower WFQ LQs.Therefore, in Part 6, higher scores on clinical questionnaires were uniformly predicted by more leftward lateralization.For Part 7, we found few significant differences for categorical asymmetry measures with respect to clinical questionnaires, which remained significant after correction for multiple testing.In detail, mixed-footed (WFQ) as well as mixed-handed (EHI) participants showed higher SPQ scores than right-footed (WFQ) and right-handed (EHI) participants.
Hence, these effects seemed to be driven by the middle (mixed-footed/mixed-handed) category in large parts.The WFQ was further associated with the STAI-T and the BDI scores but post-hoc tests did not reach significance.Noteworthy, most effects only reached statistical significance when not controlling for multiple comparisons.We consider it important to highlight the small effect sizes of basically all observed effects.That is, eta squared was η 2 = 0.01 for most significant associations and η 2 = 0.02 at the maximum for differences in the ASRS score and the SPQ score between participants of different handedness categories.Consequently, the associations found to be significant do not explain a considerable part of variance in handedness or other forms of functional lateralization.Along these lines, the great majority of associations tested in the current study did not reach statistical significance at all.Therefore, we did not interpret the significant effects functionally.
The prevailing pattern of a large number of non-effects alongside small effect sizes of nominally significant results in the current study is in accordance with existing literature.Most prominently, de Kovel et al. 24 also report only few statistically significant associations with negligible effect sizes between adult left-handedness and a plentitude of birth factors.Amongst the few factors that turned out to be significant in the publication by de Kovel et al. 24 , we also assessed birth weight, birth size, breastfeeding, twin status, and birth complications.Therefore, in Part 5b we attempted to directly replicate the findings of de Kovel et al. 24 in modelling handedness as classified by the writing hand item of the EHI as a function of these predictors.Noteworthy, none of the predictors reached significance.Failure to replicate the results by de Kovel et al. 24 may be attributed to the fact that our sample size was substantially smaller.Importantly, de Kovel et al. 24 used large-scale data from the UK Biobank covering ~ 500,000 participants.However, having said that most of the associations tested in our data did not reach statistical significance nor convincing effect sizes, one may conclude that they lack decisive importance at the population level.However, in the larger study by de Kovel et al. 24 effects reached statistical significance and probably did not so by chance.It is rather conceivable that effects are significant and thus important for the single individual.For instance, being part of a multiple birth may be of great importance in triggering the development of atypical brain asymmetry in some individuals, but not in others.
However, de Kovel et al. 24 conclude that the current literature does not support the notion that specific environmental variables (in their as well as in our study) may fill the gap between variance explained by genetic factors and so-far unexplained variance in handedness (and other functional asymmetries).This is the case for many clinical phenotypes in the epidemiological literature, for which twin studies show a substantial amount of variance explained by non-shared environmental factors.However, the role of non-shared environmental factors is likely to be heavily overestimated (and overinterpreted), as it is based on simple subtraction; it equals the variance not explained by additive genetic and shared environmental factors.Therefore, what is typically called non-shared environmental variance not only includes measurement error and gene-environment interaction, but also chance or random events 27 .De Kovel et al. 24 also accounted for this perspective in discussing their findings in the context of randomness in fetal brain development as already elucidated in the introduction.Notably, for the birth factors it has been proposed that their effect is mediated via epigenetic mechanisms.However, a large EWAS found only little handedness variance to be captured by epigenetic modifications of DNA 17 , casting doubt not only on strong associations between birth factors and functional asymmetries, but also on strong associations between epigenetic factors and functional asymmetries.
While de Kovel et al. 24 only tested for associations with self-reported writing hand, we included diverse functional laterality phenotypes and assessed them by means of deep phenotyping.Since we replicated the gross pattern of non-effects found by de Kovel et al. 24 for self-reported hand preference, deep phenotyping of multiple asymmetry measures does not seem to enhance the power of unraveling relations for the research question at hand.However, it should be noted that birth factors as well as clinical questionnaires in the current study were also assessed by means of self-report.Self-reports are typically prone to certain biases as well as reporting errors.Especially for the birth factors collected as self-report (i.e., birth weight), we had to exclude several data points based on plausibility (for details, see the "Method" section of the supplementary material).As a consequence, we cannot rule out that a more precise/objective measurement of the included birth factors as well as of clinical symptoms (e.g., by means of clinical interviews) would have led to a different pattern of results.
Moreover, for the current study, it is worth mentioning, that statistical power might have been impeded by the fact that our data consistently violated required assumptions of the appropriate statistical models.To counteract, we applied diverse transformations which did not always lead to perfect distributions.In this regard it seems debatable in how far data on lateralization phenotypes may represent a special case (bearing in mind their often J-shaped distribution).
Similarly, one might question whether the clinical questionnaires used in the current study were actually suitable in our healthy sample.Since participants were only included reporting no mental, psychiatric, or neurological disorder, at best we might have covered preclinical manifestations of the psychopathological constructs.As a result, the variance in our clinical questionnaires might not have been sufficient enough to unravel putatively existing effects.This assumption gains further plausibility considering the fact that clinical samples frequently produce large effects with respect to laterality measures.For instance, as already mentioned in the introduction, meta-analyses univocally confirm a certain relation between atypical handedness or other forms of functional lateralization and diverse clinical diagnoses (e.g., schizophrenia 31,32 , ASD 35 , PTSD 36 ).Therefore, one might conclude there is some sort of rubicon covering noticeable qualitative differences between healthy and clinical samples regarding asymmetry measures.Regarding the different kinds of associations that may possibly link atypical functional lateralization and psychopathological outcomes 9 , non-occurrence of effects in the www.nature.com/scientificreports/context of our healthy sample may rather point towards a broader, generic and transdiagnostic effect of atypical lateralization on psychopathology, if any.Still, for the results that reached statistical significance for a categorical operationalization of asymmetry measures in the current study, it is striking that they often concerned the "middle" category.Indeed, effects often seemed to be driven by mixed-handed/mixed-footed participants while more extreme forms of lateralization towards the left or the right side of the continuum did not seem to be influential.Hence, one might speculate that it is not left-but mixed-handedness/-footedness that shows the closest association with clinical constructs.Indeed, this is in line with several meta-analyses suggesting that disorders such as PTSD 36 and schizophrenia 32 are related to mixed-handedness in particular, rather than left-handedness.Therefore, it has been put forward that a reduction or an absence of asymmetries (such as mixed-handedness) rather than a reversal (such as lefthandedness) is of relevance for clinical outcomes 36 .
In conclusion, the current study further confirms previous findings of mostly negligible associations between birth factors and functional asymmetry measures in healthy individuals.Deep phenotyping did not lead to any substantial changes in this overall results pattern, confirming the robustness of previous findings using shallower phenotyping.Likewise, effects between functional lateralization and diverse psychopathological outcomes did not achieve noticeable predictive power in our sample of healthy individuals.Further research might identify qualitative differences between healthy and clinical samples as studying the latter typically renders strong effects for lateralization indices.Future studies might also benefit from the inclusion of social laterality phenotypes and biological markers as well as from the application of longitudinal approaches.

Figure 1 .
Figure 1.Quantitative asymmetry measures as a function of quantitative birth factors (linear regression).

Figure 2 .
Figure 2. Quantitative asymmetry measures as a function of binary birth factors (ANOVA).

Figure 3 .
Figure 3. Categorical asymmetry measures (L-M-R) as a function of quantitative birth factors.

Figure 4 .
Figure 4. Clinical questionnaires (overall scores) as a function of quantitative asymmetry measures (linear regression).

Table 2 .
Overview over the statistical models we applied to account for different effective directions between the measured variables (predictors vs. outcomes) alongside their scale level (binary/categorial vs. quantitative).M eff effective number of tests used for FDR correction.Equals the product of the effective number of tests determined separately for predictors and outcomes (see Table1).Vol.:(0123456789) Scientific Reports | (2023) 13:15348 | https://doi.org/10.1038/s41598-023-42563-7

Table 3 .
Descriptives of quantitative variables: laterality indices, quantitative birth factors, and clinical questionnaires.

Table 4 .
Descriptives of categorical variables: laterality indices, quantitative birth factors, and clinical questionnaires.