Introduction

Childhood aggression is the most common reason for psychiatric referrals in children, comprising 64% of all referrals [1]. Youth are responsible for up to 200,000 homicides every year [2], and over 1000 children need emergency care for youth aggressive and physical assault-related injuries daily in the United States [3]. Aggression can be defined as behaviors that intend to create physical or emotional harm on another individual [4]. From an evolutionary standpoint, aggression has been seen as an advantageous adaptive strategy in obtaining and defending food and mates. Therefore, certain levels of aggression have continued to be positively selected for and maintained through the generations. Among preschool-aged children, temper tantrums can be considered normal [5], yet certain aggressive behaviors can develop into more severe pathological forms. Increased anger, irritation and frustration accompanied by persistent aggressive behaviors can have negative consequences throughout life such as peer rejection, relationship problems, poor academic performances and lower graduation rates, substance abuse issues, criminal behaviors, and financial and occupational difficulties [2, 3, 6,7,8,9]. Pathological aggressive behaviors not only lead to social and financial problems for the aggressor and the victims’ families, but can also have significant societal costs through increased needs for health and medical services, unemployment and welfare services, social services, and criminal justice services [3, 10, 11]. With a high prevalence of problematic aggressive behaviors, up to 30% of children in low income and single parent homes exhibit aggression [12,13,14]. Childhood aggression is a major public health concern that requires further understanding for better prevention and treatment strategies.

Although aggression can take on different forms and behaviors, researchers categorize aggression into two main categories: proactive and reactive aggression [15, 16]. Proactive aggression represents aggressive behaviors that are predatory and have premeditated purposes to harm others for external or internal personal gains [15]. On the other hand, reactive aggression is the reaction to a perceived threat [16, 17]. Proactive and reactive aggression are highly correlated and can co-occur or be expressed separately [15].

While maladaptive aggressive behaviors can exist without fitting into a specific diagnostic category, they can also be the core symptom of some of the psychiatric diagnoses such as oppositional defiant disorder (ODD), conduct disorder (CD), intermittent explosive disorder (IED) and antisocial personality disorder (ASPD) (For a review of disorders related to aggression, please see Blair et al. (2014) [18]). Symptoms of excessive aggressive behaviors start early and are highly stable [19], however expression of disruptive behaviors can change through development and can be different in children versus adults [18]. While persistent and extensive aggressive behaviors can be a symptom of ODD around age 12 [20], it can develop into CD during adolescence [21] and then further into ASPD in adulthood [18]. Therefore, it is crucial to assess aggressive behaviors from early childhood and adjust measurement methods and criteria based on the age and developmental stage of the patients. However, it is also important to note that aggressive behavior is not necessarily an essential symptom for these diagnoses.

Studies over two decades have demonstrated that there is a prominent genetic component to aggressive behaviors. It has been found that aggressive behaviors are highly heritable and genetic factors account for roughly 50–65% of the risk of high aggression [22, 23]. Initially, chromosomal abnormalities were studied in relation to aggressive behaviors. While XYY individuals have shown to have increased aggressive behaviors [24], they do not form the whole picture [25]. Therefore, there have been numerous studies analyzing the association between genes and aggression. One of the first and landmark studies that found the genetic contributions to aggressive behaviors is the study by Brunner et al. [26]. Brunner and colleagues investigated a family with a history of criminal behaviors and found that all males were lacking monoamine oxidase A enzyme activity, which encodes the monoamine oxidase A (MAOA) enzyme that regulates catecholamine and serotonin levels [27]. Following that, Caspi et al. (2002) [28] further investigated the association between MAOA genotypes and aggressive behaviors in abused males, further supporting that variants within genes may influence aggressive behaviors. There has been significant research conducted on the genetics of aggression in adults [29, 30]. Nonetheless, a study on two large population cohorts with ages from 12–73 years reported that effects of polygenic risk scores for childhood aggression appeared to decrease from childhood and adulthood to later life [31], suggesting that while child and adult aggression are genetically similar, it is conceivable that some genetic factors underlying ADHD in children and later life may be different [32], thus emphasizing the importance of studying the genetics of aggression in different age groups separately. There have been various genes in different biological pathways investigated in association to childhood aggression, including dopaminergic, serotonergic, vasopressin and oxytocin system genes (For an earlier review of the genetics of aggressive behaviors, please see Anholt & Mackat (2012) [33]). Researchers agree that childhood aggression is a polygenic trait, with numerous genes of small effects contributing to the phenotype.

Although numerous studies demonstrate evidence for genes underlying childhood-onset aggression and there are previous reviews focusing on certain genes, there has not been a review that systematically considers every gene that has been studied in relation to childhood aggression. The objective of this study was to systematically review the literature to provide a comprehensive summary and informed analyses of the genes influencing aggressive behaviors specifically in child and adolescent populations.

Methods

Literature search

The literature search was performed on May 20, 2022 using the PUBMED, MEDLINE, and PsycINFO databases. Pubmed search yielded 256 hits using the following search terms: ((aggression [MESH] OR aggressive behav* [TIAB] OR aggressive trait* [TIAB])AND (genes [MESH] OR genetics [TIAB] OR epigenetics [TIAB] OR genom* [TIAB] OR genot* [TIAB] OR GWAS [TIAB]) NOT (neoplasms [MESH] OR tumor* [TIAB] OR cancer* [TIAB])) and applying the filter for Child (birth to 18) and human studies. Ovid MEDLINE and PsycINFO searches yielded 111 and 280 articles respectively using the following search terms: ((aggression.mh. or aggressive behav*.ab. or aggressive behav*.ti. or aggressive trait*.ab. or aggressive trait*.ti.) AND (genes.mh. or genetics.ab. or genetics.ti. or epigenetics.ab. or epigenetics.ti. or genom*.ab. or genom*.ti. or genot*.ab. or genot*.ti. or GWAS.ab. or GWAS.ti.)); FILTER for “Child (0 to 18 years)” and (“human”). Five relevant articles were subsequently added as a result of manual search and articles available to the authors. Of the 652 hits, 215 articles were found to be duplicates and were removed, leaving 437 studies qualifying for initial screening. As a result of initial title and abstract screening, 137 articles were found to be irrelevant, leaving 300 articles eligible for full-text review.

Inclusion and exclusion criteria

Since the purpose of our systematic review was to provide an overview of studies examining genes associated with childhood aggression, only articles examining aggression in children and adolescents aged 18 years or younger were included in this study. There were a few articles, however, that were included in our review despite the participants being older than 18 years of age at the time of assessment, because the participants were asked to rate their aggression retrospectively for when they were younger than age 18.

Studies were excluded from further review if they were: (1) written in a language other than English, (2) dissertations & conference abstracts, (3) full text was not available, (4) review articles, (5) wrong patient population (ex: only studying children who were victims of aggression) or study design (ex:case studies), (6) included adult participants or (7) tested phenotypes other than aggression. As a result of these criteria, 212 studies were excluded. The most common reasons for exclusion were the inclusion of the adult population (n = 128), wrong patient population or study design (n = 36), and the outcome not being related to child aggression (n = 22). Eighty-seven articles were subject to data extraction and quality assessment. The PRISMA flow chart is shown in Fig. 1.

Fig. 1: PRISMA flow chart.
figure 1

Flow chart showing the number of studies from our literature search and the number of studies removed during title/abstract screen and full-text review together a text box showing the reasons for exclusion.

Full size image

Data extraction and quality assessment

Data extraction was performed by three independent reviewers (CZ, TK, and EK). The following information was extracted for each study: First author, year of publication, population characteristics, study type (twin/pedigree studies, longitudinal, candidate gene etc.), participants’ ancestry or country of origin, sex, age, sample size, genes assessed, assessment of aggression, and key findings related to aggression (Table 1). The quality of each article was evaluated regarding the risk of bias on a 4 point scale (ranging from low risk to critical) using the following criteria: sample size, confounding, participant selection, measurement of outcomes, selection of reported results and overall risk of bias (Table S1). Abstract screening, full text review, quality assessment, and data extraction were managed in Covidence.

Table 1 Summary of reviewed articles.
Full size table

Results

Eighty-seven studies were included in our data extraction and quality assessment. The general summary of these articles can be found in Table 1. Seventy percent (n = 61) of the articles examined samples from European ancestry. Eighty percent (n = 70) of studies included both females and males, while 19% (n = 16) studies only included males. Only one study included only females in their study.

The first research study retrieved using our search strategy was by Twitchell and colleagues in 2001 [34], which examined the genetics of child aggression among offspring of alcoholic fathers. Until 2006, childhood aggression was examined in association with particular psychiatric disorders, such as ADHD, CD and ODD, and studies examining aggression as a transdiagnostic behavioral phenotype were rare. The first studies used the candidate gene approach, which continues to be the major research method (74% of all studies) used in studying the genetics of childhood aggression. Until 2007, studies focused on several key genes, including serotonin transporter, MAOA, and dopamine D4 receptor. From 2011, genome-wide association studies started to be published, although the sample size remained relatively small (n < 1000) until 2016 [35].

Twin / pedigree studies

Throughout the years, researchers focused on twin studies for heritability and understanding the contribution of genes to aggressive behaviors. Studies have demonstrated a significant heritability for aggressive behaviors of up to 60% [36,37,38]. However the influence of genes can be augmented by the environment such that it decreases with decreasing positive social feedback [37] or increasing parental negativity [38]. From the genetic influences, mainly additive genetic factors have been found to explain the variability; they accounted for 15-77% of the variance in social aggression [39] and up to 90% of the individual differences in baseline impulsive aggression in the longitudinal Quebec Newborn Twin Study [40]. Interestingly, the effects of genetic factors on aggressive behaviors can change over time [41] and the influence of genes on the variation of aggressive behaviors can change over development (18% genetic influence in early-middle adolescent to 47% genetic influence in middle adolescent) [42]. Although the magnitude of the influence may change, genetic factors still account for the stability of physical aggression as well as can explain the individual differences in the initial levels and the rate of change of aggressive behaviors over time in a longitudinal Canadian sample [41].

Candidate gene studies

The majority (77%; n = 67) of genetic studies in childhood aggression used a candidate gene approach. The most commonly investigated genes include monoamine oxidase A (MAOA; n = 17), dopamine D4 receptor (DRD4, n = 13), and catechol-o-methyltransferase (COMT, n = 12). Many of these candidate gene studies, however, suffer from methodological issues including small sample size (often <500), lack of psychometrically sound assessments of aggression, and inconsistencies in the cutoffs and categorization of repeat polymorphisms, thus making generalization of findings difficult.

Monoamine oxidase A—MAOA

The MAOA gene is located on the X-chromosome (Xp11.23) and encodes for MAOA, an enzyme which catabolizes monoamine neurotransmitters such as serotonin, epinephrine, norepinephrine and dopamine. In our systematic review, MAOA was assessed in over a quarter of the studies assessing candidate genes (n = 17; one study was categorized as epigenetic studies). The most commonly examined polymorphism is MAOA-uVNTR, the 30-bp repeat polymorphism, which can exist in 2 repeats (2 R), 3 R, 3.5 R, 4 R, and 5 R. Many studies regard the 2 R, 3 R, and 5 R to be low activity variants (MAOA-L), and the 3.5 R and 4 R to be high activity variants (MAOA-H) [27]. Nevertheless, there is great variation across studies in the categorization of low and high activity repeats (see Table 1), making direct comparisons of findings across studies challenging.

The first study that investigated the association between MAOA-uVNTR and childhood aggressive behavior in the context of ADHD reported that low transcription (3 R only) alleles of MAOA-uVNTR was associated with higher aggression among children with ADHD [43]. Similarly, different studies have also found that the low-expression (2 R, 3 R, and 5 R) MAOA-uVNTR to be associated with childhood aggression, although their categorization of low-expression alleles differ [44]. On the other hand although fewer in number, some studies have reported that the high-expression variant was associated with increased aggressive behavior in children [45].

While the reason for these inconsistent findings across studies has not been fully examined, several researchers have suggested that developmental age may have an influence on the observed effects of MAOA risk alleles. Pingault et al. [46] examined the age-dependent contribution of six MAOA SNPs on childhood physical aggression using a longitudinal dataset of 436 boys followed annually from ages 6 to 12 in Quebec, Canada.The results showed that the T-allele carriers for rs5906957 had lower initial levels of physical aggression and also a less steeper decline in physical aggression over time compared to the C allele carriers. In a similar vein, Kant et al. [47] examined the effect of MAOA-uVNTR on aggression and psychopathic traits by developmental age. The 194 male participants were divided into those below age 13 (n = 132) and those at or above age 13 (n = 62). While MAOA-uVNTR was not significantly related to aggression in either age group, there was an interesting pattern in that, in the younger age group, oppositional defiant problems and conduct problems were associated with the high-activity MAOA 4 R allele (MAOA-H), whereas in the older age group, oppositional defiant problems and callous-unemotional traits were more significantly associated with the low-activity MAOA 3 R allele (MAOA-L). More studies are required to confirm the age-dependent association of MAOA on aggressive behavior.

It should be noted, however, that the majority of studies reported no significant main effects for MAOA gene variants on childhood aggression [48]. Despite this, following the example of Caspi et al.’s [28] seminal study which reported no significant main effect of MAOA-uVNTR but a significant MAOA gene by childhood maltreatment interaction, several studies examining gene-by-environment interaction have been conducted. Various types of environmental exposure have been examined, including child maltreatment, abuse [49,50,51], and parenting behaviors [52,53,54,55]. These gene-environment interaction studies have sometimes been used to test theories regarding the role of genes and environment in childhood aggression. For example, Zhang et al. [54] sought to test two related hypotheses regarding the role of gene and environment: Diathesis-stress (i.e., carriers of certain genetic risk variants will show greater aggression when exposed to adverse environments) and differential susceptibility (i.e., not only do carriers of certain genetic variants show greater aggression when exposed to adverse environments, the carriers of the same genetic variants will show less aggression when exposed to supportive environments). This study with 1399 healthy Han Chinese adolescents supported the differential susceptibility hypothesis; males who had the T allele and females who had the homozygous for the T/T genotypes for MAOA rs6323 (T941G) were more likely to exhibit reactive aggression when the mothers exhibited low levels of positive parenting but were less likely to exhibit reactive aggression when mothers exhibited high levels of positive parenting.

Catechol-O-Methyltransferase—COMT

Catechol-o-methyltransferase (COMT) is an enzyme that metabolizes catecholamine neurotransmitters including dopamine, epinephrine, and norepinephrine. It has two isoforms, a longer membrane-bound (MB-COMT) isoform that is expressed mainly in neurons in the brain [56], and a shorter soluble (S-COMT) isoform that is expressed in other tissues such as blood, liver, and kidney. It is coded by the COMT gene, which is localized on chromosome 22q11.21. A common single-nucleotide polymorphism, rs4680 (Val158Met), within the coding region of COMT changes the amino acid Valine at position 158 of MB-COMT to Methionine, which decreases the thermostability and activity of COMT enzyme. The role of COMT in aggression was initially supported by observations of hostility in mice deficient in Comt and the negative correlation between COMT levels and hostility in men with behavioral problems as children [57, 58]; reviewed in Qayyum et al. (2015) [59].

The COMT genetic variants, particularly rs4680, have received a similar level of attention as MAOA in association studies of child aggression. The first published study that examined a possible association between COMT gene and child aggression was in the context of ADHD, where Caspi et al. [60] reported an association among ADHD patients of Val/Val with increased aggression compared to Met-carriers; the association was replicated across three samples within this study. Other studies reported high aggression being associated with either Met-allele carriers [61], or no significant association [62]. Few studies examined SNPs other than rs4680, with one reporting rs6269 A/G heterozygotes being over-represented in cases compared to adult controls [63] and another reported non-significant results for rs6267 (Ala22/72Ser in S/MB-COMT) [64].

The possible association of COMT with child aggression has been examined in the context of interactions with environmental or demographic variables. For example, in a birth cohort study, among children who scored high on disorganized attachment, Val/Val carriers exhibited greater increase in aggression from 4 years to 6 years of age than Met-allele carriers [65]. The same research group also reported that, among those with stressful life events, rs4680 Val/Val homozygotes were more aggressive than Met-allele carriers, while the reverse was observed among those without stressful life events [66]. In a study on Chinese early adolescents, Val/Val carriers were reported to display higher reactive aggression compared to Met-allele carriers in the context of higher positive parenting scores, but lower reactive aggression compared to Met-allele carriers with lower positive parenting scores [54]. Age and sex may also be an effect modifier for the effect of rs4680 on risk of child aggression. For example, Kant and colleagues [67] demonstrated that among European males at least 13 years of age, Val-allele carriers had higher CBCL aggressive scores than non-carriers (p = 0.03). In contrast, among those younger than 13, Met/Met genotype carriers had increased conduct problems compared to Val-allele carriers (p = 0.03). These associations were not observed in females in their sample. A three-way interaction was reported, where carriers of the COMT low-activity rs6267 T allele and MAOA rs6323 T allele displayed higher aggressive behavior in the presence of high academic pressure than those with low academic pressure; this association was not observed in carriers of other genotype combinations [64]. Further efforts in large samples are needed to confirm these preliminary interaction findings and pursue more complex interaction analyses.

Dopamine system genes

The dopamine system is vital to the regulation of motor and cognitive behaviors, and dopamine dysregulation has been implicated in multiple psychiatric and behavioral disorders.

Within the dopamine system, aside from COMT mentioned above, the most studied dopamine system gene in child aggression is the dopamine D4 receptor-encoding DRD4, which is localized on 11p15.5. The 48-bp exon III variable number tandem repeat (VNTR) polymorphism [68, 69] is the extensively studied DRD4 polymorphism, for which between two to eleven repeats (R) have been observed in humans, with the 4-repeat (4 R), 2 R, and 7 R being the most commonly observed alleles. Functional significance of this polymorphism has been demonstrated [70,71,72,73,74,75]. The 7 R has been shown to reduce in-vitro DRD4 expression [73] and to be less likely to form heterodimers with the dopamine D2 receptor [76], while the 4 R allele appears to be less responsive to quinpirole-mediated DRD4 upregulation [74].

The larger repeat (7 R or 6-8 R) alleles were associated with high aggression in an Italian sample [77], the Mannheim Study of Children at Risk study [78], and the Ben-Gurion University Infant Developmental Study [79], while the 3 R allele (p = 0.014) and rs3758653 C/C genotype were nominally associated with aggressive behavioral impulsivity in the International Multicenter ADHD Genetics (IMAGE) study [80]. The VNTR was not associated with externalizing behavior in a longitudinal community sample of 87 boys [81], within our earlier sample of 48 clinically referred aggressive boys [81], or our later sample of 144 high aggression child cases and adult controls [82].

A number of gene-environment interaction findings have been reported for DRD4. In a study on the Dutch Twin Registry sample, a significant VNTR-by-maternal sensitivity interaction was observed. More specifically, larger repeat allele (7 R or 6-8 R)-carrying genotypes were associated with higher externalizing behaviors or aggression compared to 7 R non-carrying genotypes (e.g., 2–4 R) only in the context of maternal insensitivity [83], high maternal prenatal stress [84], or low-aggression peer play environment [85]. Besides COMT and DRD4, only few other dopamine system genes have been examined in child aggression, with our group reporting that DRD2 rs1799978 (A-241G) G-allele carrying genotypes, rs1079598 C/C genotype, and rs1800497 (TaqIA) T/T genotype were overrepresented in high aggression cases compared to adult controls [82]. A significant DRD4-by-socioeconomic status interaction in high aggression scores has been reported, where the 6–8 R carriers with low socioeconomic status had higher aggression scores compared to other comparison groups [77].

Serotonin system genes

Under the serotonin system genes, the most extensively studied polymorphism is the 5-hydroxy-tryptamine-linked polymorphic region (5-HTTLPR) polymorphism of the SLC6A4 gene.

5-HTTLPR long allele (L/L genotype) was associated with higher CBCL aggressive behaviors score in 607 Italian children [77]. Interaction between low socioeconomic status and 5-HTTLPR long alleles further demonstrated significant effects on aggressive behaviors [77]. A smaller sample consisting of 62 European participants similarly reported an increased risk for behavioral disinhibition and aggressive behaviors with the L/L genotype when compared to S/S and S/L [34]. On the other hand, Beitchman and colleagues [86] reported a significant effect of 5-HTTLPR on aggression with the low expressing (S/S, Lg/S, Lg/Lg) genotypes in children with clinically severe aggression. Similarly, the S-allele was significantly associated with teacher reported aggressive behaviors at age 9 for both boys and girls [87] and with increased aggressive behaviors and hostility in a group of female Caucasian Russian swimmers [88].

There are also studies that did not yield significant 5-HTTLPR main effect findings on childhood aggression. Several studies on European children and adolescents [78, 89] and Chinese adolescents [51] did not report a significant main effect of 5-HTTLPR on aggressive behaviors and related phenotypes. Similarly, the initial analyses of the study on 87 adopted children from the United States of America further failed to detect a main effect of 5-HTTLPR and aggressive scores [90]. However interestingly, when the biological parent status and sex of the children were included in the analyses, the results were significant. Male children with S/S or S/L (short) demonstrated increased aggressive behaviors while females with the SS and SL demonstrated lower levels of aggression. Moreover, when the biological parent of the child was considered antisocial, adolescents, but not preadolescents, demonstrated a significant increase in aggressive behaviors with the L/L genotype [90].

Although some studies failed to report a significant main effect of 5-HTTLPR on aggressive behaviors, they demonstrated a significant gene-gene interaction on behavior. Zhang and colleagues [51] reported that there was a three-way interaction between MAOA high activity, 5-HTTLPR and sexual abuse on aggressive behaviors. Children with MAOA high activity, 5-HTTLPR S/S allele and with increased sexual abuse experience exhibited higher aggressive behaviors [51]. Furthermore, there was a significant interaction between 5-HTTLPR S/S genotype and DRD4 7 R on increased aggression scores [78], while Nobile and colleagues [77] demonstrated increased aggression with DRD4 VNTR 6-8 R and 5-HTTLPR L/L genotype.

Other polymorphisms of serotonin system genes that have been studied in relation to childhood aggression include SLC6A4 VNTR polymorphism and tryptophan hydroxylase 2 (TPH2) gene polymorphisms. Neither the SLC6A4 VNTR [86] nor the TPH2 rs4570625 polymorphism [91] demonstrated significant associations with childhood aggression. Furthermore, four SNPs of the 5-hydroxytryptamine receptor 2 A (HTR2A) gene that encodes for one of the receptors for serotonin failed to have a significant difference between the conduct disorder cases and controls in adolescents [92]. However, in adolescent cases, G/G genotype or the G allele carriers of rs2070040, C-allele carriers of rs9534511 and G-T haplotype of rs2070040- rs9534511 were associated with increased aggressive scores [92]. On the other hand, in adolescent controls T/C haplotype of rs4142900-rs9534512 was associated with the increased aggressive behaviors [92]. Other serotonin receptor encoding gene polymorphisms, including 5-hydroxytryptamine receptor 1B [93, 94], 1E [95] and 2 C [96] further demonstrated significant effects on childhood aggression (Table 1). Lastly, a recent comprehensive study analyzing the association between polygenic score indexing serotonin functioning and aggression demonstrated that adolescents with higher serotonin polygenic risk (lower levels of serotonin functioning) had an increased risk for aggressive and antisocial behaviors [97].

Hypothalamic-pituitary-adrenal (HPA) axis and hormonal signaling genes

HPA axis refers to the neuroendocrine system that involves the hypothalamus, pituitary, and adrenal glands and is responsible for stress response and regulation of various biological processes such as food digestion and immune response. The hypothalamus secretes corticotropin-releasing hormone (CRH) and arginine vasopressin (AVP) in response to physical or psychological stress. These two neurohormones are transported to the pituitary through blood vessels and bind to the CRH and AVP receptors respectively and stimulate the release of adrenocorticotropic hormone (ACTH). ACTH then stimulates the secretion of glucocorticoids such as cortisol. Glucocorticoids in turn provide a negative feedback signal to inhibit the secretion of CRH and AVP from the hypothalamus and ACTH from the pituitary gland, respectively. Animal studies have consistently shown a robust association between the HPA axis and aggressive behavior [98]. Nevertheless, relatively few studies have interrogated genes along the HPA axis with regards to childhood aggression.

Several researchers have investigated the arginine-vasopressin related genes in association with childhood aggression. Zai et al. [99] examined eleven SNPs from the AVP receptor 1 A (AVPR1A), AVPR1B, and AVP genes in 177 children with high aggression and ethnicity and sex matched adult controls and found a significant association between childhood aggression and AVPR1B rs35369693, as well as the two-marker haplotype containing rs35369693 and rs28676508. Similarly, Malik et al.’s [100] study compared 182 clinically aggressive children of European ancestry with 182 sex, age and ancestry-matched non-aggressive controls and found that the A allele and the AA genotype of the rs3761249 SNP of the AVP gene was underrepresented in highly aggressive male cases, whereas AVPR1A rs1174811 G allele was over-represented in highly aggressive female cases. While the former studies examined SNPs in the AVP pathway, Vollebregt et al. [101] examined two microsatellites, RS1 and RS3, of the AVPR1A gene among children with pervasive aggression and non-aggressive age-matched controls. They found that the RS3 long repeat variants were nominally associated with non-aggressive status. It is noteworthy that Pappa et al.’s [35] genome-wide association study also found an association between the AVPR1A gene and childhood aggressive behavior in a post-hoc gene-based analysis, warranting further investigations of AVPR1A gene variants.

With regards to the corticotropin releasing hormone (CRH), Liu et al. [102] reported that the carriers of the G allele and the GG genotype for rs24924 of the corticotropin-releasing hormone receptor CRHR1 gene were overrepresented among young offenders of violent crime compared to non-violent control adults in a Han Chinese sample.

The FKBP5 is a co-chaperone of the glucocorticoid receptor. Studying the association of the FKBP5 gene with childhood aggression, Bryushkova et al. [103] did not find significant main effects with any of the SNPs within the gene, but found a significant gene-environment interaction in that A allele carriers of the FKBP5 rs4713916 who were exposed to maltreatment exhibited the highest levels of aggression.

Oxytocin (OXT) is a nonapeptide most widely known for its stress-reducing effects and has been shown to affect prosocial behaviors, emotional recognition and feelings of trust [104,105,106]. The gene coding for OXT receptor (OXTR) has been examined in association with childhood aggression. The study by Malik et al. [100] found that OXTR rs237898 A allele was over-represented in high aggression children. Other studies have found a gene-by-environment interaction between variants in the OXTR gene and stressful life events [107]. Glenn et al., [108] found that the variations in the OXTR gene moderated the effectiveness of, Coping Power, a disruptive behavior modification program.

Genome-wide association studies

We have identified 12 genome-wide association studies (GWASs) of child aggression, the majority of which were performed on children and adolescents with European ancestry. The first GWAS focused on the Dysregulation Profile from CBCL (which consists of Attention Problems, aggressive behavior, and anxious/depressed clinical subscales) among 341 ADHD children from 339 ADHD affected trio families [109]. This study found no genome-wide statistically significant associations (P < 5 × 10–8); however, TMEM132D, LRRC7, and STIP1 were identified as nominally significant [109]. The second GWAS on 398 ADHD child cases from Cardiff and 5,081 controls from the Wellcome Trust Case Control Consortium (Phase 2) found higher polygenic risk scores for ADHD (ADHD-PRS) scores in ADHD cases with diagnosis of conduct disorder compared to those without, and positive correlation between ADHD-PRS and the number of aggressive conduct disorder symptoms within cases [110]. The first GWAS by the EAGLE (Early Genetics and Lifecourse Epidemiology) Consortium performed quasi-Poisson regression on aggression scores across nine cohorts with a total of 18,988 participants from early childhood and mid-childhood/ early adolescence and reported a near genome-wide significant variant (rs11126630, P = 5.3e–8) at 2p12 and a significant gene (AVPR1A) [35]. They also reported that the 450,000 tested common variants accounted for between 10% and 54% of the variance in aggression across three sample sets ranging from 3 to 6 years of age [35]. The authors suggested that the large range of observed SNP heritability could be due to different sample characteristics, environmental contributions, and ages across these samples [35]. With additional samples across the age ranges, we may be able to capture the pattern of genetic components across development. [35] The authors followed up with GWAS of aggression subtypes as well as a cross-trait gene-based meta-analysis of GWAS of aggression with GWAS of volume of amygdala, nucleus accumbens, or caudate nucleus [111]. They found the MECON (MDS1 And EVI1 Complex Locus) gene to be associated with cross-trait construct of aggression and nucleus accumbens volume, and the AVPR1A gene to be associated with the construct of aggression and amygdala volume. Another GWAS of aggressiveness during childhood on 1050 adult ADHD patients and 750 child ADHD patients reported the top suggestive variant in a long non-coding RNA gene on chromosome 10 (rs10826548) and the top suggestive gene to be WD repeat domain 62 (WDR62) [112].

A Polygenic risk score (PRS) is an estimate of the genetic risk for a phenotype of interest and is generally calculated based on the number of risk alleles each person possesses and the effect sizes of these risk alleles [113]. Genetic correlation is an estimate of the genetic similarity between two complex phenotypes by calculating the correlation of phenotypic effects across genetic variants [114]. In a longitudinal study of children of diverse low-income families from the Women, Infants, and Children Nutritional Supplement Programs (WIC) study, PRS for child aggression [35] based on all SNPs with p < 0.05 or SNPs mapped to gene regions were not significantly associated with aggression at any age from early to mid-childhood, while PRSs enriched for SNPs with putative biological function were associated with aggression, with effect estimate appeared to change through early childhood (age 2–5 years) to mid-childhood (age 7.5–10.5 years) [115]. In a more recent study on the WIC sample, higher aggression-PRS based on the EAGLE Consortium GWAS [35] appeared to predict greater co-occurring internalizing/externalizing problems at age 14 via negative affectivity observed during parent-child play at age 3 [116]. In a sample of 404 participants from a school-based program consisting of two preventive interventions for early learning and aggressive/ disruptive behaviors, polygenic risk scores for conduct disorder from the SAGE (Study of Addiction: Genes and Environment) sample, an interaction between polygenic risk scores and exposure to community violence was observed such that among those who endorsed witnessing violence, conduct disorder PRS was negatively associated with likelihood of being in the high-aggression group (or positively associated with likelihood of being in the lowest aggression group [117]). In the most recently published GWAS of aggression with multiple observations in 87,485 children from ages 1.5–18 across multiple sites, instruments, and study designs, SNP heritability was reported to be 3.31% [118]. Though no genome-wide significant SNPs were found, three genes emerged as showing association with childhood aggression from gene-based analysis: ST3GAL3 (p = 1.6e–6), PCDH7 (p = 2.0e–6), and IPO13 (p = 2.5e–6). The authors also reported significant genetic correlation between aggression and 36 phenotypes, including positive correlations between aggression and ADHD, smoking, major depressive disorder, and autism spectrum disorder, as well as negative correlations between aggression and age at smoking initiation, intelligence, and educational attainment [118].

Mendelian randomization studies

One potentially powerful way in which genes have been used in the research literature is to clarify the causal mechanism between a predictor variable and outcome. This approach, known as Mendelian Randomization, uses genes as an instrumental variable, that is, a variable that predicts the predictor variable but not other confounding variables. Because genetic variants are inherited at random from the parents to their child, it can act as a quasi-randomized experiment.

Only one study was identified that used Mendelian Randomization to examine childhood aggression. Chao et al. [119] sought to examine the causal effect of alcohol consumption during adolescence and externalizing behaviors (including aggression; evaluated by Youth Self Report [120]) in 1608 Chinese adolescents. The Glu504Lys (rs671) polymorphism within the aldehyde dehydrogenase 2 family member-encoding ALDH2 gene, having established effects on enzyme function [121,122,123] and consistent associations with alcohol use-related phenotypes [124, 125], was used as the instrumental variable. The results showed that decreased ALDH2 function was significantly associated with lower alcohol use, and also with lower aggression problems. Alcohol use was found to be a significant mediator of the relationship between ALDH2 and aggression, thus supporting the hypothesis that alcohol use causes adolescent aggression.

Epigenetic studies

Our search resulted in two epigenetic studies. Provençal and colleagues [126] conducted a case-control study for 8 high-aggression case and 12 control participants and studied T cell DNA methylation using methylated DNA immunoprecipitation (MeDIP) followed by hybridization to microarrays. Their results reported that 227 and 171 distinct gene promoters were methylated significantly more in the control and high aggression group, respectively. From the differentially methylation genes, AVPR1A, HTR1D and GRM5 were less methylated while DRD1 and SLC6A3 were more methylated in the high aggression group. More recently, Cecil and colleagues [127] demonstrated that there were seven differentially methylated sites across the genome in children who developed early onset conduct problems from an epigenome-wide association study (EWAS). Results of their follow-up studies with 15 candidate genes that were previously studied in relation to childhood aggression demonstrated that MAOA, BDNF and FKBP5 were further associated with early onset of conduct problems in children [127].

Discussion

To our knowledge, this is the first systematic review that specifically focuses on the genetics of childhood aggression. Overall, there is growing interest in this research area, as evidenced by the growing number of studies since 2001 (Fig. 2). Twin and pedigree studies support a prominent genetic component in liability for childhood aggression, which encourages further research to replicate and clarify findings from existing literature. The majority of gene association studies were candidate gene studies, which have focused on the MAOA, DRD4 and COMT genes with mixed findings of their main effects. For the majority of candidate genes we reviewed, the positive findings (if any) have not been replicated in childhood aggression GWASs thus far [128]. It should be noted that many of the earlier childhood aggression candidate gene studies and GWASs were limited by insufficient sample sizes, lending itself to potential spurious relationship reportings and overestimation of effect sizes, a phenomenon known as the winner’s curse [129, 130]. Nonetheless, we found converging evidence for a role of AVPR1A in child aggression coming from genome-wide association [35], epigenomic [126], and candidate gene [101] studies. This warrants further investigation into the mechanism through which AVPR1A affects risk of child aggression and demonstrates that the use of diverse genetic study methodologies can facilitate genetic discoveries.

Fig. 2
figure 2

Number of childhood aggression studies. A histogram showing the number of childhood aggression genetic studies published per year.

Full size image

The conclusions from this review should be interpreted with the following considerations. Firstly, only studies that were in the English language were included, which may have biased the results to studies examining primarily European participants. Second, because our main focus for this systematic review was to shed light on the genetics of childhood aggression using only mesh terms of variants of the word aggression, studies that did not have direct assessments of childhood aggression or used only psychiatric diagnoses (e..g, ADHD, conduct disorder, oppositional defiant disorder) as proxies for aggressive behaviors would have been excluded. Furthermore, with null findings possibly not being reported and statistically significant findings tending to be published, publication bias is likely when drawing generalized conclusions from these published results [131].

Quality assessment of studies included in this review identified a number of areas where improvements will help advance the field of child aggression genetics. Sample size is a major limitation, with 74% of the studies being rated as moderate to serious in our quality assessment (Supplementary Table S1). Although it is more apparent in earlier candidate gene studies, it also remains a limiting factor in identifying genetic markers for child aggression GWASs. As many genetic studies also examined gene-gene and/or gene-environment interactions, even larger sample sizes are required. Another consideration is the definition and measurement of child aggression, for which 75% of the studies have been rated as moderate to critical (Table S1). The assessment methods of aggression varied substantially across studies in terms of tools and informants (Table 1), which may have increased heterogeneity and limited comparability across studies. Another consideration is the variability in the inclusion of potential confounding factors as well as that of environmental factors being examined in interaction with genetic factors, for which 95% of the articles have been rated as moderate to serious (Table S1). There are numerous prenatal and postnatal environmental factors, such as socioeconomic status, childhood trauma, abuse and maltreatment, parenting styles, maternal sensitivity, prenatal stress, parental psychiatric disorders and alcohol use, that may influence the effects of genes on aggressive behaviors (reviewed in [132]). Studies using standardized measures of aggression and considering multiple environmental and confounding factors will help to disentangle the complexity surrounding child aggression.

Moreover, studies were limited with their participant selections where 92% of the studies has been rated as moderate to serious (Table S1). While the majority of the studies included only participants of European ancestry (Table S2) in order to limit spurious findings due to population stratification, the results may not be generalizable to participants of other ancestries [63, 133]. More studies on participants of non-European ancestries are important in gaining additional insights into biological pathways for child aggression [134], as demonstrated in multi-ancestry GWASs of other phenotypes such as asthma [135] and rheumatoid arthritis [136]. Moreover, 16 studies included in this review only included male participants, while only one study included only female participants. Sex is a major factor that may modify the gene-behavior association. While males are three times more likely to exhibit aggressive behaviors than females due to both biological and cultural factors [137,138,139], the effects of genes on behavior may also be modified by sex-specific factors such as the levels of testosterone and Y-chromosome genes [140, 141]. Therefore more studies focusing on females are necessary to understand the genetics underlying female youth aggressive behaviors.

Furthermore, developmental age is another major factor that may change the effects of genes on childhood aggression [142] due to factors including the changing levels of gene expression, hormones, and enzymatic activity during development [142,143,144]. Study designs and data analyses that account for age and/or development in their study designs and data analyses, as age-stratified analyses [47, 67] and longitudinal assessments for changes in aggressive behavior and related factors may uncover novel associations and clarify mixed findings in the literature.

Lastly, it is important to note that GWAS does not directly interrogate other types of genetic variants, such as repeat polymorphisms (e.g., MAOA-uVTNR, 5-HTTLPR, DRD4 exon III VNTR, AVPR1A RS1 and RS3). Examining the correlation between SNPs and these repeat polymorphisms will help in incorporating this type of polymorphisms in GWAS. Incorporation of rare variants, copy number variants, and other genetic variants besides SNPs and repeat polymorphisms in whole-genome analyses would likely help in explaining additional portions of the risk for child aggression and understanding the genetic architecture underlying child aggression [145,146,147,148,149]. Building consensus on the designation of risk vs. non-risk alleles, low vs. high activity genotypes, and short vs. long allele cutoffs in repeat polymorphisms will also facilitate the interpretation and generalizability of research findings across studies.

There are many future directions that can be followed from the results and limitations found from our systematic review of the literature. Most prominently, with candidate gene studies continuing to dominate the field of childhood aggression research, there is a greater need for more varied approaches, including epigenetic studies, gene expression studies, interrogation of rare [145] and/or more complex variants [148] in addition to SNPs, gene system studies, longitudinal studies that track changes in risk/ameliorating factors and aggression-related outcomes, as well as studies examining causal mechanisms related to aggressive behavior.

With the exception of ADHD and autism spectrum disorder, there is a paucity of well-powered GWASs in pediatric populations [150, 151], especially for aggressive behaviors and related phenotypes such as disruptive behavior disorders, conduct disorders, as well as externalizing and internalizing behaviors. There are a few studies that have investigated aggression-related phenotypes in the context of ADHD and other psychiatric disorders using summary data such as the Psychiatric Genetics Consortium, with one study noting an increased contribution of common genetic variants to ADHD with disruptive behavior disorder compared to ADHD without disruptive behavior disorder, with a portion of that increase attributed to genetic variants associated with aggression [152]. Genomic analyses of the genetic architectures of aggression and related phenotypes in youth as well as their co-occurrences will improve our understanding of the unique and shared genetic components across these phenotypes and across the lifespan [110, 152]. Therefore, future research is warranted focusing on the shared genetic architecture of aggression and the related phenotypes.

Conclusion

Extreme and persistent childhood aggression continues to be a public health concern worldwide with potentially serious lifelong consequences to the perpetrator, the victim, and their loved ones, as well as incurring major costs to the society as a whole [153]. To devise effective early identification, intervention, and prevention strategies, an understanding of the biological mechanisms and environmental determinants of excessive childhood aggression is paramount. However, it is crucial to consider the factors such as sex, environment, development, and ethnicity when analyzing the effects of genes on child aggression. Although we found that the quality of the reviewed studies improved over time, the overall risk of bias for 95% of current evidence were rated as moderate to serious (Table S1). Improvement to the research design including larger sample size and standardized, reliable assessment of aggressive behavior, as well as triangulation of research evidence using diverse genetic research methodologies, will facilitate the advancement of genetic research in childhood aggression.