ARTICLE https://doi.org/10.1038/s41467-020-16710-x OPEN Transcriptional and imaging-genetic association of cortical interneurons, brain function, and schizophrenia risk Kevin M. Anderson 1, Meghan A. Collins1, Rowena Chin1, Tian Ge2,3, Monica D. Rosenberg 1,4 & ✉ Avram J. Holmes 1,3,5 fl 1234567890():,; Inhibitory interneurons orchestrate information ow across the cortex and are implicated in psychiatric illness. Although interneuron classes have unique functional properties and spatial distributions, the influence of interneuron subtypes on brain function, cortical specialization, and illness risk remains elusive. Here, we demonstrate stereotyped negative correlation of somatostatin and parvalbumin transcripts within human and non-human primates. Cortical distributions of somatostatin and parvalbumin cell gene markers are strongly coupled to regional differences in functional MRI variability. In the general population (n = 9,713), parvalbumin-linked genes account for an enriched proportion of heritable variance in in-vivo functional MRI signal amplitude. Single-marker and polygenic cell deconvolution establish that this relationship is spatially dependent, following the topography of parvalbumin expression in post-mortem brain tissue. Finally, schizophrenia genetic risk is enriched among interneuron-linked genes and predicts cortical signal amplitude in parvalbumin-biased regions. These data indicate that the molecular-genetic basis of brain function is shaped by interneuron-related transcripts and may capture individual differences in schizophrenia risk. 1 Department of Psychology, Yale University, New Haven, CT 06520, USA. 2 Psychiatric and Neurodevelopmental Genetics Unit, Center for Genomic Medicine, Massachusetts General Hospital, Boston, MA 02114, USA. 3 Department of Psychiatry, Massachusetts General Hospital, Harvard Medical School, Boston, MA 02114, USA. 4 Department of Psychology, University of Chicago, Chicago, IL 60637, USA. 5 Department of Psychiatry, Yale University, New ✉ Haven, CT 06520, USA. email: [email protected] NATURE COMMUNICATIONS | (2020) 11:2889 | https://doi.org/10.1038/s41467-020-16710-x | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-16710-x amón y Cajal theorized that the functional diversity of the be a core locus of disruption in schizophrenia, leading to altered human brain arises, in part, from the vast assortment of gamma-band signals and working memory deficits that are a R 1 21 neurons that pattern cortex . Inhibitory interneurons are hallmark of the disorder . However, a direct link between the most varied neuronal class2, exhibiting divergent morpholo- PVALB-related genetic variation and human brain activity has yet gical and physiological properties while coordinating information to be established. Linking cortical interneurons to individual flow across the brain’s collective set of connections (functional differences in human brain function would yield deep biological connectome)3. Converging animal and human work provides insight into the hemodynamic BOLD signal, providing an engine evidence for the role of interneurons in healthy brain function as for the discovery of functional connectome-linked genes and well as their dysregulation in psychiatric illnesses, including associated risk for illness onset. schizophrenia4 and major depressive disorder5. The development Here, we bridge genetic, transcriptional, and neuroimaging of dense spatial transcriptional atlases now enables the study of data to advance three lines of inquiry linking interneurons to cellular and molecular associates of functional brain networks. human brain function. First, we describe the organization of SST Early work in this area identifies genes encoding ion channels6 and PVALB expression in human and non-human primates, and those enriched in supragranular layers of cortex7 as correlates demonstrating a robust pattern of anti-correlation across cortex of large-scale network organization. Other research indicates that and subcortex. We perform single-cell polygenic deconvolution22 cortical resting-state signal fluctuations follow the expression of of bulk cortical tissue data to infer spatial distributions of inhi- neuron-enriched genes8. However, the transcriptional correlates bitory, excitatory, and non-neuronal cells across cortex, providing of brain function have rarely been validated in-vivo6 and little is converging evidence with single-marker analyses. Second, we known about how the spatial distribution of specific interneuron establish that the relative density of SST and PVALB tracks subtypes shapes cortical function and associated risk for psy- regional differences in cortical brain activity. In a large sample (N chiatric illness in humans. = 9713)23, genetic variation among PVALB-correlated genes The spatial distribution of interneuron subtypes is theorized to explained an enriched proportion of heritable variance in resting- contribute to regional specialization of cortex, partly by altering state signal amplitude (RSFA), in a manner that mirrors the the relative balance of excitation and inhibition for a given cor- spatial expression of PVALB measured in independent post- tical area9–11. Interneurons comprise 20–30% of cortical neu- mortem brain tissue. These discoveries suggest that the rons12 and form stereotyped connections with excitatory molecular-genetic basis of cortical function is spatially nonuni- projection neurons2,3. The majority of interneurons express one form and that genes linked to PVALB interneurons explain of a limited set of genetic markers: somatostatin (SST), parval- heritable aspects of the BOLD signal. Third, we link PVALB bumin (PVALB), and vasoactive-intestinal peptide (VIP; a subset interneurons and psychotic illness, demonstrating that genetic of HTR3A interneurons)2. Each subtype possesses unique risk for schizophrenia is enriched among interneuron-linked synaptic and functional characteristics, leading to the hypothesis genes and predicts reduced resting-state signal amplitude in a that the ratio of interneuron classes underpins local differences in spatially heterogeneous manner that follows PVALB expression. neural activity11,13. For example, SST interneurons preferentially These data help address a deep-rooted challenge in neuroscience target dendrites of cortical projection neurons to regulate their to understand how cytoarchitecture shapes human brain function input, whereas PVALB interneurons synapse on perisomatic and related vulnerability for psychiatric illness. regions to regulate output2. Consequently, increased relative density of SST interneurons may facilitate filtering of noisy or task-irrelevant cortical signals and promote recurrent excitation Results required for higher-order cognition11,14. Conversely, relative Anti-correlation of SST and PVALB interneuron markers increases in PVALB may produce stronger feedback inhibition on across cortex. The properties of interneuron subtypes emerge excitatory neurons, leading to shorter activation timescales suited early in development and are partly determined by their spatial for processing constantly changing sensorimotor stimuli11,15. origin in the embryonic ganglionic eminence24. SST and PVALB Previous work documents cortical gradients of gene transcription interneurons originate in the medial ganglionic eminence (MGE) that mirror the hierarchical organization of timescales from fast along negatively correlated spatial gradients25. That is, PVALB- (unimodal cortex) to slow (multimodal cortex)8,10,16,17. Further, and SST-destined neurons differentially cluster within the dorsal SST and PVALB interneuron markers are differentially expressed and ventral MGE, respectively26. Evidence in humans9,10 and within distributed limbic and somato-motor cortico-striatal net- rodents11,17 suggests that SST and PVALB transcripts maintain a works, respectively9. These observations suggest that spatial dis- negative spatial correlation in adulthood, indicating that tributions of interneuron subtypes could underlie regional embryonic organization may constitute a “proto-map” of mature signaling differences across the cortical sheet, as indexed by blood cortex. Although prior research suggests that SST and PVALB oxygenation level-dependent (BOLD) functional magnetic reso- markers are differentially expressed across cortex9,10, the current nance imaging (fMRI). However, the psychiatric and functional work directly establishes transcriptional anti-correlation between consequences of spatially variable SST- and PVALB-related these two cell types across multiple techniques, human datasets, transcription in cortex have yet to be fully characterized. non-human primate data, and human neurodevelopment. This Establishing the principles by which cellular diversity influ- comprehensive profiling of SST and PVALB interneuron ences brain function is a long-standing challenge in neuroscience expression is necessary for deep profiling of their relationship to and could reveal biological mechanisms of individual variability in-vivo brain function and subsequent schizophrenia risk through of the human brain. Consistent with this aim, cross-species evi- statistical genetic approaches. The functional consequences of a dence indicates the importance of PVALB interneurons for fMRI negative spatial SST to PVALB relationship are not well under- measures of brain function18. PVALB interneurons orchestrate stood, but the presence of replicable and evolutionarily conserved gamma-band oscillations