Background
Episodic memory is the memory of autobiographical events that occur at a particular time and place, engaging the medial temporal lobe (MTL) and prefrontal cortex (PFC) [
1‐
3]. Episodic memory can be categorized into emotional and non-emotional stimuli [
4]. Aversive content and emotional arousal have consistently been shown to increase memory of previously encoded items compared to neutral ones. The hippocampus plays an important role in processing negative or stressful information. Therefore, inhibiting negative memory engrams in the hippocampus could be a novel therapeutic approach for treating the cognitive symptoms of depression [
5]. The striatum is also important for episodic memory formation, and successful memory is associated with greater activity in the striatum during encoding [
6].
Deficits in both emotional and non-emotional episodic memories have been observed in various neuropsychiatric disorders [
6]. In a neutral encoding task, patients with schizophrenia and their healthy siblings showed reduced parahippocampal activation and hippocampal-parietal coupling during the encoding of neutral stimuli compared with normal control participants [
7]. Patients with depression demonstrated impairments in selecting relevant positive information [
8]. Individuals with anxiety disorders showed enhanced memory for threatening disorder-related material [
9], but impaired memory for neutral information [
10]. In addition, police officers with post-traumatic stress disorder (PTSD) exhibited smaller hippocampal volumes, higher cortisol levels, and memory impairments [
11]. Other studies have suggested that the amygdala has heightened responsivity in symptomatic states of PTSD during the processing of trauma-unrelated affective information, whereas the responsivity of the medial prefrontal cortex (MPFC) is inversely associated with PTSD symptom severity [
12].
Urbanicity is a major socio-ecological change especially in this century. By 2050, the urban population is expected to increase to 66%, whereas the rural population is expected to decline [
13]. Previous studies have suggested that the environment during childhood affects brain development [
14]. Urban childhood was negatively correlated with the gray matter volume (GMV) of the MPFC in developed and developing countries [
15,
16], while positively correlated with the GMV of the dorsal lateral prefrontal cortex (DLPFC) only in developing countries [
16]. Meanwhile, activation of the pregenual anterior cingulate cortex (pACC) in a social stress task was affected by childhood urbanicity [
17] and interacts with polygenic risk score to affect brain activation under social-stress working memory task [
18]. Different urban and rural childhood environments can affect early memory development. Among children aged 10 to 13 years, those with early rural childhoods were more likely to remember information about social interactions, while those with early urban childhoods were more likely to report individual memories, and these memories appeared to contain more words [
19]. However, the impact of childhood urbanicity on the neural correlates of episodic memory remains poorly understood.
Although urban environments can facilitate a higher average quality of life [
20], they may also be accompanied by an increased risk of neuropsychiatric disorders, including depression, autistic spectrum disorders, and psychosis [
21‐
24]. Additionally, a higher genetic risk for psychiatric disorders has been reported to affect individuals’ choice of residence [
25]. From twin studies, living in an urban environment is itself partially heritable [
26]. Although the idea that childhood environment is heritable may seem counterintuitive, work on behavioral genetics has long documented the heritability of many exposures perceived as environmental [
27]. This heritability is referred to as gene–environment correlation (rGE), and potential rGE mechanisms may be posited to explain the heritability of childhood environments [
28]. One such mechanism is “active” rGE, where individuals with genetic variants associated with certain behavioral phenotypes may be more prone to selecting adverse situations. For instance, genetic factors related to individuals’ nature experiences may lead children who experience more nature to benefit more from it [
29]. Therefore, we hypothesized that differences in episodic memories affected by urbanicity may have genetic influences, as reflected in brain activity. The significant loci found in genome-wide association studies may provide clues about the mechanism of partial heritability on the impact of urbanicity on episodic memory.
China has undergone large-scale urbanization since the 1980s, accompanied by its economic development [
30]. This gave us a unique opportunity to leverage China’s recent urbanization to examine different childhood rural-urban environmental effects on brain development, which are currently poorly understood. The aim of this study was to explore how the childhood environment could affect episodic memory brain function. To achieve this, we first compared aversive or neutral episodic memory performance across participants from rural and urban childhood environments. Second, we explored the effect of urbanicity on brain activity and investigated the correlation between brain activity and performance. Third, considering that the mechanism of urbanization effects on the brain activity associated with episodic memories remains unknown and may be influenced by genetic variation, we also conducted genome-wide genotyping of the discovery sample. Subsequently, we performed a genome-wide association study with urban-rural differences in brain activity as the dependent variable to explore the potential genetic effects. Participants in this study were balanced between their current urban environments and genetic backgrounds [
18], thus maximizing the effect of childhood urban versus rural upbringing.
Discussion
In this study, we examined the effects of urban and rural childhood on episodic memory and related brain function in a large and genetically homogeneous sample of young healthy Han Chinese adults. Despite having similar current educational and occupational status, these individuals differed in their urban or rural childhood during China’s rapid and large-scale urbanization. Rural childhood appeared to result in stronger brain activation during encoding, especially in the striatum; and stronger brain activation during retrieval, especially in the hippocampus. These findings were replicated in an independent sample. The different urban–rural striatal emotional encoding functions were associated with the NTRK2 common variant, which exhibited a higher distribution of the T allele in the urban group. These findings suggest that effects on striatal function may be related to childhood urbanicity, emotional memory processes, and genetic variability within NTRK2.
Both groups exhibited reduced accuracy to aversive stimuli during the encoding session, indicating a tendency to avoid aversive stimuli. However, the urban group demonstrated a smaller decrease, which suggests that the urban participants were less likely to neglect aversive stimuli compared to their rural counterparts. The absence of a significant interaction effect between urbanicity and valence in the replication sample may be attributed to the limited sample size, necessitating a more cautious interpretation of behavioral measures. During the retrieval session, episodic memory for aversive stimuli was generally better than for neutral stimuli, consistent with the memorability and ease of recall of aversive stimuli [
38]. However, participants with rural childhoods showed less of this effect under aversive stimuli in both cohorts, which may reflect a tendency toward enhanced memory for negative stimuli among urban participants.
During the encoding phase, the rural group showed greater activation in the dorsal striatum than the urban group in both the discovery and replication samples, particularly in response to aversive stimuli. The striatum, especially the caudate nucleus [
42], is involved in the formation of declarative memory, the process by which episodic elements are bound to a complete memory trace. One possible explanation for this activation is that networks involved in incremental learning, including the striatum, contribute to the binding process in the formation of integrated episodes [
43]. Considering the role of the striatum in processing reward stimuli, the removal of an image may serve as a reward. The lower engagement of the striatum in the urban group may reflect weaker encoding of this positive stimuli. Given that striatal activation has also been associated with aversive learning, a potential reason may be that participants are contemplating how to avoid the potentially negative outcome [
44]. In our sample, rural volunteers, when confronted with aversive stimuli, exhibited higher levels of brain activation and lower recall accuracy, aligning with this hypothesis. Furthermore, the higher accuracy demonstrated by the urban group when facing aversive stimuli may suggest a more automatic processing of negative information among adults with an urban upbringing.
During the retrieval phase, the rural group showed greater activation in the hippocampus and parahippocampal gyrus compared to the urban group, particularly in response to aversive stimuli. The hippocampus is involved in emotional memory recall and regulation [
45,
46]. While rural participants exhibited increased hippocampal engagement, suggesting a greater investment in memory recall, they still displayed lower accuracy than their urban counterparts when facing aversive stimuli. However, their accuracy levels were similar when encountering neutral stimuli. These results may reflect a protective mechanism of emotional regulation that successfully inhibits negative memory recall.
Previous studies have hypothesized that memory encoding competes for striatal processing in the hippocampus [
47,
48]. If this hypothesis holds true, the stronger striatal activation observed in rural participants during encoding could indicate selective neglect of aversive episodic memory through inhibition of hippocampal function. During the retrieval session, despite rural participants displaying increased hippocampal activity, their accuracy remained lower than that of the urban participants. This could be attributed to their selective suppression of the encoding of negative stimuli. Therefore, we suggest that rural participants possess an adaptive striatal function that is less engaged by aversive stimuli, and therefore less affected by or more resilient to stress.
On the other hand, another possibility is that urban participants may more automatically focus on and remember aversive stimuli than their rural counterparts. The pathological process of enhancing negative memories or neglecting positive memories has been observed in conditions including depression, anxiety, and PTSD [
8,
10,
49]. A meta-analysis indicated that, when exposed to negative stimuli, patients with depression exhibited lower striatal response levels compared to healthy volunteers, possibly due to decreased striatal dopamine levels when confronted with negative information [
50]. Compared to rural volunteers, the brain activation patterns of urban volunteers were analogous to that observed in the depression models. In addition, the reduced prefrontal engagement of urban participants under aversive stimuli may indicate hypersensitivity to negative events, suggesting a potential “sensitizing effect” in urban participants [
51].
Our exploratory Genome-Wide Association Studies (GWAS) analysis suggesting genetic variability within
NTRK2 gene affected the striatal processing of aversive stimuli in relation to childhood urbanicity was based on a modest sample of over 400 participants (power = 98.62% calculated by Quanto [
52]).
NTRK2 (also known as Tropomyosin receptor kinase B,
TrkB) is activated by several neurotrophins and serves as a high-affinity receptor for brain-derived neurotrophic factor (BDNF). Although these findings are consistent with previous suggestions that correlations between SNP data and task-related brain imaging data can offer clues about genetic mechanisms [
53], some caution is warranted. Firstly, we note that GWAS of complex diseases including in psychiatry, and in psychological constructs (e.g., intelligence quotient, psychological traits) [
54,
55] has generally implicated genetic variants with small effect sizes. Our limited sample size thus limits the power to detect additional small genetic and/or environmental effects that could be present, despite the large environmental differences in this unique, genetically homogenous sample, and the relatively large neuroimaging samples herein. For similar reasons, it is also possible that effect sizes here are overestimated. Replication in independent samples is thus needed, which we provide. Here, we again observe the differential striatal activity between two genotype groups, suggesting that the involvement of BDNF signaling in these effects could be robust. Nevertheless, we suggest that future work should be needed in larger populations, and indeed in populations with improved characterization of features in the urban environment (e.g., local-level pollution, density, noise). BDNF expression and downstream signaling through the TrkB receptor are essential for memory formation in aversive domains [
56]. BDNF-TrkB signaling within the ventral tegmental area (VTA) - nucleus accumbens (NAc) circuit has also been reported as a pathological mechanism during periods of chronic stress, resulting in depression [
57]. Accumulating evidence suggest a relationship between
NTRK2 and a broad range of psychiatric disorders, especially those associated with stress, including depression, schizophrenia, and anxiety disorders. Genotype-dependent differences in
NTRK2 have been observed in white matter properties among patients with depression [
58] and have been linked to emotional arousal in healthy individuals [
59]. Furthermore,
NTRK2 plays a role in modulating fear learning and synaptic plasticity in the amygdala [
60]. Our discovery of
NTRK2 common variations associated with differential urban-rural striatal encoding activities in aversive conditions through genome-wide association analysis supports the potential role of
NTRK2 in emotion dysregulation. The high expression levels of
NTRK2 in the putamen and frontal cortex throughout the lifespan, along with genotype-related expression patterns, suggest its involvement in general neurodevelopmental processes underlying stress and emotion, which may have implications for the mental health of individuals undergoing urban migration.
This study had several limitations. Firstly, the rural group of our subjects was slightly older, and this could potentially have influenced the results, although we did control for age as a factor. Secondly, because our analysis was performed on a population of relatively highly educated individuals in large cities, it may not fully represent people who still reside in rural areas. Further research should aim to include participants with varying socioeconomic statuses, encompassing both patients and healthy individuals, while incorporating variables that could directly reflect early life rural–urban features and considering additional potential confounding factors. Thirdly, our GWAS findings that suggest the involvement of BDNF signaling in the striatal processing of aversive stimuli were exploratory. Although the differential striatal activity had been observed between two genotype groups in the replication sample, more replications were needed in a larger sample. Fourthly, the use age 12 in defining childhood urbanicity is not meant to suggest any empirical evidence that the age of 12 constitutes some critical juncture for the impact of urbanicity on neurodevelopmental processes. Rather, this category variable was selected on the basis that at least in our population context (and many others), age 12 is a convenient demarcation between elementary and middle or secondary school. Moreover, we note that if instead of using such a categorical urbanicity variable, we use a continuous urbanicity variable such as the urbanicity score [
17] (Additional file
1: Figure S5), similar results are obtained.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.