Depression: Human brain systems have significantly different pattern of adolescent development in females vs males; developmental differences that are located in cortical areas/subcortical nuclei are psychologically, genomically, clinically relevant

Sexually divergent development of depression-related brain networks during healthy human adolescence. Lena Dorfschmidt et al. Science Advances, May 27 2022, Vol 8, Issue 21, DOI 10.1126/sciadv.abm7825

Abstract: Sexual differences in human brain development could be relevant to sex differences in the incidence of depression during adolescence. We tested for sex differences in parameters of normative brain network development using fMRI data on N = 298 healthy adolescents, aged 14 to 26 years, each scanned one to three times. Sexually divergent development of functional connectivity was located in the default mode network, limbic cortex, and subcortical nuclei. Females had a more “disruptive” pattern of development, where weak functional connectivity at age 14 became stronger during adolescence. This fMRI-derived map of sexually divergent brain network development was robustly colocated with i prior loci of reward-related brain activation ii a map of functional dysconnectivity in major depressive disorder (MDD), and iii an adult brain gene transcriptional pattern enriched for genes on the X chromosome, neurodevelopmental genes, and risk genes for MDD. We found normative sexual divergence in adolescent development of a cortico-subcortical brain functional network that is relevant to depression.

DISCUSSION

This study was motivated by the twin hypotheses that there are sex-divergent differences in brain functional network development of healthy adolescents and that these normative developmental differences are located in cortical areas and subcortical nuclei that are psychologically, genomically, and clinically relevant to depression. In this accelerated longitudinal fMRI study of healthy young people, we first identified human brain systems that demonstrated a significantly different pattern of adolescent development in females compared to males. We found sex differences in several aspects of FC: Females had lower global mean FC across all ages and reduced nodal strength of connectivity in most regional nodes at 14 years, FC14. However, there were more anatomically specific sex differences in two developmentally sensitive parameters: The rate of change in FC during adolescence, FC14 − 26, was significantly reduced in females for connections between one cortical nucleus (nucleus accumbens) and 27 cortical structures, and the MI, a coefficient of the linear relationship between edgewise FC14 and FC14 − 26 at each node, was significantly more negative in females for 107 cortical areas concentrated in the DMN, ventral attentional, and limbic networks, as well as subcortical nuclei.

The MI can be used to define two modes of adolescent brain functional network development (6). A conservative node is defined by a positive MI, indicating that it is highly connected or “hub-like” at baseline (14 years) and becomes even more strongly connected over the course of adolescence (14 to 26 years). Theoretically, conservative nodes could also be weakly connected at baseline and become even more weakly connected during adolescence; however, empirically, we found that this was not the case (fig. S14). A disruptive node is defined by a negative MI, indicating either that it is weakly connected at age 14 but becomes more strongly connected or hub-like during adolescence or that it is a strongly connected node at 14 years but becomes more weakly connected or less hub-like during adolescence. The disruptive developmental profile of weak-getting-stronger during adolescence hypothetically represents a “rewiring” in the functional connectome, which could be relevant to the acquisition of social, cognitive, and other skills (6). Similar selective strengthening of connections has also been observed on the cellular level in the developing Caenorhabditis elegans connectome (63). It has also been argued that brain networks that are most developmentally active during adolescence are most likely to contribute to the coincidentally increased risk of mental health symptoms, i.e., “moving parts get broken” (11). For these reasons, our analysis focused particularly on sexual differences in weak-getting-stronger disruption in cortico-subcortical networks; results for strong-getting-stronger or conservative development are summarized in fig. S16.

The first explanation that we considered for this sex difference in developmental fMRI parameters is that they were attributable to sex differences in potentially confounding variables, including head motion during scanning. Head movement is known to be a potentially problematic confounder in developmental fMRI (19–21), and males, especially younger males, had more head movement than females in this sample. We initially addressed this issue by a two-stage preprocessing pipeline that statistically corrected each participant’s functional connectome for between-subject differences in head motion, indexed by FD. These preprocessed data passed the standard quality control criteria for movement-related effects on FC. In addition, we conducted three sensitivity analyses of head movement, repeating the entire analysis for male and female data separately, for a “motion-matched” subset of the data in which there was no significant sex difference in FD, and for all data after GSR (figs. S20 to S33) (24). In parallel, we conducted two additional sensitivity analyses to assess whether the male > female differences in intracranial volume, or global FC, might have confounded our principal results. In all five sensitivity analyses, our key results were qualitatively and quantitatively conserved, e.g., ΔMI maps estimated by the principal analysis were strongly correlated (mean r ∼ 0.8) with corresponding maps estimated by each sensitivity analysis. We therefore consider that sex differences in head movement, intracranial volume, and global FC can be discounted as sufficient explanations for sex differences in these parameters of brain network development.

An alternative explanation is that sex differences in FC14 − 26 and MI reflect divergent development of specific cortico-subcortical circuits. In particular, females have a significantly more disruptive pattern of adolescent development, indexed by negative ΔMI, because functional connections that were weak at 14 years became stronger, and connections that were strong became weaker, over the course of adolescence. This sex difference in terms of FC could be related to sex differences in an underlying process of reconfiguration or remodeling of cortico-subcortical connectivity at a synaptic or neuronal scale. To assess the plausibility of this biological interpretation, we used preexisting data on human brain gene expression, and the dimension-reducing multivariate method of PLS to identify the set of genes that were most over- or underexpressed in brain regions corresponding to the divergent system defined by developmental fMRI. Enrichment analysis demonstrated that the genes that were most strongly expressed in brain regions with more disruptive (or less conservative) development in females included significantly more X chromosome genes than expected by chance. The same set of genes was also significantly enriched for genes that are known a priori to be expressed in cortical areas during early (perinatal) development and in subcortical structures, such as amygdala, during adolescent development.

Sexual differentiation of the brain has been proposed to occur in two stages: an initial “organizational” stage before and immediately after birth and a later “activational” stage during adolescence (64). It has long been argued that these events are driven by gonadal hormones. However, more recent work suggests a complex interplay of sex chromosomes and their downstream products leading to sexual differentiation of brain cells (65–67). The results of our enrichment analysis, indicating colocation of the sexually divergent fMRI-derived map with brain regions enriched for expression of X chromosomal and neurodevelopmental genes, are compatible with interpretation of adolescent change in fMRI connectivity as a marker of an underlying program of transcriptional changes in genes previously linked to postnatal sexual differentiation at a neuronal level.

We assessed the relevance to depression of this sexually divergent profile of brain network development in several ways. Anatomically, the DMN and subcortical structures that had more disruptive development in females, e.g., ventral medial prefrontal cortex, medial temporal gyrus, and anterior and posterior cingulate cortex, have previously been implicated as substrates of depressive disorder (68, 69). This anatomical convergence was quantified by the significant spatial correlation between the whole brain map of sex differences in MI and an independent map of MDD case-control differences in nodal degree of FC. Cortical and subcortical areas with reduced degree of connectivity or “hubness” in MDD cases had more disruptive development in adolescent females. Genomically, the list of genes transcriptionally colocated with this divergently developing network was enriched for risk genes from prior genome-wide association studies of MDD. Further contextualizing the genes that were found to be significantly overexpressed in regions displaying more disruptive development in females, we noticed that this list included two (SST and NPY) of three genes previously reported (70), as specifically expressed by adult neuronal and glial cells and linked to neuroimaging phenotypes of depression (fig. S44). It is also notable that MDD has been previously associated with up-regulation of X-linked escapee genes and genes that control X-inactivation (71). Psychologically, by meta-analysis of a large prior database of task-related fMRI studies, we found that brain regions comprising the sexually divergent system were psychologically specialized for reward- and emotion-related processes that are fundamental to core depressive symptoms, e.g., anhedonia. Collectively, these results do not prove that there is a causal relationship between sexually divergent brain development and risk of depression. However, they demonstrate that there is a sexually divergent process of adolescent development of a cortico-subcortical system that is anatomically, genomically, and psychologically relevant to depression. These insights motivate and focus future studies purposively designed to test the hypothesis that sexual divergence of adolescent brain development causes contemporaneous or subsequent sex differences in the risk for mood disorders.

It is increasingly recognized that clinical phenotypes and genetic and environmental risk factors may be shared in common between depression and other mental health disorders arising in adolescence (72, 73). In particular, abnormalities in fMRI connectivity have been reported as trans-diagnostic phenotypes, characteristic of multiple, diagnostically distinct disorders (72), and risk genes associated with individual mental health and neurodevelopmental disorders have been found to overlap across disorders, implying that some genes confer trans-diagnostic risk for multiple neuropsychiatric disorders (73). In this context, it is reasonable to ask whether the significant associations that we have demonstrated between ΔMI and both fMRI and genetic data on MDD are specific to depression or whether they are representative of a trans-diagnostic association between ΔMI and functional dysconnectivity and/or risk genes for mental health disorders more generally. As a first step in addressing this question, we tested for spatial colocation of the ΔMI map and a map of functional dysconnectivity derived from a prior case-control fMRI study of schizophrenia. We found no significant association, indicating that the abnormalities of FC associated with adult schizophrenia do not coincide anatomically with the cortico-subcortical network that demonstrated sex differences in adolescent development. In a second step, we tested for enrichment by schizophrenia-associated genes of the list of genes that were identified by PLS analysis as transcriptionally colocated with the ΔMI map. We found no evidence for significant enrichment of this gene list by risk genes for schizophrenia. In summary, these two specificity analyses indicated that the brain systems demonstrating sexually divergent development in adolescence were not anatomically or genetically linked to schizophrenia, suggesting that this normative neurodevelopmental process may be specifically relevant to depression. However, we note that we have only tested for a relationship between ΔMI and two mental health disorders (MDD and schizophrenia). It will be important in the future to explore this relationship across a wider range of disorders to characterize its diagnostic specificity more comprehensively and conclusively. It is conceivable that sex differences in development of this system could be relevant to sex differences in risk for other mental health disorders.

Methodological limitations

It is a strength of the study that our analysis of sexually divergent brain network development is based on a large, accelerated longitudinal fMRI dataset with approximately equal numbers of males and females in each stratum of the adolescent age range. However, previous work has found substantial overlap in male and female distributions of multiple brain measures (74, 75), and the metrics analyzed here (FC14, FC14 − 26, and ΔMI) are group-level parameters. Thus, all reported sex differences are reflective of a group mean difference, estimated from FC distributions that substantially overlap between the sexes. On this basis, we are not arguing that female and male brains are distinctly dimorphic (76). Furthermore, this study included only data on biological sex such that we cannot comment on the effects of gender.

Limitations of the study include our reliance on gene expression maps from postmortem examination of six adult, mostly male, brains. This dataset is used widely and has been invaluable in shedding new light on the molecular correlates of neuroimaging phenotypes (77). Biological validation of sexually divergent adolescent development of this cortico-subcortical system derived from fMRI would be more directly informed by sex-specific human brain maps of whole-genome transcription in adolescence, but to the best of our knowledge, these data are not currently available. It will also be important in the future to test the hypothesis that an anatomically homologous cortico-subcortical system has divergent adolescent development in animal models that allow more precise but invasive analysis of the cellular and molecular substrates of fMRI phenotypes than is possible in humans.

Here, we used spin tests to correct for the confounding effects of spatial autocorrelation. Spatial autocorrelation of statistical brain maps can cause inflated estimates of the probability of spatial colocation or correlation between two maps (78). The spin-test procedure addresses this issue by conserving the spatial autocorrelational structure of the maps by randomly “spinning” or spherically rotating each map over the surface of the brain and calculating the spatial colocation statistic after each spin permutation (79). Other methods for testing spatial colocation in the context of spatial autocorrelation have been proposed, and this remains an active focus for ongoing research, especially in relation to colocation of neuroimaging phenotypes and brain gene transcriptional maps (78).

Social and environmental factors are relevant modulators of psychiatric disorders (80) and have not been assessed in this study. These factors (i) can be neurodevelopmentally relevant, i.e., childhood socioeconomic status influences the pace of brain development (81), and (ii) can help explain sex and gender differences in mental health outcomes, i.e., previous studies have demonstrated a relationship between social inequality and gender disparities in mental health (82). This naturally leads to the question of how sexually divergent functional network development might be modulated by socioeconomic deprivation or other environmental risk factors for mental health disorder. We suggest that deeper understanding of these potential interactions between biological programs of sexually divergent brain development on one hand and gendered or generic social stressors in childhood and adolescence on the other hand will be an important strategic goal for the future of mental health science.