sign in sign up
<<
Home , History of neuroscience

ViewpoinT

medium of conceptual space is: how do we combine these twin paths to understand Reflections on the past two decades the transmission of representations in the network of the ? Although answering of that question will surely build on the prior paradigms, neither seems equipped to Danielle S. Bassett , Kathleen E. Cullen , Simon B. Eickhoff , provide a complete solution; the former offers the dots within, whereas the latter Martha J. Farah , Yukiko Goda , Patrick Haggard , Hailan Hu , offers the lines between. Yet obtaining such Yasmin L. Hurd , Sheena A. Josselyn , Baljit S. Khakh , Jürgen A. Knoblich , a solution is important for deciphering the Panayiota Poirazi , Russell A. Poldrack , Marco Prinz , Pieter R. Roelfsema , mechanism of the brain’s fundamental goal: Tara L. Spires-Jones , Mriganka Sur and Hiroki R. Ueda information processing. The efforts towards such a solution that I find particularly Abstract | The first issue of Nature Reviews Neuroscience was published 20 years promising are those that move beyond ago, in 2000. To mark this anniversary, in this Viewpoint article we asked a selection descriptive characterization or correlative of researchers from across the field who have authored pieces published in the approaches, and towards theory; that is, journal in recent years for their thoughts on notable and interesting developments theory instantiated in an interpretable and generalizable mathematical model in neuroscience, and particularly in their areas of the field, over the past two that encodes a conceptual principle, and decades. They also provide some thoughts on current lines of research and theory as supported by (but not composed questions that excite them. of) computational and machine-learning approaches applied to data shared publicly in the spirit of open science. Progress along Broadly, how has the field of neuroscience and a growing (rather than diminishing) this tack will require more interdisciplinary changed over the past two decades, and undercitation of women in the reference lists graduate training programmes that provide what do you regard as the most notable or of neuroscience journal articles in the past equal depth in theory and experiment. interesting developments within your area of the 20 years. Each of us as citizens of science field over that period? What do you think are can choose to own and mould our culture Kathleen E. Cullen. Traditionally, the field the most interesting, big-picture questions being to realize a more equitable future. of neuroscience has asked the question: how asked in your area currently? Which lines of As a practice in the acquisition and is sensory input (afference) transformed research in your area of neuroscience are you curation of knowledge, neuroscience has into motor output (efference) to generate particularly excited about? expanded upon prior paradigms. In human appropriate behaviour? This conceptual neuroscience, I have been particularly struck framework — initially made popular by Danielle S. Bassett. Science is a culture by a concerted effort to move beyond the Sherrington — has long served as a common as much as it is a practice in the acquisition activation studies that were characteristic of foundation for systems . and curation of knowledge. As time passes, the brain-mapping era. Leading from that A fundamental development over the past the field changes its culture and practice. point in scholarly space, two particularly two decades, however, has been a collective As a culture, neuroscience is now composed pioneering paths of formal study extend: realization that it is necessary to reverse this of a more diverse group of scientists, along that of neural representations and that of traditional perspective. A series of recent the dimensions of sex, gender, sexual network systems. The former considers studies has established that motor signals orientation, race, ethnicity and disability. the pattern of activity across neural units profoundly influence sensory processing With the creation of anneslist (anneslist. that encodes an object, concept or state of during voluntary behaviour. In particular, net), Women in Neuroscience Repository information. The latter considers the pattern experiments across sensory systems have (winrepo.org) and BiasWatchNeuro of connections between neural units that can provided circuit-level insight into how (biaswatchneuro.com), there exist resources support the transmission of information and motor-related inputs to sensory areas to increase awareness of the work of women the modulation of network state in a manner selectively cancel self-generated sensory scientists, and to encourage a greater balance that alters the repertoire of computations inputs. This neural strategy underlies an in the speaker line-ups of conferences and accessible to a given neural population. essential property of the brain, namely its symposia. Yet explicit bias exists now as it The twin paths of representation and ability to distinguish externally applied did in 2006 when transgender Professor transmission move beyond the mapping of a (exafferent) from self-generated (reafferent) Ben A. Barres (Stanford University) wrote country’s borders, to record a city’s statistics sensory inputs. For instance, work by my his commentary for Nature entitled ‘Does and the transportation networks along group has established that at the Gender Matter?’ And implicit bias is perhaps which traffic flows. first central stage of vestibular processing even more pervasive, as reflected in peer To me, the question that creates giant preferentially encode sensory exafference. review, paper acceptance rates, grant funding resonance when dropped into the liquid Moreover, researchers studying other

Nature Reviews | NEURoSCiEnCE Viewpoint

sensory pathways (for example, visual, based on motor-related inputs is vital for and cognitive processing that is central to somatosensory and auditory) have both perceptual stability and accurate our subjective awareness of initiating and shown suppression of reafferent sensory . This computational strategy controlling our own volitional actions in the information by motor-related inputs also provides the flexibility required for world (that is, our sense of agency). In this at higher levels of sensory processing, fine-tuning and updating relationships context, a fundamental question logically including in the cortex. between motor signals and resultant arises: how exactly does the brain actually Indeed, much of the sensory input sensory feedback in order to compensate predict the sensory consequences of our that we experience in our everyday lives for changes to our body or the environment. actions in a dynamically changing world? is reafferent as it is generated by our own Furthermore, the ability to distinguish Two particularly exciting lines of research behaviour. The brain’s ability to predict between sensory exafference and reafference are currently addressing this question. the sensory consequences of our actions is a hallmark of higher-level perceptual First, emerging technologies including

The contributors Danielle s. Bassett is the J. Peter skirkanich Professor at the university role of astrocytes in neural circuits. The Khakh laboratory develops of Pennsylvania, Departments of Bioengineering, Electrical and Systems and uses novel genetic and optical methods to explore glial biology Engineering, Physics and Astronomy, , and . Her in vivo. expertise centres on the architecture, function, design and control of Jürgen A. Knoblich is Director of the Institute of Molecular Biotechnology complex systems, with a particular focus on large-scale neural systems of the Austrian Academy of Sciences in Vienna. He is a developmental in humans. studying development and psychiatric Kathleen E. Cullen is a professor of biomedical engineering at the Johns disorders. Researchers in his group have developed a method for growing Hopkins University School of Medicine. She holds joint appointments human brain tissue in the laboratory. They can recapitulate human in neuroscience and in otolaryngology. Her expertise in systems and embryonic brain development during the first trimester and analyse the computational neuroscience uses state-of-the-art methodologies that developmental defects leading to neurological disorders. manipulate and monitor neural circuits to understand the neural Panayiota Poirazi is a Research Director at the Foundation for Research computations that predict the consequences of natural self-motion and and Technology-Hellas (FORTH) in Heraklion, Crete, Greece. She uses the implications for perception and action. mathematical and computational techniques to develop experimentally Simon B. Eickhoff is a researcher at the Heinrich-Heine constrained models of single neurons and neuronal networks, aiming at University Düsseldorf and the Forschungszentrum Jülich. At the interface inferring the contributions of dendritic computations to complex brain between , data science and medicine, his work combines functions. , modelling of inter-individual differences and brain– Russell A. Poldrack is the Albert Ray Lang Professor in the Department of phenotype relationships as well as machine learning for single subject Psychology and Professor (by courtesy) of Computer Science at Stanford prediction of behavioural traits and precision medicine. University, and Director of the Stanford Center for Reproducible Martha J. Farah is a cognitive neuroscientist at the University of Neuroscience. His research uses neuroimaging to understand the brain Pennsylvania, where she directs the Center for Neuroscience & Society. systems underlying decision-making and executive function. His laboratory Her current research is focused on the relations between socio-economic also develops tools to help improve the reproducibility status and the brain. and transparency of neuroscience. Yukiko Goda is Deputy Director and a team leader at the RIKEN Center Marco Prinz is a neuropathologist at the University of Freiburg, Germany. for Brain Science. Her research addresses the cellular principles by which Currently, his research group aims to understand myeloid cell biology in synaptic strengths are set and dynamically modified at individual the CNS during health and disease and studies the impact of the CNS in relationship to other synapses in defined neural circuits. Recent endogenous immune system on the pathogenesis of neurological efforts in the group also target roles for astrocytes in regulation. disorders such as neurodegenerative and neuroinflammatory diseases. Patrick Haggard leads the ‘Action and Body’ research group at the Pieter R. Roelfsema has been Director of the Netherlands Institute for Institute of , University College London. Neuroscience, Amsterdam, since 2007. He studies visual perception and Hailan Hu is a professor at Zhejiang University School of Medicine, China. perceptual organization, as well as plasticity rules. He aims to develop a Her research interest focuses on how emotional and social behaviours are neuroprosthesis for the visual cortex to restore a rudimentary form of encoded and regulated in the brain, with a main focus on the neural vision in blind individuals. circuitry underlying depression and social dominance. Tara Spires-Jones is Professor of and a UK Dementia Yasmin L. Hurd is Professor of Psychiatry, Neuroscience and Research Institute Group Leader at the University of Edinburgh. Her group Pharmacological Sciences at the Icahn School of Medicine in New York studies the role of synapses in brain degeneration and resilience. Using and Director of the Institute in the Mount Sinai Behavioural pioneering imaging techniques, her group has discovered links between Health System. Her research focuses on opioid abuse and the genetic risk factors for Alzheimer disease and synaptic degeneration. developmental consequences of early drug exposure (particularly Mriganka Sur is a neuroscientist at the Massachusetts Institute of cannabis) using multidisciplinary and translational approaches in Technology, where he directs the Simons Center for the Social Brain. humans and animal models. His laboratory studies mechanisms of brain wiring and processing, Sheena A. Josselyn is Senior Scientist at the Hospital for Sick Children especially in the , and how they go awry in disorders (SickKids), a Professor at the University of Toronto in Canada and a of brain development. Senior Fellow in the Canadian Institute for Advanced Research (CIFAR). Hiroki R. Ueda is a professor at the University of Tokyo and a team leader She is interested in understanding how the brain uses information. in RIKEN BDR. He has an expertise in systems biology and focuses on Her laboratory uses a variety of experimental techniques to probe mammalian –wake cycles. He invented the whole-brain and memory function in mice. whole-body clearing and imaging methods called CUBIC and the Baljit S. Khakh is a Professor of and Neurobiology at University next-generation mammalian genetics for one-step production and of California Los Angeles. His expertise centres on glial biology and the analysis of knockout and knock-in mice without crossing.

www.nature.com/nrn Viewpoint

high-density recording across brain regions In this context, one of the effects, and subjects of study with the prefix ‘neuro’ and the ability to target perturbations in further driver, of large sample sizes is the have proliferated, from neuroaesthetics to neural circuits are providing new insight growing focus on individual inference using neurowarfare. This historical turn in the uses into how specific brain regions, such as machine learning. The mass-univariate of neuroscience arrived right around the the , actually perform this investigation of within-sample associations turn of the new century, so we are now about essential computation at the level of single is being increasingly abandoned in 20 years into the era of ‘neuroeverything’. neurons. Second, recent advances in our favour of multivariate models that are The idea of neuroeverything understanding of sensory processing better suited for capturing the richness understandably evokes eye rolls, within during natural behaviours have motivated of MRI data. The latter also allow the neuroscience and without. Many of these the investigation of neural circuits in prediction of inter-individual differences exuberant interdisciplinary connections will real-world environments. For example, in phenotypical (for example, behavioural) not turn out to be useful, but I believe that the brain must recalibrate its encoding of characteristics in new, previously unseen some will. Why? Because neuroscience is — sensory information in conditions in which individuals. Machine-learning approaches in principle — relevant to any field that seeks relationships between motor signals and may thus hold the key towards delivering to understand, predict or influence human reafferent sensory feedback are altered. Yet on some of the promises that have been behaviour. Whether and when it becomes studies of the neural basis of our sense of abundant in the literature ever since the relevant in practice is an empirical issue, direction and place have traditionally used start of neuroimaging. Throughout but I see no reason to rule it out yet. Indeed, virtual reality set-ups that unintentionally the decades, many neuroimaging studies there are glimmerings of real progress in generate persistent sensorimotor were ultimately motivated by clinical some of these fields. mismatches compared with natural issues, which can hardly be realized by took, as its initial focus, navigation and/or recorded only coarse group-mean differences but may become questions of criminal responsibility and measurements of behaviour (for example, possible using multivariate, predictive punishment. Neuroscience has helped shore head position). Ongoing experiments approaches. Even though further maturation up the rationale for separate standards of focused on understanding circuit-level and consolidation of the field is needed, punishment for juvenile offenders, given neural coding strategies during navigation it has brought real-life applications of their incomplete brain development, and has — using more precise and comprehensive neuroimaging research into reach. been cited in US Supreme Court opinions measurements of complex animal behaviour Finally, although hardly a new on the matter. Neurolaw now encompasses — are now beginning to elucidate the neural development but rather a controversy issues beyond responsibility, reflecting the computations underling navigation in going back 150 years to Broca and law’s many intersections with psychology changing real-world environments. Lichtheim, the continued change in focus and behaviour. For example, cases involving between localizing and connectionist claims of are common in the US legal Simon B. Eickhoff. Cognitive and systems perspectives is intriguing. The former system, with hundreds of billions of dollars neuroscience has gone through a period of became massively popular with the ability in awards at stake each year. Pain is, of explosive growth over the past two or three of task-based functional MRI (fMRI) to course, invisible, and there are obvious decades following the advent, expansion discern differences in activity between motivations for malingering. Brain imaging and maturation of in vivo neuroimaging. conditions, whereas the latter surged with research on pain has revealed potential Among the many noteworthy developments the development of methods for studying biomarkers, which may eventually prove over that period, the increase in sample size structural and functional connectivity useful in such cases (indeed, they have certainly stands out. In the late 1990s, MRI in vivo. More recently, novel methods already been used, but this was undoubtedly studies involving fewer than 10 individuals for in vivo brain cartography at the level of premature). Neuroeducation has evolved were the norm. However, this number groups or individuals on the one hand, and beyond vague, in-principle statements has grown continuously and nowadays increasing focus on network-theory and about and learning to specific new studies with several dozen if not hundreds dynamical system modelling on the other, insights about how children learn to read of participants are common. Moreover, have further advanced both perspectives. and do maths, and has delineated different in particular for task-free data, analyses What is missing, though, and hence types of learning difficulties and ways to of 10,000s of individuals have become remains a key challenge for the future, are individualize instruction based on these feasible due to increased data-sharing and unified accounts for regional segregation types. is based on the large-scale efforts such as the UK Biobank. and system-wide interactions in the brain. idea that people often cannot or will not Although these developments could be How to define the building blocks of the articulate their true preferences, whether rightfully attributed to increasing awareness human brain and determine their likely for product packaging or political speeches, about power and replicability, another main context-dependent integration into larger but that neural correlates of emotion and driver is the evolution of study objectives. systems, their inter-individual variability and attention offer a new window on their Earlier studies often focused on robust their relationship to behaviour continue to inclinations. The academic literature backs effects that are often well discernable even be grand but crucial challenges. up many of these claims. Companies around in individuals, but paradigms became more the world use neuromarketing, and political complex and differences between conditions Martha J. Farah. For most of the long campaigns in a number of counties are became increasingly subtle, highlighting history of neuroscience, the field has been reported to have consulted neuromarketers. the aforementioned concerns. Moreover, defined by two main goals: to understand Has neuroscience revolutionized the although the search for the ‘neural correlates the brain and to alleviate neurological and way we teach, do business or pursue justice? of …’ in group analyses has long dominated psychiatric disorders. These two goals have Of course not, but that is an unreasonably the field, more recently, inter-individual recently been joined by a third: to enhance high bar. Has it contributed to progress in differences have come more into focus. diverse non-medical endeavours. New these previously unrelated fields? My answer

Nature Reviews | NEURoSCiEnCE Viewpoint

is a cautious, provisional ‘yes’. The long-term the behaviour of individual synapses understanding internally generated, as well impact of neuroscience on these and other constituting the network. For example, how as stimulus-guided, actions. For example, fields remains to be seen, but what else might synapse-specific Hebbian plasticity when stimulus evidence is ambiguous, would we expect after scarcely 20 years? drive memory engrams that are represented internal motor noise can still drive a The history of neurology and psychiatry by cellular networks? Furthermore, given decision variable over the action threshold. shows that applying neuroscience to our increasing understanding of Hebbian A causal role of biological noise in voluntary complex problems is slow and hard. Patience plasticity mechanisms, how do neighbouring action remains difficult to reconcile with and an open mind are necessary, if not synapses interact and balance distinct our subjective feelings of deciding and sufficient, for progress. So, for the time forms of synaptic plasticity, in some cases controlling. However, being clear about the being, let us resist rolling our eyes. involving communication to the nucleus, role of noise can at least help to distinguish to impact dendritic computation? Also, in the putative signals — internal drives, goals Yukiko Goda. Neuroscience research contrast to spine plasticity, little is known and dispositions — that underlie voluntary over the past two decades has been of the contribution of presynaptic plasticity action. catapulted by technological advances to particular circuit functions and related A second area of progress comes from enabling manipulation and monitoring behaviours. reinforcement learning (RL). Action choices of neural activity in behaving mice at Further cell biological insights into the are often based on learning from experience ever-higher spatial and temporal resolution workings of synapses are also warranted. what worked previously in similar over increasingly broad brain areas. For example, super-resolution microscopy contexts. The past two decades have seen Such developments have dramatically has provided a glimpse into the adhesive increasing focus on RL in human action. stimulated circuit-based analysis of the organization of a synapse where presynaptic ‘Model-based’ RL can be viewed as an neural systems underlying cognitive release sites are aligned to nanoclusters of explicit account of how thinking translates functions and behaviour. Omics approaches postsynaptic scaffolds. How plastic are such into acting — perhaps the fundamental — transcriptomics, proteomics and nano-columns and how are they regulated challenge of voluntary action. Exploratory — are also providing novel to guide transmitter release? Moreover, or exploitative choices in action–outcome views into the molecular and cellular besides the presynaptic and the postsynaptic learning may correspond to the innovative architecture of the brain, revealing an mechanisms, a better understanding or creative aspect of human voluntary extraordinary degree of complexity of of the contribution of glial signalling and action, and to more routine and habitual individual cell types and their subcellular the extracellular matrix that occupy the aspects, respectively. The significance of RL compartments across different brain areas interstitial space would be crucial, as are lies in emphasizing that humans learn to and brain states. the roles of neuropeptidergic signalling. be voluntary. The movements of newborn In synapse biology, the past two decades Finally, the vast number of synapses in the infants seem purposeless, yet healthy have witnessed substantial insights into brain are highly heterogeneous, even in adults learn to control highly complex the molecular underpinnings of synapse the same . Deciphering how synapse movements, so as achieve desired outcomes. organization and plasticity. We have learned, diversity arises in the context of specific Interestingly, childhood amnesia means that for example, how synapse specificity is circuit functions might shed new insights we generally cannot remember the period driven by a combinatorial synapse adhesion into synaptic information processing in of our lives where we lacked voluntary code, that a neuron can pack different the brain and may help in tackling diseases agency. Recent learning studies suggest that transmitters into single synaptic vesicles, linked to synapse dysfunction, of which experienced linkage between action and sometimes even both glutamate and GABA, there are many and with mechanisms yet outcome is boosted immediately after an and that the principal mediators of synaptic to be clarified. error, suggesting an important connection transmission can dynamically move beyond between sense of agency and persistence in the boundaries of individual synapses. Local Patrick Haggard. Since classical times, trying. protein translation, whose occurrence in scientists have noted how human actions Voluntary action lies at the nexus of the dendrites is now well recognized, may also on the one hand seem to have a degree of stimulus–response chain and the action– occur in presynaptic terminals. Collectively, unpredictability, and on the other hand outcome chain. The past two decades have the findings underscore the broad are accompanied, at least sometimes, seen substantial progress in both areas, often plasticity repertoire possessed by synapses. by a subjective feeling of being open, driven by the combination of computational Dendritic spine plasticity during learning or ‘up to me’. These observations often models and mechanistic studies with has continued to attract attention over the lead to theories of human action being animals. The space that volition must occupy past two decades. The causal relationship somehow exceptional, or escaping from is therefore more clearly defined. However, between learning and spine structural normal mechanistic causation. Advances in we still know little about how people select changes has helped establish the hypothesis theory and methods over the past 20 years and maintain their goals. For this reason, originally proposed by Donald Hebb have shown that several key aspects of recent neuroscientific developments have that synapses are the building blocks of human voluntariness can be explained not undermined concepts of individual memory. neuroscientifically, pushing back against psychological autonomy, but rather Excitatory principal neurons typically exceptionalist theories. described how such autonomy may arise, carry tens of thousands of synapses, each Important progress has come from how it is acquired and how it is deployed. capable of expressing various forms of studies on decision-making. The The tension between what neuroscience plasticity. A key issue is to close the widening introduction of formal decision theory teaches us about our volition and what we gap between our circuit-level understanding into perceptual neuroscience, notably are taught by cultural traditions and by our of various neural systems, which is gaining through drift diffusion models, turned own subjective experience remains to be momentum, and our understanding of out to be surprisingly informative in resolved.

www.nature.com/nrn Viewpoint

Hailan Hu. One area I have seen booming is research on mechanisms of action, but from the Human Connectome Project, psychiatric neuroscience, the research into ‘straight roads’ towards the core brain which is mapping structural and functional psychiatric disorders. Unlike neurological mechanisms of psychiatric disorders. Why neural connections, will significantly impact disorders, which are often associated with so? Psychiatric disorders are well known for our understanding of neuropsychiatric discrete anatomical lesions, psychiatric their slow progression, complex symptoms disorders. One other notable discovery is disorders used to be thought of as diseases and multiplex genetic and epigenetic factors, that classic monoamine , of unknown aetiology. Indebted to many making it a daunting task to understand such as serotonin and dopamine, can new mechanistic discoveries, the paradigm the underlying aetiologies. However, some directly attach to chromatin, moving the is shifting and the boundary between rapid-acting drugs quickly alleviate or field to consider these neurotransmitters neurology and psychiatry is becoming create a mentally disordered state, providing not only as cell-to-cell receptor mediators less distinct. This new paradigm is led a convenient and clinically relevant of intracellular signalling but also as direct by revolutionary tools for brain imaging perturbation tool to unlock the hidden regulators of gene transcription. and manipulations, which permit direct doors of psychiatric diseases. For example, Several of the interesting big-picture testing, with unprecedented temporal and as antagonists of the NMDA receptor, questions relate to understanding the spatial precision, of the causal relationship phencyclidine can rapidly induce symptoms long-term consequences of early-life between the activity of specific neural of schizophrenia whereas ketamine can events for neuropsychiatric vulnerability. circuits and behavioural or disease states. rapidly alleviate symptoms of depression. Thus, important questions relate to how In particular, aided by optogenetics or Their actions imply that NMDA receptors we conduct comprehensive longitudinal chemeogenetics, researchers can now mimic are key players and strong target candidates assessments (neuroimaging, biospecimens or mitigate aspects of different psychiatric for these disorders. This knowledge, and behavioural) of individuals — from fetal states, including obsessive–compulsive combined with brain-circuit-specific development, childhood and adolescence disorder, addiction and depression, in information acquired through localized — and how we identify early markers of risk animal models, by activating or inhibiting drug application and circuit manipulation, including epigenetic mechanisms. Some of relevant neural circuits. By targeting these should shed light on the diagnosis, the big-picture questions directly impact neural circuits, researchers can rigorously prevention and treatment of mental health governmental policies and society, such tackle the aetiology of psychiatric disorders, disorders. as what are the potential adverse neural characterizing neural activity patterns In terms of lines of research, I am effects of cannabis and cannabinoids on and pinpointing molecular and cellular personally most excited about how social mental health, and what are their potential abnormalities in relevant brain areas. dominance shapes different behavioural therapeutic properties? Moreover, the recent Another interesting and exciting traits, and what special pharmaceutical opioid epidemic has also raised big questions development in recent decades is the rise features enable ketamine’s rapid and of how to develop novel treatments for of , which explores sustained antidepressant effects. chronic pain conditions that are devoid of the neurobiological mechanisms of social addictive properties. behaviour. As social behaviour involves Yasmin L. Hurd. Each decade has normally In terms of lines of research, I am dynamic interactions between more than brought revolutionary advances to the particularly excited about the growing use one animal or human, there are inevitably neuroscience field, and the past two decades of machine-learning strategies that leverage more variables that affect behavioural have been exceptional. A few advances that neural-derived omics and other complex big outputs. Owing to this complexity, studies in instantly resonate with me include those data sets to improve prediction models about this domain used to be conducted more at that have continued to address a major neurobiological substrates linked to specific the correlative level or focused on hormonal challenge in neuroscience, that being phenotypes. The potential for enhanced or genetic traits. Now, ecologically relevant, understanding the heterogeneous mosaic precision treatment of neuropsychiatric well-controlled behavioural paradigms of brain cells. Recent single-cell sequencing disorders is also particularly exciting with in laboratory settings are being created, strategies and genetic tools have revealed vast the continued development of optogenetic, and new technologies for behavioural differences in the molecular profiles of cells, DREADSS, CRISPR and nanotechnology monitoring, recognition and manipulation indicating the existence of potentially over tools, which have the potential to be are being implemented. These advances a hundred different mammalian neuronal applied in humans, to target cell-specific will enable more direct causality studies at and non-neuronal cell types. Moreover, and circuit-specific neural processes. multiple levels, including the molecular, major technological advances combined In addition, the new research on histone cellular and neural circuitry levels. Most information about the genetic phenotype monoaminylation is exciting as it opens excitingly, some studies have pioneered of cells with optics, such as optogenetics, up new epigenetic processes to target tracking behavioural patterns, physiological and enabled fine-tuned control of neuronal therapeutically. Last, I continue to be excited parameters or even neural activities of activity to elucidate the functional role of about efforts to expand molecular, structural multiple animals simultaneously. Such individual cells. From such advances, we and functional knowledge about the human advancements will be particularly insightful have gained better insights into the role of brain. Such knowledge is essential to for studying social interactions in complex distinct cells localized within discrete neural improve the translation of animal studies and naturalistic environments. circuits to directly modulate behaviours and to improve treatment strategies. Overall, In my view, one interesting, big-picture relevant to, for example, decision-making, my excitement stems from research efforts question in psychiatric neuroscience is reward, fear and negative affect. Large that provide in-depth neurobiological how do some drugs rapidly ameliorate steps have also been taken in acquiring knowledge about neuropsychiatric disorders or induce certain psychiatric symptoms? knowledge about the transcriptome and that is directly relevant to the human Studies related to this question should not epigenetic landscape of the human brain. condition and that can provide critically be considered as standard pharmacological This knowledge, taken together with data needed new therapeutic interventions.

Nature Reviews | NEURoSCiEnCE Viewpoint

Sheena A. Josselyn. Understanding how remembered distinctly or organized into In relation to neurons, glia are still an we learn and remember has fascinated boarder conceptual knowledge), how the understudied cell population. I think the philosophers and scientists for hundreds of passage of time or new information changes most interesting questions being asked in years. In the past 20 years, there has been engrams, the extent to which engrams are my area relate to how glia regulate neural a seismic shift in how we (the field) study static or dynamic and how information circuits, how glia contribute to behaviour memory. This methodological tremor becomes momentarily inaccessible (as in and how glia contribute to the function of was triggered by our newfound ability to the tip-of-the-tongue phenomenon) or the brain as a highly evolved multicellular gain a firm experimental grip on the basic irretrievably forgotten (more permanent organ and not just as a collection of neuronal cellular unit of memory storage, the engram amnesia). circuits. Including glia into how complex (or memory trace). The general idea that That being said, many many questions functions of the brain are performed also a specific memory is stored as enduring remain. For instance, exactly how (and holds the promise of shedding new light on changes in the brain is centuries old, but it where) is information stored in the brain? disease mechanisms as well as brain–body was not until the beginning of the twentieth Is it stored in cell ensembles, synapses, and neuroimmune interactions. century that Richard Semon introduced molecules or DNA? And, more basically, I am particularly excited about the the term ‘engram’ to describe the neural what do we even mean by ‘information’? prospect of manipulating and targeting glia substrate for storing and recalling memories. The answer to these ‘hard’ memory as endogenous neuromodulators to produce An experience, Semon essentially proposed, questions will likely require continued disease-related phenotypic benefits in activates a small subset of cells that undergo crosstalk between researchers examining complex brain disorders. This is a worthwhile persistent chemical and/or physical changes memory in humans, rodents, fish, flies and direction to follow: pursuing neuronal to become an engram. Memory retrieval even in silico. mechanisms alone over the past 20 years has occurs when sensory cues later reactivate not been as fruitful as was once anticipated this engram. Although the idea of an engram Baljit S. Khakh. Twenty years ago, in terms of disease-modifying treatments. may have been theoretically appealing, neuroscience consisted largely of a Additional parallel and orthogonal similar to hunting a mythical creature, as cottage industry of researchers pursuing approaches are now also necessary. soon as scientists tried to experimentally open-ended questions for which they At a broader level, I think mechanistic grab an engram, it seemed to slip away. had amassed a great deal of specialized studies and computational neuroscience Over the past 20 years, though, the knowledge and skills. Since then, we have are going to be critical to make sense development of an ever-increasing array of witnessed the dawn of ‘industrial scale of the vast amount of precisely tracked, new tools allowed scientist not only to find but neuroscience’, pursued as large organized technique-driven, but essentially descriptive also to manipulate ‘engram cells’ across several initiatives, often in private institutes. These behavioural data that are emerging in the model species. The ability to catch an engram types of projects seek to tackle big problems field as a whole. proved to be an absolute game-changer. For in defined periods. They are having an instance, previously we attempted to examine inexorable and substantial impact in the Jürgen A. Knoblich. Like other fields, the biochemical and electrophysiological details of the science itself and, increasingly, neuroscience was profoundly influenced substrates of memory by grinding up an in how others plan and do their own science. by the arrival of the post-genome era. The entire brain region or ‘listening’ to the Pundits, academic institutions and funding ability to compile complete inventories electrophysiological changes of a broad agencies are also pondering how they can of receptors, channels and other elements of population of neurons (as in field long-term align faculty to attack key problems at a scale neuronal information processing, together potentiation recordings) after a learning event. that is not possible in single research groups. with emerging methods for measuring With engram technology, changes in engram The promise of such approaches is that their expression and function across cell cells alone could be studied, free from the progress could be rapid and efficient, but types in a comprehensive manner, have potentially obscuring effects in non-engram the potential risk is that laboratory research, brought neuroscience to another level. cells. The mnemonic wheat could finally be unless thoughtfully shepherded, may lose A second groundbreaking development separated from the chaff. some of the creativity that has and continues was the establishment of optogenetics. The With engram technology, we can now to drive discovery and innovation in small, ability to turn neurons on or off has merged implant ‘artificial memories’ in mice. hands-on research groups. Undoubtedly, functional neuroscience with cell biology. Indeed, a handful of laboratories’ recent both models deliver in distinct ways. Mapping brain circuits and connecting them studies showed it is possible to create In my own field of glial biology, to specific behaviours has become possible and entirely artificial percepts or memories in the availability of new genetic, optical will fill the gap between neuroanatomy mice, without any external stimuli, simply and instrumentation tools has allowed and the functional analysis of individual by manipulating cell ensembles. These neuroscience to be explored in ways that neurons and their electrical properties. science-fiction-type findings verify that we were previously not possible, and this In neurodevelopment, my own field are on the right track. In addition, studying has changed the field to the extent that of research, the past two decades saw the memory at the level of engrams is leading even cellular–molecular projects can now, identification of radial glia cells and their to novel insights into the very fundamental for example, evaluate the behavioural various descendants as the progenitors for building blocks of memory. For instance, implications of mechanistic findings. all neurons in the cortex. Importantly, some we are beginning to understand how cells Another development has been the growth subtypes of these progenitor cells are unique (initially excitatory neurons, and now of big data — including genomic, gene to primates and this has triggered a new era expanding to include other cell types) and protein expression, electrophysiology, in which analysing unique features of the are allocated to an engram, how the imaging and behavioural data — and the human brain is within reach. architecture of an engram impacts memory necessity to process and analyse them with I am convinced that the next decade in retrieval (for instance, whether events are computational methods. neuroscience will be the era of the human

www.nature.com/nrn Viewpoint

brain. It is increasingly clear that human but as powerful computational units that computations previously attributed to brains are not just enlarged versions of greatly expand the processing and storage networks. Finally, I believe that adopting animal brains. They have unique features capacity of individual neurons. dendritic properties in artificial and develop in a unique way, and we are Today, the technological advances of (AI) holds great promise. Incorporation at the verge of understanding why that is. the past decades allow the testing of these of dendritic processing in both software We have identified human-specific cell propositions through causal probing of (for example, deep neural networks) and types and are beginning to understand multiscale processes that operate in the hardware (for example, CMOS devices) their role in making our brains so powerful. brains of behaving animals. New tools technologies is underway, hoping to achieve Genome-editing, functional genomics and allow the tracking of synaptic activity and brain-like performance in complex tasks organoid models can be combined to allow plasticity in vivo (for example, SynTagMA such as reasoning and one-shot learning. genetic analysis in human brain tissue and GluSnFR), the characterization of These efforts are expected to open an models. inputs to individual neurons (for example, entirely new era in bio-inspired machine Eventually, those tools will endow us with via rabies viruses) and the selective learning and AI in the following years. the ability to ask fundamentally important manipulation of activity (optogenetics), questions for understanding ourselves: often restricted to specific compartments Russell A. Poldrack. In my view, what makes us human and what genetic located far from the surface (for example, one notable development in cognitive changes are responsible for the enormous via prisms), while recording from hundreds neuroscience over the past two decades complexity and cognitive power of the of neurons across different brain areas. has been the advent of computational human brain? What makes us individual? Voltage indicators are also coming of theories that are being used to both Why does my son think like me and where age and will soon replace slower molecules, drive experimentation and interpret is this encoded in the genome? And why do and computational modelling, driven by neuroimaging data. This has occurred humans with alterations in specific genes this slew of multimodal data, has advanced across many subfields, most notably think differently and develop symptoms to account for dendritic processing at the in visual object recognition, but also in we call autism or schizophrenia? Besides single neuron and circuit levels. These content areas as diverse as decision-making, providing fundamental biological insights, new tools enable unprecedented access memory and motor control. Together the answers to those questions will also to the microscopic, mesoscopic and with new methods for modelling and the profoundly change the medical treatment macroscopic levels of computation and analysis of neuroimaging data, these models of neurological and psychiatric disorders. promise exciting new discoveries. The have provided a tool to more clearly link Needless to say, I am particularly excited past decade has also witnessed a closer functional imaging signals to underlying about organoids. It may be surprising to and more fruitful collaboration between computational mechanisms. some that I do not think that they will theorists and experimentalists. Numerous Another important development replace animal experiments or that they dendritic contributions, many of which were has been a transition from the highly are or will ever become a comprehensive originally predicted by models, have been localizationist approaches of early model for the human brain — but they now confirmed in awake behaving animals neuroimaging research to a focus on the allow us to take the last step towards asking for critical brain functions such as learning hierarchical structure and dynamics of brain specific questions about our own brain. and memory, sensation, navigation and even networks. Much of this focus has arisen Currently in their infancy, I am confident perception and anaesthesia. from task-free fMRI studies, which have that organoids will develop into models Yet most of our current understanding is demonstrated structured correlations with for investigating certain neurobiological limited to dendritic effects on single-neuron clear functional relevance at multiple levels processes. Establishment of organoid models processing or circuit activity. Establishing of spatial organization. In addition, we have for functional brain circuits and for major causal links between dendritic function seen the advent of ‘network neuroscience’, long-range connections within the brain and behaviour is the ultimate challenge and, leveraging the tools of network science to are within reach and can be combined with to me, the most exciting direction for mathematically characterize brain networks. optogenetics to probe human-specific brain the years to come. A synergistic, online It might once have been legitimate to refer features at the circuit level. I am also excited approach will be critical for achieving to neuroimaging as ‘blobology’, but in most about the power of combining in vitro this goal. For example, monitoring of areas of neuroimaging research those days models with ever-improving fMRI brain synaptic, dendritic and neuronal activity are largely gone. imaging methods to understand human from different cell types and brain areas For me, one of the most interesting brain circuits. during various behaviours could be used to questions is how we can reconcile the constrain large-scale network models in a views of brain function that are currently Panayiota Poirazi. The turn of the continuous, online manner. The models will inherent in the computational neuroscience twenty-first century was marked by predict how changes in dendritic properties and network neuroscience approaches. remarkable discoveries regarding the may alter aspects of behaviour and online Network neuroscience has provided cellular and subcellular properties of tests will refute or verify predictions, thus important insights into the structure neurons. As a dendrites’ aficionado, closing the loop and updating both models and dynamics of brain networks, but the I was thrilled by the unveiling of their rich and hypotheses. I am also fascinated by the methods used for these analyses generally conductance repertoire, and their ability to quest for understanding dendritic functions treat ‘nodes’ as fungible beyond their specific compartmentalize electrical, chemical and across different cell types and species. location within the network. By contrast, molecular signals and support local spiking. Exciting new discoveries suggest that computational neuroscience has provided Empowered by computational modelling, interneuronal dendrites can also integrate insights into the specific computational these discoveries suggested that dendrites incoming signals in non-linear ways and that functions that are performed by individual should no longer be viewed as passive cables human dendrites can implement advanced brain circuits, but has had relatively

Nature Reviews | NEURoSCiEnCE Viewpoint

little to say about the complex dynamics microglia discoveries in the following years. Pieter R. Roelfsema. Advances in that occur when many such circuits are For example, the unexpected finding that technology have driven progress in systems interconnected at the scale of an entire brain. ‘resting’ microglia are continuously active neuroscience in important ways over Bridging these two approaches will require in the living mouse brain shed new light on the past 20 years. New methods became a rapprochement between very different microglial functions in the normal brain available to measure and control the activity analytical approaches coming from physics and made clear that these cells are highly of many neurons with high precision. and machine learning, but it will be essential active guardians of brain homeostasis. Another noteworthy development is to developing a fuller understanding of how Along this line, microglial cells were the increasing influence of AI, allowing the brain gives rise to high-level cognitive subsequently identified as key CNS intrinsic neuroscientists to better understand synaptic functions. architects that shape neuronal circuits plasticity rules and information processing I am particularly excited about recent during development and homeostasis by in the brain. work that has begun to define the functional a process called synaptic pruning that was A major change in visual neuroscience architecture of individual human brains later similarly found to be present during has been the focus on the mouse as a model with much higher precision than before. pathological processes, such as Alzheimer to study vision. Many researchers would not After my initial work in the MyConnectome disease. have predicted this in 2000, because mice study, a number of laboratories have begun The most burning questions in do not have a particularly well-developed using a ‘deep phenotyping’ approach in dealing with the innate visual system. The disadvantage of which individuals are scanned repeatedly immune arm are how microglia contribute studying a species with poor visual acuity and characterized individually. This to CNS diseases and how this contribution is offset by the many powerful tools now approach has already provided new insights can be modulated. Big international available for mice, including optogenetics, into the fine-grained organization of the genome-wide association studies in recent two-photon imaging, tracing techniques entire human brain, highlighting both years have identified an ever-growing and interventions using viruses and the universality and diversity of functional number of genetic risk factors associated transgenic mouse lines. Even good-old organization at different levels, as well as with several disorders — such as multiple electrophysiology has been improving. identifying new functional features that were sclerosis, Alzheimer disease, Parkinson Researchers used to make recordings from not visible in earlier group-averaged fMRI disease and others — that are expressed one cell at a time, but new probes permit studies. Going forward, I am particularly by CNS myeloid cells, like microglia. The the study of many hundreds of neurons excited to see these approaches taken into major challenge now is to understand simultaneously. Hence, recent papers the study of mental illness. Given the great how intergenic variations of several loci report the activity of tens of thousands of deal of variability in the time course of jointly lead to changes in transcription neurons, distributed across many cortical mental illness, understanding the dynamics or translation and, ultimately, contribute and subcortical regions during a single of brain function in relation to the to disease. Hopefully, rare monogenetic behavioural task. For the development of dynamics of disease is crucial to enable disorders primarily and solely affecting local the field, it will be important to make these mechanistic insights that can lead to new tissue macrophages (for example, disorders novel techniques available to use in species, treatments. caused by CSF1R mutations) will help to including non-human primates, with brains decipher the underlying mechanisms of that more closely resemble human brains. Marco Prinz. In the past two decades, we normal and disturbed immune regulation An important trend has been the have been eyewitnesses to a renaissance in the CNS and might finally facilitate the move towards open science and open data, of neuroimmunology research within future development of myeloid-focused providing researchers access to large neuroscience. In particular, a special focus therapeutic approaches. data sets. This leads to a more efficient has been made on understanding the There are several exciting novel use of resources and will increase the function and heterogeneity of CNS myeloid technological developments that currently reproducibility of results by making it easier cells including parenchymal macrophages electrify neuroimmunologists around the to evaluate the validity of the conclusions of (that is, microglia) and macrophages globe. A new era of microglial research has an article. By itself, access to large amounts at CNS interfaces, such as perivascular, just begun. The urgent need to understand of data is not enough to understand the meningeal and choroid plexus macrophages. microglial biology led, for example, to the brain, however. Fortunately, neuroscientific Conversely, when I started my research development of microglia-like cells from insight has started to benefit from the on microglia in the late 1990s at the Max embryonic stem cells or induced pluripotent progress in AI. In vision, artificial neural Delbrück Center for Molecular Medicine in stem cells. Further thrilling techniques in the networks are able to transform raw sensory Berlin, neuroscience was almost exclusively field include brain organoids, assembloids input into object categories; that is, they neuron-centred and glial researchers were and xenotransplantations that will hopefully extract meaning. Importantly, the tuning considered as rather exotic foreigners, permit the study of environmental and of units of these artificial neural networks neither belonging to neuroscientists nor to genetic factors in microglia-mediated resembles the tuning of neurons in the immunologists. CNS diseases. In the future, multi-omics brain, helping neuroscientists to understand The big bang of microglia research approaches to profile genetic, epigenetic, how the many areas of the brain’s own was certainly the development of the proteomic and metabolomic cellular ‘deep networks’ contribute to the step-wise CX3CR1GFP mouse line by Steffen Jung states at the single-cell level, novel animal transformation of pixels into semantic in 2000 when he was a postdoc in Dan R. models, imaging and fate-mapping tools, codes. Furthermore, AI is now also inspiring Littman’s laboratory in New York. This patient-derived microglial cells and other insights into how connections in deep brain pioneering work allowed for the first innovative techniques will help us to networks can be trained efficiently. time the visualization of microglia in vivo decipher more microglial secrets around The enhanced understanding and and paved the way for several important CNS development, homeostasis and disease. accessibility of neural codes through AI also

www.nature.com/nrn Viewpoint

opens up new possibilities for interfacing researchers face many difficulties in suggest the widespread expression of signals, with the brain. A recent study used AI to pursuing an academic career. Although especially those related to action and reward, decode language from brain activity of much more work is needed, groups such throughout the brain. These studies point to humans. This methodology can be applied in as the FENS-Kavli Network of Excellence the importance of disentangling correlative the future to help people with severe forms are striving to support people coming up in versus causal signals, and the roles of specific of paralysis in their communication with the field and to enhance the diversity and yet dynamic local and long-range circuits in the outside world. In combination with new inclusivity of neuroscience. In my view, we behaviour. high channel count interfaces, researchers also collaborate more in neuroscience than High-resolution imaging of synapses may soon read and write to the brain with we did 20 years ago, with collaborations across and synaptic molecules has led to a richer unprecedented precision, opening new disciplines and industry, and through large conceptual understanding of cortical avenues for therapeutic intervention. Indeed, strategic initiatives like the UK Dementia plasticity. At excitatory synapses, plasticity approaches are underway to create new Research Institute, the US Brain Initiative and is largely implemented by Hebbian learning prostheses for blind individuals by directly the EU . The impending rules, whereby coactive synapses are writing information captured by a camera to global health crisis that dementia poses and strengthened; yet unregulated Hebbian the visual cortex, skipping the malfunctioning the current also highlight the strengthening leads to saturated synapses eyes. As soon as these high-density brain continued need for scientists to engage with and unstable cell assemblies. Thus, synaptic reading and writing are applied in humans, the public and policy-makers to help develop renormalization is necessary to stably encode there are likely to cause a new wave of evidence-based policies to meet societal information in neurons and networks. scientific insights into the neuronal processes challenges. A key advance is the realization that underlying the human mind. For the next 20 years, I am excited about neuronal plasticity is implemented translation. Perhaps I am a hopeless optimist by multiple cooperative mechanisms: Tara L. Spires-Jones. Exciting new lines in the face of decades of clinical trial failures Hebbian changes at specific synapses are of investigation have emerged over the for Alzheimer disease, but I think that we complemented by locally coordinated past two decades in the field of Alzheimer are close to life-changing treatments. It is plasticity at adjacent synapses that disease research, including understanding extremely gratifying and humbling that renormalizes synaptic weights over dendritic how pathological proteins damage the brain, our work increasing understanding of stretches, and by longer-term synaptic why certain neurons and brain regions are the fundamental brain changes that cause scaling which further renormalizes weights vulnerable to degeneration, bridging the dementia symptoms is contributing, albeit over the entire postsynaptic neuron. gap between genetic and epidemiologic risk in small part, to the development of clinical These ideas, derived from measurements factors and the brain changes that cause approaches and trials. I am also excited in cortical neurons, likely apply to plasticity dementia, and understanding how glia about the advancing understanding of the of most neurons in the brain during contribute to neurodegeneration. fundamentals of the brain and what makes us development and learned behaviours in The field of dementia research has been human, our ability to think, learn, remember. adulthood. dramatically advanced by technology over I agree with the European Brain Council’s Plasticity is integrally related to cortical this period. For example, although the ‘Brain Mission’, which places understanding dynamics and computations. A host of pathological lesions that define Alzheimer the brain and treating brain disorders at the molecular and cellular processes are engaged disease were described in the early 1900s of a twenty-first-century mission by activity to modify cortical synapses and the ability to study pathology with analogous to the twentieth-century missions and circuits. For example, the induction histology was well established over 30 years to put humans on the Moon. of plasticity in the visual cortex requires a ago, the past 20 years has seen a boom in critical level of inhibition. Plasticity is gated the ability to image these pathologies Mriganka Sur. Two complementary fields by modulators of attention and arousal, in living people over time with positron have witnessed notable developments over such as acetylcholine and norepinephrine, emission tomography, to study the the past two decades: understanding cortical which regulate inhibitory–disinhibitory pathological proteins in exquisite detail with dynamics at the level of brain regions and circuits as well as astrocyte–neuron cryo-electron microscopy and to examine neuronal populations, and cortical plasticity at interactions. The implementation of the toxic oligomeric forms of these proteins the level of synapses and clusters of synapses. plasticity requires signalling molecules that within individual synapses in animal models Recording the activity of large numbers mediate cooperative synaptic potentiation and human tissue. GWAS starting in the of neurons, via multiphoton imaging and and depression, and molecules that regulate early 2000s highlighted new pathways high-density probes, has given us novel synaptic scaling. How might particular involved in disease pathogenesis, which are views of brain-wide activity, at single-neuron mechanisms relate to specific circuits and being followed up using new technologies resolution, in at least small brains. tasks? like single-cell transcriptomics and CRISPR Manipulating regional and cell-specific Understanding cortical plasticity and gene editing. activity with optogenetics has revealed dynamics has two crucial applications. Some of the biggest changes in the specific contributions of brain regions and First, it provides the basis for treating brain field over this time frame have not been cell classes to information processing. disorders. Indeed, we are increasingly seeing neuroscience-specific. The explosion of the Applying these tools to mice performing imaginative proposals for treating disorders ability to share data has been of real benefit to behavioural tasks, we have derived a of brain development and ageing that neuroscientists, and this is likely to become computational understanding, building on derive, importantly, from basic mechanisms even more important over the next 20 years. previous studies in non-human primates, of of plasticity and dynamics. Second, it can We have also seen major changes in regional circuit contributions to decisions, help build efficient computing machines. the way we conduct neuroscience. There short-term memory and actions. At the Unlike computers with von Neumann is an emerging awareness that early-career same time, recent findings puzzlingly architecture where processors and memory

Nature Reviews | NEURoSCiEnCE Viewpoint

are separate, the confluence of plasticity and Regarding lines of research, I am 15Department of Psychiatry of First Affiliated Hospital, Zhejiang University School of Brain Science and Brain processing in the same circuit elements is particularly excited about the trans- Medicine, Zhejiang University School of Medicine, a core component of brain computations. hierarchical studies to investigate the role Hangzhou, China. Cortical plasticity almost certainly mediates of particular residues of specific molecules 16The MOE Frontier Research Center of Brain & Brain– our robust perceptual schemas driven by in specified cells in organisms, which are Machine Integration, Zhejiang Laboratory for Systems few examples, in contrast to large example currently only possible for limited model & Precision Medicine, Hangzhou, China. set-driven, yet brittle, machine perception organisms such as nematodes and fruit flies. 17Department of Psychiatry, Icahn School of Medicine that is the current state of the art. This is because the mysteries of biological at Mount Sinai, Addiction Institute of Mount Sinai, systems often come from unique and New York, NY, USA. Hiroki R. Ueda. Just 20 years ago, we irreplaceable molecules and/or cells that 18Department of Neuroscience, Icahn School of successfully obtained a catalogue of play essential roles in the systems. To fully Medicine at Mount Sinai, Addiction Institute of Mount Sinai, New York, NY, USA. molecules, in the form of a working draft of and deeply understand the uniqueness and 19Program in & Mental Health, Hospital the human genome. Now, we are acquiring irreplaceability of an essential molecule, it for Sick Children, Toronto, Ont., Canada. catalogues of cells in entire brains and even is important to investigate its underlying 20Department of Psychology, Physiology & Institute of entire bodies in some mammals. One of the atomic network. Likewise, to elucidate Medical Sciences, University of Toronto, Toronto, Ont., greatest contributions to these achievements those of a pivotal cell, it is key to study Canada. has come from the remarkable development its underling molecular network. Twenty 21Brain, Mind & Consciousness Program, Canadian and seamless integration of tissue-clearing years from now, by the 40th anniversary of Institute for Advanced Research (CIFAR), Toronto, Ont., methods and light-sheet microscopy. Nature Review Neuroscience, I hope that such Canada. Tissue-clearing methods have become trans-hierarchical identification, analysis, 22Department of Physiology, David Geffen School of increasingly high-performance, faster, safer control and design of atomic, molecular Medicine, University of California Los Angeles, Los Angeles, CA, USA. and easier to use. Light-sheet microscopy has and cellular networks will be possible in 23 become more precise, with higher resolution, higher organisms, which will eventually Institute of Molecular Biotechnology of the Austrian Academy of Sciences (IMBA), Vienna BioCenter (VBC), faster and more isotropic. As a result, it is lead to a wider and deeper understanding of Vienna, Austria. now possible to perform whole-body or brain organismal systems in mammals, primates 24Institute of Molecular Biology and Biotechnology profiling of cells and circuits in mice and and even humans. (IMBB), Foundation of Research and Technology-Hellas other mammals. In fact, a three-dimensional, Danielle S. Bassett 1,2,3,4,5,6 ✉ , (FORTH), Heraklion, Crete, Greece. 7,8,9 ✉ 10,11 ✉ single-cell atlas of the adult mouse brain Kathleen E. Cullen , Simon B. Eickhoff , 25Department of Psychology, Stanford University, 12 ✉ 13 ✉ is already available. In addition to the Martha J. Farah , Yukiko Goda , Stanford, CA, USA. Patrick Haggard 14 ✉ , Hailan Hu 15,16 ✉ , 26Institute of , Medical Faculty, acquired positional information of cells, Yasmin L. Hurd 17,18 ✉ , Sheena A. Josselyn 19,20,21 ✉ , University of Freiburg, Freiburg, Germany. tissue-staining methods and image Baljit S. Khakh 22 ✉ , Jürgen A. Knoblich 23 ✉ , informatics are also rapidly being developed Panayiota Poirazi 24 ✉ , Russell A. Poldrack 25 ✉ , 27Signalling Research Centres BIOSS and CIBSS, to characterize various cells in the brain and Marco Prinz 26,27,28 ✉ , Pieter R. Roelfsema 29,30,31 ✉ , University of Freiburg, Freiburg, Germany. 32 ✉ 33 ✉ body. In the near future, we may expect a Tara L. Spires-Jones , Mriganka Sur and 28Center for Basics in Hiroki R. Ueda 34,35 ✉ colourful single-cell atlas of adult mouse, rat (NeuroModulBasics), Faculty of Medicine, University 1Department of Bioengineering, University of of Freiburg, Freiburg, Germany. and marmoset brains and bodies. Technically Pennsylvania, Philadelphia, PA, USA. 29 speaking, the completion of even the human Department of Vision and Cognition, Netherlands 2Department of Electrical and Systems Engineering, Institute for Neuroscience, Amsterdam, Netherlands. whole-brain atlas may not be a dream. University of Pennsylvania, Philadelphia, PA, USA. 30Department of Integrative , What comes after the catalogues of cells? 3 Department of Physics and Astronomy, University VU University, Amsterdam, Netherlands. The catalogue of molecules was followed of Pennsylvania, Philadelphia, PA, USA. 31Department of Psychiatry, Academic Medical Centre, 4 by the emergence of a new biology field Department of Neurology, University of Pennsylvania, Amsterdam, Netherlands. called systems biology. Systems biology at Philadelphia, PA, USA. 32Centre for Discovery Brain Sciences and UK 5Department Psychiatry, University of Pennsylvania, cellular to organismal levels has been hard Dementia Research Institute, University of Edinburgh, Philadelphia, PA, USA. to realize until recently, because it first Edinburgh, UK. 6Santa Fe Institute, Santa Fe, NM, USA. requires the identification of the cellular 33Picower Institute for Learning and Memory, 7 networks where a catalogue of cells plays a Department of Biomedical Engineering, Johns Hopkins Department of Brain and Cognitive Sciences, fundamental role. Now that the catalogues University, Baltimore, MD, USA. Massachusetts Institute of Technology, Cambridge, 8 of cells are being built and whole-brain cell Department of Otolaryngology — Head and Neck MA, USA. Surgery, Johns Hopkins University, Baltimore, MD, USA. profiling technology is available, we may 34Department of Systems Pharmacology, 9Department of Neuroscience, Johns Hopkins The University of Tokyo, Tokyo, Japan. be able to identify these cellular networks, University, Baltimore, MD, USA. 35Laboratory for Synthetic Biology, RIKEN BDR, Suita, if the aforementioned approaches are 10 Institute of , Medical Faculty, Osaka, Japan. combined with perturbational technology, Heinrich Heine University Düsseldorf, Düsseldorf, ✉e-mail: [email protected]; kathleen.cullen@ Germany. such as next-generation genetics (genetics jhu.edu; [email protected]; mfarah@ 11 without mating) and the systemic delivery Institute of Neuroscience and Medicine, Brain & psych.upenn.edu; [email protected]; p.haggard@ of adeno-associated virus. Both of these Behaviour (INM-7), Research Centre Jülich, Jülich, ucl.ac.uk; [email protected]; yasmin.hurd@ technologies do not depend on mating, and Germany. mssm.edu; [email protected]; bkhakh@ 12 hence can perturb the molecules and cells Center for Neuroscience & Society, University of mednet.ucla.edu; [email protected]; Pennsylvania, Philadelphia, PA, USA. [email protected]; [email protected]; of interest to test hypotheses in vivo within 13Laboratory for Synaptic Plasticity and Connectivity, [email protected]; p.roelfsema@ a couple of weeks in mice. Identification RIKEN Center for Brain Science, Wako-shi, Saitama, nin.knaw.nl; [email protected]; msur@ of cellular circuits will further facilitate the Japan. mit.edu; [email protected] analysis, control and even design of cellular 14Institute of Cognitive Neuroscience, University https://doi.org/10.1038/s41583-020-0363-6 networks. College London, London, UK. Published online xx xx xxxx

www.nature.com/nrn Viewpoint

Acknowledgements MH104069), CHDI Inc., an Allen Distinguished Investigator M.S. is supported by NIH grants EY028219, MH085802 D.S.B. is grateful to the mentees, colleagues and mentors who Award through The Paul G. Allen Frontiers Group, a UCLA and DA049005, and the Simons Foundation Autism constantly broaden her appreciation of the diversity of scien- David Geffen School of Medicine Seed Grant and the Ressler Research Initiative. H.R.U. is supported by Brain/MINDS tific inquiry and deepen her excitement to pursue the most Family Foundation. J.A.K. would like to thank all his lab mem- JP21dm0207049, Science and Technology Platform Program puzzling questions. K.E.C. was supported by the National bers for exciting discussions. Work in J.A.K.’s laboratory is for Advanced Biological Medicine JP21am0401011, Institutes of Health (NIH; grant numbers DC002390, supported by the Austrian Federal Ministry of Education, Grant-in-Aid for Scientific Research (S) JP25221004 DC018061 and UF1 NS111695). S.B.E. is supported by the Science and Research, the Austrian Academy of Sciences, the (JSPS KAKENHI) and HFSP Research Grant Program Deutsche Forschungsgemeinschaft (DFG; EI 816/16-1 and EI Austrian Science Fund (Z_153_B09), the City of Vienna and (HFSP RGP0019/2018). 816/21-1) and the European Union’s Horizon 2020 Research the European Research Council (ERC, grants 695642 and Innovation Programme under Grant Agreement No. and 693184). P.P. thanks the members of her laboratory Author contributions 945539 (HBP SGA3) and 826421 (Virtual Brain Cloud). Y.G. (www.dendrites.gr) for their dedication and hard work. Her The authors contributed equally to all aspects of the article. is supported by funds from the RIKEN Center for Brain laboratory is funded by the European Union’s Horizon 2020 Science, The Uehara Memorial Foundation and the Japan research and innovation programme (FET Open RIA GA Competing interests Society for the Promotion of Science (JSPS) Core-to-Core 863245: NEUREKA, MC RISE GA 873178 INAVIGATE), the P.R.R. is a co-founder of and shareholder in a neurotechnol- Program A Advanced Research Networks. P.H. is partly sup- EINSTEIN Foundation-Berlin and the Fondation Santé. M.P. ogy start-up, Phosphoenix (Netherlands). T.L.S.-J. receives ported by a joint grant from the John Templeton Foundation is supported by the Sobek Foundation, the Ernst-Jung collaborative funding from three industry partners (Autifony, and the Fetzer Institute. The opinions expressed in this pub- Foundation, the DFG (SFB 992, SFB1160, SFB/TRR167, Cognition Therapeutics and one anonymous industry donor), lication are those of the author(s) and do not necessarily Reinhart-Koselleck Grant and Gottfried Wilhelm Leibniz Prize) is on the Scientific Advisory Board of Cognition Therapeutics reflect the views of the John Templeton Foundation or the and the Ministry of Science, Research and Arts, Baden- and is an editor for several journals, including founding Editor Fetzer Institute. Furthermore, P.H. was supported by a Chaire Wuerttemberg (Sonderlinie ‘Neuroinflammation’). M.P. is fur- in Chief of Brain Communications. None of these influenced de Recherche Jean D’Alembert held at Paris-Saclay University, ther supported by the DFG under Germany’s Excellence this piece. D.S.B., K.E.C., S.B.E., M.J.F., Y.G., P.H., H.H., Y.L.H, and from a fellowship at the Paris Institute for Advanced Strategy (CIBSS — EXC-2189 — Project ID390939984). S.A.J., B.S.K., J.A.K., P.P., R.A.P., M.P., M.S. and H.R.U. Study (France), with the financial support of the French State, P.R.R. was supported by the Netherlands Organization for declare no competing interests. programme ‘Investissements d’avenir’ managed by the Scientific Research (NWO; grant P15-42 ‘NESTOR’), the Agence Nationale de la Recherche (ANR-11-LABX-0027-01 European Union’s Horizon 2020 Framework Programme Publisher’s note Labex RFIEA+). H.H. is supported by grants from the National Agreement No. 650003 (HBP FPA) and the Friends Springer Nature remains neutral with regard to jurisdictional Foundation of China (31830032 and Foundation of the Netherlands Institute for Neuroscience. claims in published maps and institutional affiliations. 81527901) and the Fountain-Valley Life Sciences Fund of T.L.S.-J. is funded by the European Research Council (ERC) University of Chinese Academy of Sciences Education under the European Union’s Horizon 2020 research and inno- Related links vation programme (Grant agreement No. 681181), the UK Foundation. Y.L.H. is supported by grants DA050323, anneslist: https://anneslist.net/ Dementia Research Institute — which receives its funding DA048613, DA043247 and DA030359 from the National Biaswatchneuro: https://biaswatchneuro.com/ from DRI Ltd, funded by the UK Medical Research Council, Institute of Drug Abuse (NIDA). S.A.J. acknowledges discus- women in neuroscience Repository: https://www.winrepo.org/ sions with P. W. Frankland (SickKids) and members of the Alzheimer’s Society and Alzheimer’s Research UK — and the Josselyn and Frankland laboratories for continued inspira- University of Edinburgh. M.S. thanks the members of his lab- tion. B.S.K. is supported by the NIH (NS111583 and oratory for their comments. Research in the laboratory of © Springer Nature Limited 2020

Nature Reviews | NEURoSCiEnCE