INSIGHTS | PERSPECTIVES NEUROSCIENCE Neurotechnology to address big questions Profiling of single neurons in tissue allows structure and function linkage in brain circuits By Thomas Knöpfel specific transcript [messenger RNA (mRNA)] lar diversity and brain state–dependent tran- by a probe from a library of gene-specific scription. STARmap can provide information motions, cognition, and conscious- primers. These cDNAs were then rolling-cir- about the transcriptome (up to 10,000 genes) ness emerge with the processing of cle-amplified so as to generate a DNA nano- of up to 30,000 cells in a cortical volume pre- neuronal information provided by ball (amplicon) that contains multiple copies pared from an individual mouse brain. memory readouts and senses. How of the cDNA templates (6). Importantly, this Hydrogel-tissue chemistry (5) was needed this actually works seemed for a long process takes place in structurally conserved to fix the amplicons in the cubic millimeter time to be a question too big to ask. brain tissue. Formation of the circular cDNAs tissue samples. Classification of cortical cell EAdvances in neuroscience have always been included a barcode as a distinct identifier, type based on anatomical location, gene ex- driven by methodological inventions, and which was decoded with a newly designed pression pattern, and functional properties current efforts to develop neurotechnolo- error-reduced in situ sequencing technology has been performed in the past by accumulat- gies are motivated by experimental strate- to identify specific mRNAs (5). ing data across many experiments and many gies, including analytical dissection (that is, Since the development of techniques that specimens (7). This knowledge served the Downloaded from inverse engineering), large-scale interroga- involve mRNA harvesting from single cells authors to identify a set of 160 target genes, tion, and synthetic reconstruction of the for sequencing—single-cell reverse transcrip- including 112 putative cell-type marker genes mammalian cerebral cortex (which is im- tion–polymerase chain reaction (scRT-PCR)— and 48 genes that are expressed in response portant for higher cognitive to neuronal activity pat- functions) and connected terns. It is not known which brain structures (1, 2). As Profiling brain tissue of the selected marker genes http://science.sciencemag.org/ part of this neurotechnolog- Wang et al. provide a method to determine the activity of marker genes within a sample of are most indicative for spe- ical endeavor, mapping the brain tissue. This allows identification and mapping, for example, of subtypes of excitatory cific cell types and which of transcriptome of neurons or inhibitory neurons in the cortical layers, corpus callosum, and hippocampus. the activity-regulated genes with single-cell resolution are recruited in which cell and with known three-di- types and under specific pat- Tissue collection mensional tissue localiza- Spatial distribution of distinct excitatory neurons terns of neuronal activity. tion has been a long-sought Mouse visual Therefore, Wang et al. used enabling technology (3). On cortex Hippocampus (HPC) Corpus callosum (cc) Cortical layers this set of target genes both to page 380 of this issue, Wang benchmark their methodol- et al. (4) present the tech- ogy and workflow but also to L6 L5 L4 L 2/3 L1 on July 30, 2018 nology and workflow to ac- contribute to efforts aimed at cess transcriptional states of detecting and classifying cell more than 100 genes from Mouse brain types and corresponding tis- up to 30,000 cells in a cubic V1 cortex sue-organization principles millimeter of cortical tissue. dissection in the neocortex of adult This advanced methodology In situ DNA sequencing mouse brain (see the figure). will facilitate studies that HPC cc L6 L5 L4 L 2/3 L1 A weakness of STARmap improve our understand- is the lack of direct linking ing of the neuronal hard- of transcriptomes with cel- ware and, when combined 0.3 mm lular protein expression and with other emerging neuro- function. However, because technologies, will enable big 8 mm 1.4 mm several of the chosen marker questions in neuroscience Dissected Spatial distribution of distinct inhibitory neurons genes encode proteins—in- to be addressed. sample cluding ion channels, recep- Wang et al. integrated tors, and neurotransmitter- knowledge of mouse cortex-specific gene ex- transcriptomics has been instrumental to related enzymes that indicate functional pression, targeted signal amplification, in situ classify the many functionally distinct cell properties—an indirect link between tran- transcriptomics, and hydrogel-tissue chemis- types found in the cerebral cortex (7). scRT- scriptomics and function is already provided. try (5) into a workflow for single-cell resolu- PCR in combination with electrophysiologi- In future implementations, a larger number tion spatial transcriptomics, called STARmap cal recordings has the advantage to directly of genes indicating functional properties may (spatially resolved transcript amplicon read- link specific gene expression patterns and replace marker genes with low cell type– out mapping). Targeted signal amplification cellular functional characteristics (includ- specificity to increase the possibilities for was achieved by generating circular comple- ing information on how specific neurons functional inference. SCIENCE mentary DNAs (cDNAs) after recognition of a integrate synaptic information and on their The high data content from a single tissue synaptic connectivity) but is not practical for specimen provided by STARmap allows for Department of Medicine and Centre for Neurotechnology, more than 100 cells (7). Scaling this approach correlation of gene activation patterns within Imperial College London, UK. Email: [email protected] is required for a systematic analysis of cellu- an individual’s specific brain state at the time BICKEL/ C. GRAPHIC: 328 27 JULY 2018 • VOL 361 ISSUE 6400 sciencemag.org SCIENCE Published by AAAS DA_0727Perspectives.indd 328 7/25/18 11:51 AM of tissue preparation. As a proof of principle, GENE EXPRESSION the authors compared the transcriptome from two cohorts of mice: The first was ex- posed for 1 hour to light after 4 days of hous- Dynamic condensates ing in darkness, whereas the second cohort was housed continuously in darkness before tissue collection. STARmap analysis con- activate transcription firmed and refined our knowledge about the recruitment of activity-dependent genes and Transcriptional components exhibit their enhancer RNAs (which control tran- transient phase separation to drive gene activation scription of their target genes). In the future, a larger range of physiological brain states as well as rodent models of brain diseases, and By Aaron J. Plys1,2 and Robert E. Kingston1,2 tories,” where components are organized and even human tissue, may be analyzed this way. ready to act on a gene that goes to the cellular The STARmap data are a valuable source very aspect of human function, from location of the factory (1). On pages 378, 379, for data mining to identify patterns and proper cell differentiation and develop- and 412 of this issue, Chong et al. (2), Sabari higher-level structural and functional prin- ment to normal cellular maintenance, et al. (3), and Cho et al. (4), respectively, argue ciples, many of which are likely still hid- requires properly timed activation of that special protein domains, which interact den behind the complex nature of cortical the necessary genes. This requires tran- with each other to form fleeting or more per- circuits and their function. The first appli- scription of genomic DNA into messen- sistent interactions, form biomolecular con- cations of STARmap already revealed an in- Eger RNA (mRNA), accomplished by RNA densates that concentrate the transcription Downloaded from triguing finding: a short-range self-clustering polymerase II (RNA Pol II), which initiates machinery. Some of these condensates might organization of inhibitory neuron subtypes transcription at gene promoters. This highly even form droplets, generating a liquid phase (3). These clusters may just be relicts of cell regulated process requires hundreds of pro- separated from the rest of the nucleus. Phase differentiation and cortical morphogen- esis with little functional consequences, but more exciting are the possible functional Dynamic transcription machinery clustering during gene activation http://science.sciencemag.org/ implications. Among these is the increased Transcription factors (TFs) and coactivators condense into high-concentration clusters in the nucleus. opportunity of these inhibitory neuron sub- Condensation is mediated by low-complexity disordered regions (LCDRs) in these proteins. These clusters types to form direct electrical connections can incorporate RNA Pol II through transient interactions to efficiently activate gene transcription. with cell bodies and the proximal processes of neighboring cells. These gap junctions are Super- Gene TF LCDR or involved in the generation of high-frequency enhancer promoter coactivator LCDR rhythmic circuit activities (8). Super-enhancer clustering of Mediator or BRD4 and transient STARmap is well suited for further inter- interaction with RNA Pol II facing with other recent technologies that have been developed to address the big ques- on July 30, 2018 tions on brain functions. The technology clos- mRNA est in reach for combination with STARmap is functional in vivo optical imaging with ac- tivity indicators