The human brain underlies cognition, emotion, and behavioural control. As such it is necessarily the substrate of all psychiatric disorders—except in the view of philosophical dualists, if any remain. Psychiatry has thus been a beneficiary of progress in neuroscience and other basic fields of inquiry ranging from genetics to psychology that elucidate brain function. Neuroscience, the interdisciplinary field that studies the brain (and nervous systems, more broadly), is enjoying a period of expansion in scope (1) and significant scientific success predicated on remarkable new tools, new discoveries, and new forms of organization that support large-scale collaborations and data sharing (2–4). Nonetheless, psychiatry has lagged behind many other medical fields in its ability to translate basic research findings into understanding disease mechanisms and generating new treatments. Moreover, it can justly be said that psychiatry has reached a scientific crossroads marked by the increasingly obvious limitations of its diagnostic classifications (5,6), and by a decades-long failure to improve upon the efficacy of existing treatments. This state of affairs is not simply an academic matter. Much of the pharmaceutical industry has de-emphasized or terminated work on psychiatric disorders based on longstanding failures to make significant advances (7,8). Treatment development based on devices such as deep brain stimulation (DBS) (9) and on applications of cognitive neuroscience to psychotherapy (10) has moved forwards, but, without understanding disease mechanisms, it is not clear how substantial progress can be accomplished. Clinicians, policy-makers, patients, and families are confronted by vast unmet clinical need for people with such maladies as autism, schizophrenia, bipolar disorder, depression, obsessive-compulsive disorder, and other forms of psychiatric disorder that, in aggregate, are leading contributors to global disease burden (11). The scientific question facing psychiatry is how to revitalize research that will lead to better diagnoses and treatments.
For the class of diseases that go by the inaccurate and anachronistic (because implicitly dualistic) term of mental disorders, there are significant obstacles to understanding pathogenesis or succeeding at mechanism-based treatment development. Most notably these include the remarkable complexity of the human brain, its relative uniqueness with respect to animals that could serve as disease models, its relative inaccessibility to direct investigation, and a lack of robust neuropathological or biochemical correlates of symptoms as occurs in neurodegenerative disorders, such as Alzheimer’s disease (AD) and Parkinson’s disease (PD). Even for AD and PD, which exhibit clear neuropathology and for which we know many mutations that produce rare Mendelian genetic forms, the development of treatments that alter disease progression has proven challenging.
The brain is arguably the most complex object of scientific inquiry (with all due respect to our colleagues in physics and astronomy). The brain expresses approximately 80% of our full complement of genes, a far higher proportion than any other organ or organ system. Most organs are constructed of a relatively limited number of types of cells. In contrast, the human brain contains thousands of distinct types of neurons, the principal cells involved in the processing of information and the generation of outputs such as thought and behaviour. Each cell type is distinguished by the complement of genes that it expresses, i.e. reads out from its DNA into ribonucleic acid (RNA), and by how it processes RNA and proteins (which are translated from a subset of RNA molecules) into diverse molecular species. Ultimately the precise patterns of gene expression that define each cell type are reflected in such recognizable characteristics as their different neurotransmitters, receptors, ion channels, and other signalling molecules. In aggregate the cells of the brain possess a staggeringly rich and complex molecular vocabulary within which each cell type has its own identity. In addition to their distinct molecular composition, each type of neuron has a characteristic morphology, location in the brain, and stereotypic connections to other neurons. They are distinguished from other cell types by their connectivity via synapses, specialized connections that permit precise chemical neurotransmission. Neurons in the brain make, on average, approximately 1000 synapses with other neurons, but the patterns are highly diverse. Each of the large Purkinje cells of the cerebellum may contain 200,000 synapses. The other major cell type in the nervous system, the glia, have more complex functions than merely supporting neurons. For example, astrocytes, one type of glial cell, may participate in synaptic communication.
Overall the human brain contains approximately 100 trillion synapses that form local and large-scale circuits. The synapses, and thus the circuits in which they participate, are not unchanging features of the neural landscape, but undergo frequent remodelling or ‘plasticity’ in response to developmental processes, external stimuli including lived experience, hormones, drugs, and disease. Plasticity in synapses and circuits is responsible for the encoding of memories and for behavioural adaptations to the environment. Abnormal or maladaptive plasticity also plays an important role in the pathogenesis of many psychiatric disorders, whether highly genetically influenced as in schizophrenia and autism (12) or more strongly influenced by experience as in depression that can follow upon adversity or in post-traumatic stress disorder (PTSD). Treatments for psychiatric disorders, whether psychotherapies, medications, or neuromodulatory interventions such as DBS, must ultimately influence synapses and circuits since activity in the circuits produces all experience, thought, emotion, motivation, and behavioural control. This might sound terribly reductive unless one contemplates the extraordinary complexity and exquisite connectivity of the human brain, indeed one of the most profound mysteries of science.
The maladies that come under the purview of psychiatry involve dysfunction of many basic brain functions that humans share with other animals, which can result, for example, in abnormal sleep and circadian rhythms, appetite, energy, and motivation. Psychiatric disorders are distinguished, however, by disturbances in brain functions that are either unique to humans or significantly expanded in evolution by comparison with our near primate relatives. Thus many psychiatric symptoms affect higher cognition, emotional regulation, decision-making, and executive function that are poorly modelled in animals (13). Our limited ability to model cardinal features of psychiatric illness in animals creates a particularly high scientific hurdle because, for ethical as well as pragmatic reasons, the living human brain is not readily accessible to direct study. Even if brain biopsies were somehow imaginable for psychiatric disorders, they might provide limited information since these are not ‘cell autonomous’ diseases but the result of abnormal patterns of neural activity across widely distributed circuits.
Notwithstanding the extraordinary scientific challenges that confront psychiatry, the question for people affected by mental illness is how to make significant progress in diagnosis and treatment. In this essay I shall present an admittedly selective perspective of the historical path by which psychiatry has arrived at its current scientific crossroads, and a guardedly optimistic view of how the field might move forwards. Progress will require fresh thinking about disease classification, which exerts strong influence over translational and clinical research, new strategies to understanding aetiology and pathogenesis, and novel approaches to treatment development. If my account is correct, then the answers must lie in science. Specifically, I would argue that psychiatry has focused too narrowly (although of course, not exclusively) on pharmacological and endocrinal models of disease, and would do well to embrace broader aspects of neuroscience and other fields of modern biology. It is also worth noting that the continued separation of psychiatry and neurology (a topic beyond the scope of this essay), by inhibiting the free flow of ideas about diseases of the nervous system, remains highly unfortunate for both scientific progress and clinical practice (14).
Modern neuroscience began to coalesce as a recognizable field in the 1960s based initially on efforts to bring together scientists who had been studying the anatomy, physiology, and biochemistry of the nervous system within their own ‘silos’ (15). Soon thereafter psychologists with an interest in brain mechanisms and clinical scientists joined this increasingly interdisciplinary endeavour. As molecular biology, neurogenetics, computational biology, and bioengineering matured, more of their practitioners began to identify themselves as neuroscientists and to link up with departments and graduate programmes in neuroscience. Of course fundamental studies of the nervous system had been performed as early as the late 19th century by such pioneering scientists as Camillo Golgi and Santiago Ramón y Cajal. The latter was the main author of ‘neuron theory’, the foundational idea that communication within the nervous system occurs between distinct, morphologically polarized neurons, rather than across a continuous network of conjoined cells. During the early and mid-20th century basic findings included the ionic basis of the action potential (Hodgkin and Huxley), chemical neurotransmitters (Loewi and Dale), and the physiological basis of synaptic transmission (Fatt and Katz). However, these investigators lacked ready means of scientific interaction (15).
In 1962, the biologist Francis Schmitt created an interdisciplinary, cross-institutional Neuroscience Research Program based at the American Academy of Arts and Sciences that held symposia and published influential proceedings (16). In 1967, Stephen Kuffler at the Harvard Medical School founded the first neurobiology department. The Society for Neuroscience (SfN) was founded 2 years later. Fourteen hundred scientists attended its inaugural conference; since that time its membership has increased to over 40,000 (1). The SfN was founded as a North American society but its membership has always been international; neuroscience societies and organizations have proliferated globally as the field has expanded and matured.
In addition to its interdisciplinary character, several other factors are central to effective studies of the brain. Processes in the nervous system operate at multiple spatial and temporal scales, so understanding often demands the integration of information from diverse technologies and perspectives (17). Proceeding from ‘micro’ to ‘macro’ scales within the brain, an understanding of normal function and of disease must recognize the contributions of molecules, cells, synapses, and local and large-scale circuits. At the next level of organization, cognition, emotion, and behaviour can be viewed as complex emergent properties of activity in large-scale circuits. Finally, while the mechanisms by which our brains produce consciousness—our subjective mental lives—remain unknown, no serious thinker questions the brain’s fundamental role in the underlying processes.
As in any biological system, causal influences that act upon the brain function from ‘top down’ and ‘bottom up’. Acting bottom-up, sensory experiences are first captured by specific molecular detectors expressed by specialized sensory neurons. For example, opsin molecules on retinal neurons called rod and cone cells capture light. In the visual system, the brain analyses incoming information in complex circuits (entirely inaccessible to introspection) within the retina, thalamus, and cerebral cortex. This circuitry separately analyses the position of an object in space (the ‘where’ pathways) from its characteristics (the ‘what’ pathways). Within the latter, different circuits analyse the contours of an object, such as its colour, whether it is moving, and in what direction. All such information is synthesized by higher-level neural circuits that also integrate vision with information from other sensory modalities to produce the coherent picture of the world that we experience and that guides behaviour.
Bottom-up sensory processes are modulated from the top down. For example, some top-down control derives from circuits that have not only been shaped by evolution to recognize stimuli with survival relevance (e.g. danger and rewards), but also shaped by experience. All day humans are bombarded by an enormous amount of sensory information, but based on effective top-down influences, pay attention to only the small fraction that is ‘relevant’ or salient. Competing stimuli that would serve as distractions tend to be suppressed. In conditions such as attention deficit hyperactivity disorder (ADHD) such top-down processes work poorly and ‘distracters’ intrude upon important tasks. In depression and anxiety disorders, a different kind of pathological top-down process seems to be at work: stimuli with a negative valence tend to command attention at the expense of neutral or positive stimuli. Beyond paying attention, additional mechanisms determine what aspects of experience get encoded in memory, based on such factors as the salience of the experience and on prior experience. A top-down set of influences of interest to psychiatry is temperament influences not only the nature of responses to stimuli, but what pertains to a person to begin with.
Non-invasive neuroimaging technologies have emerged as sets of tools permitting measurement of brain structures and providing surrogates of neural activity. These technologies have made it possible to observe the living, functioning human brain, even if indirectly. However, much of the psychiatric neuroimaging literature remains correlational. The demonstration of causal mechanisms often requires the kind of invasive perturbations that can only be performed in animals. Nonetheless, combined with other technologies, such as genetics, neuroimaging has begun to provide clues to disease phenotypes and may eventually contribute to diagnostic tests and biomarkers.
In animal models it is possible to measure the activity of synapses and circuits with electrophysiological recordings and newer technologies including advanced microscopy. In living human beings neural activity is mostly studied indirectly using older tools such as electroencephalography (EEG), and more recently with magnetoencephalography (MEG) and various forms of non-invasive neuroimaging. These technologies cannot observe neural activity directly, but rely on detectable outputs that correlate with activity such as electrical and magnetic fields for EEG and MEG respectively, or other proxies of neural firing such as changes in blood flow or metabolism for neuro-imaging. Current non-invasive tools lack the fine spatial resolution of the recording electrodes that can be used in animals; EEG, MEG, and functional imaging modalities therefore reveal activity only in large ensembles of neurons.
The EEG detects endogenous electrical signals produced by the flow of ions across neural membranes when the brain is active. MEG detects the tiny magnetic fields produced by the flow of electrical current in neural processes. Each of these technologies has millisecond temporal resolution, the timescale of much neural activity. However, they have poor spatial resolution and are thus complemented by various forms of neuroimaging, which have better spatial resolution; on the other hand, their temporal resolution is of the order of a second—very slow compared to neural activity. Neuro-imaging relevant to psychiatry includes structural magnetic resonance imaging (MRI), diffusion tensor imaging (DTI), functional magnetic resonance imaging (fMRI), and positron emission tomography (PET). The signal measured in all forms of MRI is energy released; radio waves are pulsed through the brain when it is placed in a strong magnetic field. Its resolution has progressively improved based on the ability to use stronger magnetic fields. MRI used to measure brain structure has also benefited from computer algorithms that can correct for different head shapes and sizes and reliably measure the contours and volumes of specific anatomical structures like cerebral cortical thickness or hippocampal volume. Using structural MRI with such computational tools has demonstrated that prefrontal and temporal cortical grey matter loss is characteristic of schizophrenia. Similarly, measurement of hippocampal volume is a useful biological marker even in the presymptomatic stage. DTI is an MRI-based technology that makes it possible to study the white matter tracts that connect different brain regions. It exploits the fact that water molecules do not move randomly in the brain but tend to be constrained by fibre tracts. DTI reveals, for example, white matter damage following traumatic brain injury (TBI) that may not be visible through other means.
Beyond observations of the structure of the brain, functional imaging permits the study of the living human brain at work. fMRI is the predominant current technology both because of its good spatial resolution and its apparent safety. In contrast, PET requires the administration of radioactivity, albeit at very low levels. The most common implementation of fMRI technology depends on blood oxygen level dependent (BOLD) contrast as the marker of neural activity. As neurons fire, they (and surrounding glial cells) require more blood flow in order to exchange carbon dioxide for oxygen and to obtain glucose. Oxygenated haemoglobin and deoxygenated haemoglobin behave differently in the magnetic fields employed by MRI technology. Although the cellular mechanisms that couple activity to the generation of the BOLD signal are not fully clear, there is evidence that it is a good proxy for neuronal firing.
Cognitive neuroscientists and clinical investigators have used fMRI to map the circuitry involved in diverse cognitive and emotional tasks in health and disease. One well-studied example is the neurobiology of fear learning and how it changes in anxiety disorders. The amygdala is a complex nucleus (grey matter structure) within the temporal lobes that plays a significant role in the processing of emotionally salient stimuli. Among its functions, the amygdala responds to threats and is necessary in encoding into memory information about fear-inducing stimuli, including their context. When healthy people participate in fear conditioning experiments in the laboratory (e.g. associating an aversive stimulus, such as a mild shock or a fear-inducing photograph, with a previously neutral stimulus) the amygdala is activated for a brief period. In concert with this activation, the medial prefrontal cortex, which suppresses aspects of amygdala function, shows a reciprocal decrement in activity. In patients with PTSD, the amygdala becomes active at a lower threshold and remains so for longer. Moreover, fMRI shows that the medial prefrontal cortex is less active than in healthy people. Fear conditioning is a basic survival process conserved in evolution. Thus, these human investigations were predicated on antecedent animal experimentation. However, fMRI has also been used to map aspects of higher cognition that cannot be readily studied in animals.
The diversity of spatial and temporal scales, the complexity of building blocks (and their interactions) at each scale, and the multiplicity of causal influences converging on all outputs of the nervous system, require thoughtful experimental design and model-building. Reductionist strategies (i.e. breaking problems down into tractable pieces or working in simple model systems) are often required if there is to be meaningful progress. Such reduction of complexity in the service of feasibility, for example, motivated the use of simple invertebrate systems (the sea slug Aplysia californica) by Eric Kandel in his Nobel prize-winning studies of the molecular and cellular bases of memory—even though his ultimate goal has been to understand human memory (18,19). Applying such strategies in the service of progress does not imply that reduction is an ultimate goal of neuroscience. Reducing certain aspects of behaviour to molecular biology and chemistry might provide investigational tools relevant to therapeutics. However, a rich explanation of behaviour demands the integration of multiple levels of analysis. If we ask questions about behaviour, answers at the molecular level will not be useful, even if certain molecules play important roles. Thus, the attribution of depressive disorders to low serotonin levels, as found in pharmaceutical marketing material, would be anathema to reputable neuroscientists despite the roles of this neurotransmitter (20,21).
We must also appreciate that theories, hypotheses, and textbook explanations of many aspects of brain function are based on radically incomplete knowledge. Indeed, scientific understanding of psychiatric disorders remain crude. Diagnostic systems are still phenomenological (5,6) and well-validated biological markers for purposes of diagnosis or to judge efficacy of treatments remain challenging goals of research. In unusual cases, significant insights into the biology of certain psychiatric maladies have been obtained through ‘bottom-up’ approaches, as when rare, highly penetrant genetic mutations that cause syndromes associated with autism have been used to generate transgenic mouse models (22). Insights have also been gleaned looking ‘top down’ with diverse modalities of neuroimaging, often in combination with cognitive neuroscience or pharmacology (23–25). For instance, the use of neuroimaging together with biochemical measures to generate biomarkers in AD is promising (26), as is its combination with cognitive neuroscience to produce treatment biomarkers for the cognitive symptoms of schizophrenia (27). Nonetheless, understandings of pathogenesis, which demand both molecular information and better disease definitions, have progressed slowly and fitfully for autism, schizophrenia, bipolar disorder, depressive disorders, obsessive-compulsive disorder, and other morbid states, and remain far from complete.
As alluded to, the challenges come in part from the inherent nature of psychiatric disorders. For example, while anatomical abnormalities have been identified in schizophrenia based on neuroimaging and post-mortem studies, the kind of robust biochemical pathology observed in AD and other neurodegenerative disorders is not evident. Some rare forms of autism are caused by mutations in single genes but the vast majority of cases (and essentially all cases of other psychiatric disorders) appear, at least hitherto, to be genetically complex, making it difficult to identify molecular clues to pathogenesis. This is true even for disorders in which genes, in aggregate, play a large role.
In the mid-20th century, neuroanatomy, neurophysiology, and neuropathology yielded notable empirical discoveries and conceptual advances that shaped modern ideas of how emotion is processed in the brain (28,29). Despite the relevance of such findings to understanding the biology of psychiatric disorders, they did not explain specific symptoms and impairments. Historically, interest in brain anatomy and physiology within the psychiatric research community arrived later, with the emergence of non-invasive neuroimaging methods as described above. The discovery in schizophrenia of enlarged cerebral ventricles (30) and deficits in prefrontal cortical function (31) elicited great interest among both clinicians and researchers, not only because these represented replicable brain abnormalities associated with a psychiatric disorder, but also because they could be correlated with a particular subset of symptoms, namely cognitive impairment. However, it was an earlier and truly astonishing series of psychopharmacological discoveries that began to move psychiatry from a nearly exclusive focus on the psyche in the mid-20th century to a far greater focus on the brain (see Mitchell and Dusan Hadzi-Pavlovic, this volume, pp. 335–54).
Beginning with John Cade’s investigations of lithium in the late 1940s (32), a rapid succession of therapeutic discoveries initiated revolutionary changes in psychiatric practice and ignited a growing interest in biology (reviewed in 33). Astute observation of unexpected actions of chemical compounds led to their use in treatment of several severe psychiatric disorders for which specific treatment had previously been lacking. The first antipsychotic drug, chlorpromazine, synthesized in 1950, was initially tested as a pre-anaesthetic for use in surgery. The first tricyclic antidepressant, imipramine, was synthesized as a potential antipsychotic drug (which it proved not to be) as a result of chemical modifications of the three-ring structure of chlorpromazine. The first monoamine oxidase inhibitor antidepressant (MAOI), iproniazid, was originally developed to treat tuberculosis, for which it had no benefit, but was observed to improve the mood of depressed sanatorium patients. By 1957 the antidepressant properties of imipramine and iproniazid had been investigated and demonstrated. While there have been controversies over the degree to which antidepressants are efficacious and under what circumstances (34,35), early trials and clinical experience focused on more severely depressed patients, and made a convincing case that mental illnesses could be treated with at least some specificity by drugs affecting the brain. Subsequent studies of the action of antipsychotic and antidepressant drugs helped lead to the identification of molecular components of synapses, including neurotransmitter receptors and reuptake transporters that functioned as the targets of the different drugs.
This revolutionary series of discoveries gave rise to speculation that the monoamine neurotransmitters (noradrenaline, dopamine, and serotonin) targeted by the new antidepressant and antipsychotic drugs might be involved in pathogenesis. The idea that treatments identify disease mechanisms is invariably tarred by the post hoc ergo propter hoc fallacy. In fairness, however, hypothesized noradrenaline or serotonin deficits or dopamine excess were often introduced with appropriate caveats, and many proponents recognized the need for evidence independent of treatment effects (36). The hypothesis that depression might be caused by low levels of noradrenaline, serotonin, or both arose not only from the therapeutic action of drugs (tricyclics and MAOIs) that increased synaptic levels of these monoamines, but from the observations concerning reserpine, a drug that had been used as an antihypertensive and antipsychotic drug that depleted neurons of all monoamine neurotransmitters (noradrenaline, dopamine, and serotonin). Reserpine was associated with sadness, sedation, and, in a minority of cases, the onset of depression, and for such reasons fell out of clinical use. In contrast, psychostimulant drugs such as amphetamine that elevated levels of monoamines in synapses produced, at least transiently, elevations in mood and energy. Despite these apparent pharmacological clues, numerous attempts over ensuing decades failed to find monoamine-related biochemical abnormalities in depressed patients, even while the mechanism of action of antidepressants was recognized as far more complex than could be simply attributed to monoamine neurotransmitter levels (37). In subsequent decades, pursuit of a version of the monoamine hypothesis focused on the genetics of serotonin systems, specifically a common variant in the gene encoding the serotonin reuptake transporter, but no convincing results have emerged and, indeed, depression is more influenced by development and environment than by genes. The maximum heritability for depression is estimated at 35%, but even within this aggregate figure, it appears that many genes contribute small increments of risk.
Since the 1960s, much evidence has suggested that stress or significantly adverse events trigger a proportion of episodes of major depression. In depressive episodes, neuroendocrine systems release stress hormones including cortisol (38). It has subsequently emerged that marked early life stress alters neuroendocrine functioning in the long term and increases the risk of depression (39). Although the syndrome of depression differs markedly from the stress response, there is significant overlap. For example, both stress and depression may affect sleep, appetite, and cognitive functions such as attention. Based on these similarities, many animal models of depression involve administering stress. This line of research has proven fruitful but the resulting animal models have neither elucidated a convincing mechanism for depression nor identified novel drug treatments that have made it into clinical practice. Overall, the pathogenesis of depression and the mechanism of action of antidepressants, beyond their initial effects on the nervous system, have not been solved.
In the case of schizophrenia a dopamine excess hypothesis was undergirded by the observation that amphetamine, which increases synaptic monoamines, including dopamine, can produce or exacerbate psychosis when used continuingly at a high dose. Moreover, drugs that block D2 dopamine receptors (all approved antipsychotics) ameliorate psychotic symptoms. Perhaps the most direct evidence for dopamine abnormality is the observation, first made using single photon emission tomography (SPECT) and confirmed with PET, that amphetamine causes greater dopamine release in people with schizophrenia than in control subjects (25). But these findings do not explicate the panoply of symptoms that characterize schizophrenia, which include not only the positive (psychotic) symptoms that respond to D2 dopamine receptor antagonist antipsychotics, but also negative (deficit) symptoms and cognitive impairments that do not.
The dopamine hypothesis has been supplemented by a hypothesis involving impaired glutamate neurotransmission, spurred by the finding that N-methyl-D-aspartate (NMDA) glutamate receptor channel blockers, such as phencyclidine (PCP) and keta-mine, could induce psychotic-like symptoms in humans and, even at lower doses, impaired cognition. Interestingly, findings from large-scale genetic studies suggest that diverse components of excitatory synapses that utilize glutamate play roles in both autism and schizophrenia pathogenesis. However, neither the dopamine nor glutamate hypotheses in their current forms represent the complete story, given schizophrenia’s likely developmental origins leading to pathology in multiple brain circuits. Thus, abnormal dopamine or glutamate neurotransmission, while possibly playing roles in the formation of symptoms, may occur downstream of initiating as yet unknown developmental events.
These neurotransmitter- and hormone-based hypotheses of depression and schizophrenia are plausible enough to have captured substantial research attention. Frustratingly, however, despite decades of investigation, none of them has advanced sufficiently to explain the cause or the abnormalities of brain function.
As described above, despite large markets and the limited efficacy of existing treatments (40), the pharmaceutical industry has, since 2010, rapidly de-emphasized psychiatry or abandoned the field altogether. This may represent complex commercial decisions, but there is widespread doubt that adequate knowledg is available to develop better drugs that will gain regulatory approval. Specifically, no clear path exists to drugs with novel mechanisms of action that will provide subtantial benefit in terms of core features of autism or cognitive symptoms of schizophrenia, or that will prove more efficacious than drugs such as the antidepressants which action noradrenaline and serotonin neurotransmission (7,40). Given that no antidepressant is superior in effect to imipramine or the MAOIs, and no antipsychotics are better than clozapine (a drug first used in the 1960s), the dominant focus on monoamine systems has become sterile. Essentially all progress in antidepressants has come through the development of drugs, such as the selective serotonin reuptake inhibitors (SSRIs), that are safer and have fewer side effects than their predecessors but are no more effective. In the case of antipsychotics, the second-generational group again has yielded no greater benefits and even a new set of side effects. Unfortunately, there is a dearth of validated molecular targets with which to replace the old monoamine warhorses. By validation I am referring to existence of a proven biological mechanism by which the molecular target of a candidate drug can influence disease mechanisms. Even if such targets were to be discovered, it will be difficult to know whether the drug binds its intended target in the human brain. One laborious but convincing method of ensuring so-called target engagement is via PET. While it has been used in ways analogous to fMRI to study brain activity using markers of glucose utilization or of blood flow, PET is the tool of choice to label specific models in the brain. Doing so requires the synthesis of a drug known to bind a specific molecular target, such as a neurotransmitter receptor or enzyme, with the addition of a positron-emitting isotope. Ensuring that the chemistry of the added isotope does not impair the desired binding is a challenge. When this approach succeeds, the PET ligand, namely the molecule that binds a receptor, can be administered to a human subject and a range of doses tried. The goal is for the candidate to compete with, and displace, the PET ligand, thus causing a dose-dependent decrease in the radioactive signal detected in appropriate brain regions. This methodology has the added advantage of determining the required drug dosage.
The scarcity of validated targets is not the only problem that inhibits investment by the pharmaceutical industry in psychiatry. A serious hurdle is the limitations of animal models for most psychiatric disorders. In all of medicine, animal disease models play crucial roles in studying pathogenesis and testing the toxicity and efficacy of drug candidates. The paramount criterion for the validity of a disease model is whether it is produced by the same mechanism as the relevant human condition and reproduces much, if not all, of its pathophysiology. For example, genetic animal models can be produced by replacing mouse genes with adequately penetrant human disease genes by infecting an animal with a relevant pathogen, or in the case of many cancers, by placing human cancer cells in an immunodeficient mouse. Lacking knowledge of aetiology or pathophysiology, validated animal models of depression, schizophrenia, bipolar disorder, and many other conditions are elusive. It is not clear that entirely useful animal models could be produced even with relevant information about pathogenesis given that psychiatric disorders involve, in part, brain structures and functions (e.g. the prefrontal cortex) uniquely advanced in humans. Of course animal research has been central for basic neurobiology and of relevance to psychiatry, yielding disease-relevant information concerning neural circuits conserved in evolution from rodents to humans. Such circuits underlie certain basic emotions such as reward and fear, and cognitive functions like memory (41). However, the utility of rodent models is limited in studying many of the circuits and functions of the prefrontal cortex, a critical region for psychiatry given its role in executive function. Many circuits in humans are rudimentary or absent in our ubiquitous rodent models; for instance, rodent cortices lack gyri and sulci that markedly expand the surface area of the human cerebral cortex. In addition, patterns of gene expression in the human prefrontal cortex differ substantially even from those in chimpanzees, our closest evolutionary relatives.
Many animal-based assays developed to identify possibly therapeutic drugs in psychiatry have become falsely conflated with disease models. Widely used assays in rodents, such as the forced swim (Porsolt test) and tail suspension test to detect anti-depressants, or amphetamine administration to detect antipsychotic drugs, originated as empirical attempts to predict new drugs with actions that would mimic those of the prototypical drugs of the 1950s like imipramine or chlorpromazine. A widely recognized problem with deploying assays based on existing drugs is that they may fail to detect new mechanisms to treat disease. Unfortunately, this has proved to be the case in psychiatry. Indeed, the molecular targets of all commonly used psychiatric drugs are the same as those of their 1950s prototypes.
In addition to the problems related to target identification and animal models, another gap in knowledge inhibiting investment in treatment discovery and development is the phenomenological diagnostic system that characterizes psychiatry and the associated lack of biological markers for disease or treatment response. Although progress in neuroimaging, genetics, and other forms of investigation may ultimately provide objective markers, these have not yet been identified. Pharmaceutical companies are dissuaded from developing new drugs if they must rely on descriptive DSM diagnoses (as described later) that almost certainly identify heterogeneous syndromes as official indications for treatment development. Similarly, lacking objective markers to gauge treatment response, they have been forced to rely on subjective rating scales such as the Hamilton Rating Scale used for depression, which makes clinical trials almost prohibitively difficult. Depression, like many other psychiatric conditions, is characterized by waxing and waning symptoms that can readily confound the results of treatment trials.
The history of psychiatric diagnosis is very much tied to history of treatment. In concert with the discovery of the effects of lithium, antidepressants, antipsychotics, and benzodiazepines, there was resurgent interest in diagnosis that produced concerted efforts in the 1960s and 1970s. The latter was motivated in no small part by the need to match patients with the most appropriate interventions, and also by a wish to define psychiatric disorders in ways that were thought to be compatible with biology. The Department of Psychiatry at Washington University, St. Louis separated itself from the psychoanalytic mainstream and returned to Emil Kraepelin’s diagnostic concepts. He had asserted that, as in the rest of medicine, psychiatry could identify specific disorders, likely of biological origin, based on careful descriptions of symptoms and course. He illustrated this approach by carefully differentiating what he called dementia praecox (later renamed schizophrenia) from manic-depressive illness (later called bipolar disorder). In a seminal article that launched an era of research on psychiatric diagnosis in the 1970s, Robins and Guze (42) argued that it was possible to achieve reliable and valid diagnoses based on clinical description, laboratory study, exclusion of other disorders, follow-up observations, and family research. Given the lack of biological markers, their work, like that of Kraepelin, was based entirely on clinical description but had profound influences that brought psychiatry closer to both biology and other fields of of medicine.
The descriptive approach championed by the St Louis research group came to dominate psychiatric diagnosis because of its central role in the third edition of the widely influential Diagnostic and Statistical Manual of Mental Disorders (DSM) (43), published in 1980. In the service of reliability, the manual used operationalized criteria in a manner that had been pioneered by the St. Louis group in the early 1970s (44). With the sole exception of mental retardation (which was defined as a quantitative deviation from the normal based on IQ) all disorders in the DSM-III, DSM-IV (45), and DSM-V (the last published in May 2013) were defined as categories that could be distinguished from health and from one other. On the assumption that it was possible to identify homogeneous categories for the purposes of research and treatment, DSM-III created a fine-grained subdivision of psychopathology that resulted in a remarkably large number of specific diagnoses. DSM-III advanced psychiatry by providing a common diagnostic language but, paradoxically, its very success in this regard has led to questioning of its phenomenological and categorical approach to classification.
Through the identification of homogeneous patient groups, Robins and Guze (42) argued that their careful descriptive approach would lead not only to reliable diagnoses (i.e. the agreement of different raters), but also to valid entities (the identification of ‘natural kinds’). In fact, descriptive psychiatry as represented in the last three editions of the DSM is a poor reflection of nature (5,6). Robins and Guze did not foresee the remarkable heterogeneity of psychopathology, which has been progressively elucidated applying new genetic technology (see McGuffin, this volume, pp. 22–44) (12). Nonetheless, with the benefit of hindsight, DSM-III applied descriptive psychiatry and set the stage for the emergence of major clinical and scientific problems. The abundance of diagnostic silos has led to a high degree of comorbidity that is, in the main, the result of inappropriate subdivision of pathological entities sharing risk factors or neural mechanisms. Similarly, the overly narrow silos have forced scrupulous clinicians to use the ‘not otherwise specified’ (NOS) subtype, which lacks informational content. Scientifically, the DSM categories have not mapped onto the results of family and genetic studies, and assert arbitrary diagnostic thresholds.
Pathological states based on quantitative deviations from health have long been well known in medicine; these include such conditions as hypertension and type 2 diabetes mellitus. By contrast, Kraepelin, the St. Louis school, and the DSM system all treated psychiatric disorders as discontinuous categories in a manner analogous to infectious disease, such as pneumococcal pneumonia or tuberculosis. The implicit idea (occasionally explicit) in the literature of descriptive psychiatry is that a valid disorder is a homogeneous category produced by one or, at most, a limited set of related aetiologies. In fact, psychiatric disorders—like most common, enduring medical conditions—are heterogeneous and genetically complex (12); that is, a disease phenotype can result from multiple genetic pathways interacting with many non-genetic factors, with no gene either necessary or sufficient for the disorder. Psychiatric disorders that have been studied in detail also appear to be polygenic, meaning that a large number of genes contribute to risk. With the exception of some rare forms of autism, genes that confer risk for psychiatric disorders act in a statistical rather than a deterministic fashion.
Well-studied conditions such as autism (46), schizophrenia (47,48), and bipolar disorder appear to be remarkably polygenic. Gottesman and Shields (49) presciently raised the possibility as early as 1967 that schizophrenia is polygenic, but unfortunately their work had little influence on model-building for many years. Empirically, the contribution to risk conferred by variation in DNA sequences (1) is both inherited and de novo (2,46); reflects both common and rare sequence variants (3); affects protein coding genes and regulatory sequences; and involves (4) single nucleotide variants (SNVs) and—at least for a proportion of cases of autism, schizophrenia, and other neurodevelopmental disorders—larger copy number variants (CNVs). There is far greater sharing of genetic risk factors across different diagnoses (50) than was expected, not dissimilar to what has emerged across various forms of inflammatory bowel disease and several autoimmune disorders.
In sum, the relationship between genetics and nosology remains opaque (51,52) but, as is the case of many polygenic diseases, psychiatric disorders are better captured as quantitative deviations from health on multiple dimensions than as discontinuous DSM-style categories. For example, symptoms and, in some cases, laboratory findings associated with autism (53), schizophrenia, and mood disorders are normally distributed in the population with no clear natural demarcations that would justify a categorical break. The implications for science, clinical practice, and treatment development are profound. For example, it is not sensible to try to develop biomarkers for current DSM categories that in retrospect should have been seen as transiently useful fictions. Prominent attempts to start ab initio include the Research Domain Criteria (RDoCs) initiative of the National Institutes of Health (NIMH) (54), which foresees diagnoses that are entirely dimensional. In its initial iteration this project attempts to start with brain circuits for which patterns of cognition, emotion, and motivation relevant to psychiatry are adequately known; for example, fear, reward, and prefrontal cortical circuitry, the last involved in executive function.
A diagnostic system must also have clinical utility; in that regard, rough and ready categories can be extremely useful (55). Fortunately, imposing such categories on an underlying structure of quantitative dimensions has been shown to work well in many areas of medicine. For example, ‘bins’ relevant to treatment are superimposed on multiple dimensions that define hypertension (these include systolic and diastolic blood pressure) and dyslipidaemias (these include measures of total cholesterol, LDL cholesterol, HDL cholesterol, and triglycerides). Psychiatry is hardly alone in having to rethink nosology for the purposes of improving treatment. There is a dawning recognition in oncology that patterns of the particular somatic mutations within a tumour (i.e. the acquired genetic lesions that drive oncogenesis), are likely to be more important in selecting drug treatments for a cancer than its tissue of origin. As such, lung or breast cancer is turning out to be not one disease, but many, and treatment selection is increasingly made on the basis of molecular mechanisms.
Given the challenges we have described above, a critical question for psychiatry is how to develop new hypotheses that can inform neurobiological investigations relevant to disease mechanisms and treatment. How can psychiatry escape the tunnel vision created by its descriptive straitjacket and over-reliance on old pharmacological and endocrinal models of disease. Across all of biology, gaps in knowledge are being addressed through large-scale, unbiased (hypothesis neutral) modes of discovery that can complement and ultimately enrich hypothesis-driven science. As with reductionist strategies, the databases, ‘parts lists’, and tools produced by such approaches must ultimately be integrated into multi-level explanatory models. While not a panacea, large-scale unbiased approaches can, along with a deeper engagement with neuroscience, help renew psychiatric research.
Genomics, the systematic determination and analysis of the complete DNA sequence of an organism (i.e. its genome) was the first area of biology to coalesce around large-scale, systematic efforts to generate shared data. The Human Genome Project (HGP) provided organization and support to sequence entire genomes, including an initial reference sequence of the human genome. In addition to DNA sequence information per se a key result of the HGP and parallel efforts in industry was investment in technologies to increase the speed and accuracy of DNA sequencing and to decrease its cost (56). Indeed, the cost has declined a million-fold over a mere decade, making it possible to determine the DNA sequences of thousands of human genomes, as well as those of many animal species, and even to obtain nearly complete sequences from the remains of ancient humans such as Neanderthals (57). In parallel with sequencing technologies, increasingly fast and powerful computing and vast, inexpensive data storage capacities have made modern genomics possible.
Technologies and intellectual approaches developed for genomics have facilitated great strides in the ability of human genetics to analyse complex traits, including risk of genetically complex common diseases. At the population level, a trait such as disease risk may result from many different rare DNA variants acting in different individuals to produce a similar phenotype, from the interaction of many different common sequence variants each contributing a small effect, or a combination of the two. This means that to achieve the statistical power to discover variations in DNA sequences contributing to disease risk, we need to study large numbers of patients and control subjects in various research designs. New technologies have made this both feasible and affordable. Thus progress is accelerating in identifying the genetic architecture of heterogeneous and polygenic disorders like autism, schizophrenia, and bipolar disorder. Given our ignorance of the disease mechanisms in psychiatry, it is critical that genetic strategies remain agnostic concerning biological hypotheses. Research has revealed many previously unimagined genetic loci conferring risk of disorder (12,46), and far greater sharing of risk-conferring genes across diagnoses (58). Hitherto, systematic approaches (59) have also failed to confirm many reports from past small studies based on biological hypotheses. Although disappointing, the frailty of such hypotheses has reaffirmed the need for humility in confronting brain complexity, and underscored the central role of unbiased methodology.
By analogy with genomics, systematic attempts to describe and analyse all members of a class have received the fashionable suffix of ‘omics’. Although they are on the increase, big, costly, technology-driven projects, beginning with the HGP, have often been controversial (60,61) because of the concern they will divert resources from smaller laboratories engaged in hypothesis-driven science. It is also the case that not all ‘big science’ projects supported by governments and foundations have proved cost-effective. Yet it is increasingly clear that if we are to understand such complex matters as the mechanisms underlying mental disorders, we will need both large-scale unbiased discovery projects and hypothesis-driven science. The large databases and reagent collections derived from ‘omics’ research should ultimately provide a rich source of new hypotheses to drive inquiries into psychiatric conditions.
With the exception of some rare forms of autism, DNA sequence is not destiny when it comes to psychiatric disorders. This can be illustrated by the observation that in schizophrenia, bipolar disorder, major depression, and other psychiatric disorders monozygotic twin pairs (derived from a single fertilized ovum and sharing 100% of their DNA sequences) do not exhibit 100% concordance for the disorder. Indeed, in schizophrenia, which is highly genetically influenced, given a monozygotic co-twin with the disorder, the other member of the pair has only a 50% risk of full-blown schizophrenia. (The co-twin might show a degree of cortical thinning and cognitive impairment but never psychotic symptoms.) Specific environmental factors or chance might convert genetic risk into psychiatric disease. These environmental and stochastic factors might act by influencing the rates at which specific genes are transcribed into RNA. Especially when such factors act during embryogenesis, and perhaps early postnatal life, they may persist into adulthood by creating long-lived patterns in the proteins that package DNA in the cell nucleus. These patterns, which determine whether a particular gene is available to be transcribed, are captured by the term ‘epigenetics’. Historically the study of epigenetic mechanisms was limited to prenatal development. The last decade has seen the application of epigenetics to postnatal life. For example, influential models of stress due to early maternal separation in rat pups (e.g. 62) have influenced the study of human development by suggesting that early adversity might act via epigenetic mechanisms to alter responses to stress for many years. This historically recent extension of epigenetics to become a possible mechanism by which postnatal experience causes persistent changes in physiology and behaviour is still at a stage where the significance of the evidence is a matter of debate.
The global study of gene expression (the transcriptome) or of proteins (the proteome) in cells or organisms adds a new level of complexity to ‘omics’, as the production and modification of RNA and proteins are regulated processes that differ according to cell types and conditions. Thus, unlike genomics in which each individual organism has a single unique set of DNA sequences (with the caveat concerning somatic mutations), each cell type, and each of the many types of neurons and glial cells in the brain, has its own transcriptome and proteome, which change as these cells are subjected to diverse signals and stresses. Relevant research demands efficient high-throughput technologies such as gene arrays (or ‘chips’) to study patterns of global gene expression; and mass spectrometry in the case of proteomics. Alongside these approaches, enquiry into the biology of mental disorders will benefit from future studies of the epigenome (global studies of epigenetic modifications), the metabolome (the complete set of metabolic intermediates in a cell under particular conditions), and the connectome (systematic wiring diagrams of the nervous system or systematic maps of brain activity).
In many areas of medical research, such as cancer biology, diseased cells can be obtained after surgical biopsy or resection. These cells can then be examined directly, made into cell lines that will propagate indefinitely, or injected into immunologically compromised mice to study tumour formation. Here lies another hurdle for psychiatric research. Brain biopsies for psychiatric disorders—unlike those for cancer, including brain cancer—are ethically unimaginable. Moreover, a localized sample of brain tissue might not be informative about disease mechanisms since psychiatric disorders are the result of abnormal physiology affecting widely distributed circuits and are not ‘cell autonomous’. Nonetheless, the study of living human neurons is critical if we are to study genes that confer disease risk successfully. Part of the uniqueness of the human brain appears to be its patterns of gene expression in multiple regions of the forebrain. Therefore, even as psychiatry wrestles with the utility of animal models, it will be necessary to devise models based on human neurons to interrogate gene function and other domains of disease-relevant biology.
Remarkably, it is becoming feasible to study molecular and cellular aspects of psychiatric disorders with human neurons in vitro using stem cell technologies. One such approach makes it possible to generate pluripotent cells (cells that can generate any tissue in the body) from skin fibroblasts. This technology, which resulted in the award of the 2012 Nobel Prize to Shinya Yamanaka, initially used engineered viruses to carry genes that encoded four transcription factors (proteins that control gene expression) into fibroblasts that had been placed in culture. The resulting induced pluripotent stem cells (iPSCs) can then be used to manufacture diverse other cell types including neurons. The technology is moving rapidly, so that specific types of neurons have been produced, such as motor neurons. It is a pivotal goal to identify the specific neuronal cell types involved in psychiatric disorders and to be able to engineer those cells.
One major advantage of iPSC technology is that skin fibroblasts can be taken both from control subjects and from patients, and the cells can be fully genotyped. It is now readily possible to engineer disease-risk-associated DNA sequence variations into the genomes of the ‘control’ cells and to ‘rescue’ patient cell phenotypes by replacing risk-conferring DNA sequences with non-risk variants. Initial experiments relevant to psychiatry have compared neurons engineered from skin biopsies of patients with schizophrenia and neurons derived from control subjects and have reported phenotypic differences (63). It is early days in the adaptation of this technology to psychiatric disorders, but if successful, this might partially obviate the need for animal models.
While molecular and cellular approaches are critical to the pursuit of new medications, neural circuit activity underlies psychiatric symptoms and is therefore relevant to improved diagnosis, possibly to treatment biomarkers, and new non-pharmacological treatments ranging from DBS (9) to cognitive therapies (10,27). For all of these reasons, a high-resolution connectivity map of the human brain is a critical platform for psychiatric research to the same degree as a complete catalogue of genetic and epigenetic contributors to risk. Given that the brain has 100 trillion synapses underlying its connections, even achieving a rough wiring diagram poses a challenge. As in the case of genomics, however, new technologies combined with greater computing power and data storage capacity have made it possible to envision a serious effort in ‘connectomics’. The ultimate goal of such efforts is connectomes—high-resolution maps of all neurons and their synaptic connections for entire organisms, including humans. Much of the micro-scale mapping at the level of cells and synapses has perforce begun with model organisms. Much of the initial focus for the human connectome has been at the macro-scale and dependent on an array of neuroimaging technologies including methods for mapping grey matter, white matter tracts, and activity.
The ultimate utility of connectivity maps for psychiatry requires functional maps superimposed on structural ones that, to date, in humans, have largely relied on PET and fMRI. As described earlier, however, functional imaging in humans is largely correlational. New technologies, most notably optogenetics, implemented in animals, have now made it possible to stimulate or suppress activity in particular neural cells and circuits and thus to test hypotheses about their functions. Engineered viruses carrying specific genes are injected into the brain, or genes are inserted by recombination into the germ line of animals. The genes carried into the brain are designed to be turned on only within specific neural cell types where they express light-activated ion channels. The pioneering work used a gene from algae that encoded a light-activated ion channel, rhodopsin, but other light-regulated channels have since been used that can produce neuronal firing or inhibition in response to specific wavelengths of light. The desired wavelength, introduced into the animal brain through fibre optic instruments, permits the activation or inhibition of specific cells and, thus specific circuits, on a millisecond timescale (2,64). The specificity of optogenetics is not only temporal but also spatial compared to previous stimulation methods, generally involving electrodes, which involved neighbouring axons in addition to those targeted intentionally. Optogenetics has already yielded salient new insights into the control of fear, reward, aggression, and many other functions relevant to psychopathology and, in many cases, can be extrapolated to humans.
Psychiatry has reached a crossroads; the understanding of disease processes has progressed very slowly, resulting in the suspension of the development of significant novel medications. The exit of the pharmaceutical industry from the psychiatric arena is a symptom of this stasis and, at the same time, an impediment to progress in treating psychiatric patients. The half-century stasis in therapeutics reflects, above all, the exceeding difficulty of studying heterogeneous, polygenic disorders. At the same time, seemingly unnecessary obstacles have also limited advances. These encompass the continuing acceptance of a diagnostic system that has palpably limited translational and clinical science, and a set of research paradigms rooted in the pharmacological discoveries of the 1950s that have resulted in many funding grants and publications but no deep understanding of disease mechanisms (7). Despite the challenges, this is not a time for pessimism. New technologies including large-scale unbiased genetic studies, the early development of stem cell technologies with all their implications for the study of disease, and advances in understanding neural circuits can, if effectively deployed, move psychiatry past the current crossroads.
1 http://www.sfn.org/index.aspx?pagename=about_SfN#timeline [Referenced 24 August 2012].
2 Fenno, L., Yizhar, O., and Deisseroth, K. (2011). The development and application of optogenetics. Annual Review of Neuroscience 34, 389–412.
3 Fox, M. D., Synder, A. Z., Vincent, J. L., Van Essen, D. C., and Raichle, M. E. (2005). The human brain is intrinsically organized into dynamic, anticorrelated functional networks. Proceedings of the National Academy of Sciences of the U S A 102, 9673–8.
4 http://www.alleninstitute.org [Referenced 16 August 2012].
5 Hyman, S. E. (2007). Can neuroscience be integrated into the DSM-V? Nature Reviews Neuroscience 8, 725–32.
6 Hyman, S. E. (2010). The diagnosis of mental disorders: the problem of reification. Annual Review of Clinical Psychology 6, 155–79.
7 Hyman, S. E. (2012). Revolution stalled. Science Translational Medicine 4, 155cm11.
8 Insel, T. R. (2012). Next-generation treatments of mental disorders. Science Translational Medicine 4, 155ps19.
9 Holtzheimer, P. E. and Mayberg, H. S. (2011). Deep brain stimulation for psychiatric disorders. Annual Review of Neuroscience 34, 289–307.
10 Minzenberg, M. J., and Carter, C. S. (2012). Developing treatments for impaired cognition in schizophrenia. Trends in Cognitive Science 16, 35–42.
11 World Health Organization (2008). The global burden of disease, 2004 update. Geneva: World Health Organization.
12 Sullivan, P. F., Daly, M. J., and O’Donovan, M. (2012). Genetic architecture of psychiatric disorders: the emerging picture and its implications. Nature Reviews Genetics 13, 537–51.
13 Nestler, E. J., and Hyman, S. E. (2010). Animal models of neuropsychiatric disorders. Nature Neuroscience 13, 1161–9.
14 Martin, J. B. (2002). The integration of neurology, psychiatry, and neuroscience in the 21st century. American Journal of Psychiatry 159, 695–704.
15 Cowan, W. M., Harter, D. H., and Kandel, E. R. (2000). The emergence of modern neuroscience: some implications for neurology and psychiatry. Annual Review of Neuroscience 23, 343–91.
16 Schmitt, F. O., Melnechuk, T., Adelman, G., and Worden, F. (1966) Neurosciences research symposium summaries, Vol 1. Cambridge, MA: MIT Press.
17 Kopnisky, K. L., Cowan, W. M., and Hyman, S. E. (2002) Levels of analysis in psychiatric research. Developmental Psychopathology 14, 437–61.
18 http://www.nobelprize.org/nobel_prizes/medicine/laureates/2000/kandel-lecture.pdf [Referenced 25 August 2012].
19 Pittinger, C. and Kandel, E. R. (2003). Insearch of general mechanisms for long-lasting plasticity: aplysia and the hippocampus. Philosophical Transactions of the Royal Society London B Biological Sciences 358, 757–63.
20 Dayan, P. and Huys, Q. J. (2008). Serotonin, inhibition, and negative mood. Public Library of Science Computational Biology 4(2), e4.
21 Dayan, P. and Huys, Q. J. (2009). Serotonin in affective control. Annual Review of Neuroscience 32, 95–126.
22 Peça, J., Feliciano, C., Ting, J. T. et al. (2011). Shank3 mutant mice display autistic-like behaviours and striatal dysfunction. Nature 472, 437–42.
23 Thompson PM, Vidal C, Giedd JN et al. (2001). Mapping adolescent brain change reveals dynamic wave of accelerated gray matter loss in very early-onset schizophrenia. Proceedings of the National Academy of Sciences of the U S A 98, 11650–5.
24 Barch, D. M., Carter, C. S., Braver, T. S. et al. (2001). Selective deficits in prefrontal cortex function in medication-naive patients with schizophrenia. Archives of General Psychiatry 58, 280–8.
25 Laruelle, M., Abi-Dargham, A., van Dyck, C. H. et al. (1996). Single photon emission computerized tomography imaging of amphetamine-induced dopamine release in drug-free schizophrenic subjects. Proceedings of the National Academy of Sciences of the U S A 93, 9235–40.
26 Westman, E. Muehlboeck, J. S., and Simmons A (2012). Combining MRI and CSF measures for classification of Alzheimer’s disease and prediction of mild cognitive impairment conversion. Neuroimage 62, 229–38.
27 Carter, C. S., Barch, D. M.; CNTRICS Executive Committee (2012). Imaging biomarkers for treatment development for impaired cognition: report of the sixth CNTRICS meeting: Biomarkers recommended for further development. Schizophrenia Bulletin 38, 26–33.
28 Nauta, W. J. (1946). Hypothalamic regulation of sleep in rats: an experimental study. Journal of Neurophysiology 9, 285–316.
29 MacLean, P (1990). The triune brain in evolution: role in paleocerebral functions. New York: Plenum Press.
30 Johnstone, E. C., Crow, T. J., Frith, C. D., Husband, J., and Kreel, L (1976). Cerebral ventricular size and cognitive impairment in chronic schizophrenia. Lancet 2, 924–6.
31 Berman, K. F. (1987). Cortical ‘stress tests’ in schizophrenia: regional cerebral blood flow studies. Biological Psychiatry 22, 1304–26.
32 Cade, J. F.J (1949). Lithium salts in the treatment of psychotic excitement. Medical Journal of Australia 2: 349–52.
33 Mitchell, P. B. and Hadzi-Pavlovic, D (this volume).
34 Kirsch, I., Deacon, B. J., Huedo-Medina, T. B., Scoboria, A. Moore, T. J., and Johnson, B. T. (2008). Initial severity and antidepressant benefits: a meta-analysis of data submitted to the Food and Drug Administration. Public Library of Science Medicine 5(2), e45.
35 Horder, J., Matthews, P., and Waldmann, R (2011). Placebo, Prozac and PLoS: significant lessons for psychopharmacology. Journal of Psychopharmacology 25, 1277–88.
36 Schildkraut, J. J. and Kety, S. S. (1967). Biogenic amines and emotion. Science 156, 21–37.
37 Hyman, S. E. and Nestler, E. J. (1996). Initiation and adaptation: a paradigm for understanding psychotropic drug action. American Journal of Psychiatry 153, 151–62.
38 Sachar, E. J. (1967). Corticosteroids in depressive illness: I. A reevaluation of control issues and literature. Archives of Geneneral Psychiatry 17, 544–53.
39 Heim C, Newport D. J., Heit, S. et al. (2000). Pituitary-adrenal and autonomic responses to stress in women after sexual and physical abuse in childhood. JAMA 284, 592–7.
40 Khin, N. A., Chen, Y-F., Yang, Y., Yang, P, and Laughren, T. P. (2011). Exploratory analyses of efficacy date from major depressive disorder trails submitted to the US Food and Drug Administration in support of new drug applications Journal of Clinical Psychiatry 72, 464–72.
41 Fernando, A. B. P. and Robbins T. W. (2011). Animal models of neuropsychiatric disorders. Annual Review of Clinical Psychology 7, 39–61.
42 Robins, E. and Guze, S. B. (1970). Establishment of diagnostic validity in psychiatric illness: Its application to schizophrenia. American Journal of Psychiatry 126, 983–7.
43 American Psychiatric Association (1980). Diagnostic and statistical manual of mental disorders (3rd edn). Washington, DC: American Psychiatric Association.
44 Feighner, J. P., Robins, E., Guze, S. B., Woodruff, R. A. Jr, and Munoz, R. (1972). Diagnostic criteria for use in psychiatric research. Archives of General Psychiatry 26, 57–63.
45 American Psychiatric Association (1994). Diagnostic and statistical manual of mental disorders (4th edn). Washington, DC: American Psychiatric Association.
46 Neale B. M., Kou Y., Liu L., Ma’ayan A. et al (2012). Patterns and rates of exonic de novo mutations in autism spectrum disorders. Nature 485, 242–5.
47 Lee, S. H., DeCandia, T. R., Ripke, S. et al (2012). Estimating the proportion of variation in susceptibility to schizophrenia captured by common SNPs. Nature Genetics 44, 247–50.
48 Gejman, P. V., Sanders, A. R., and Kendler, K. S. (2011). Genetics of schizophrenia: New findings and challenges. Annual Review of Genomics and Human Genetics 12, 121–44.
49 Gottesman, I. I. and Shields, J. (1967). A polygenic theory of schizophrenia. Proceedings of the National Academy of Sciences of the U S A 58, 199–205.
50 Lichtenstein, P., Carlstrom, E., Rastam, M., Gillberg, C., and Anckarsater, J. (2010). The genetics of autism spectrum disorders and related neuropsychiatric disorders in childhood. American Journal of Psychiatry 167, 1357–63.
51 Kendler, K. S. (2006). Reflections on the relationship between psychiatric genetics and psychiatric nosology. American Journal of Psychiatry 163, 1138–46.
52 Craddock, N., O’Donovan, M. C., and Owen, M. J. (2006). Genes for schizophrenia and bipolar disorder? Implications for psychiatric nosology. Schizophrenia Bulletin 32, 9–16.
53 Di Martino, A., Shehzad, Z., Kelley, C. et al. (2009). Relationship between cingulo-insular functional connectivity and autistic trats in neurotypical adults. American Journal of Psychiatry 166, 891–9.
54 http://www.nimh.nih.gov/research-funding/rdoc/index.shtml [Referenced 24 August 2012].
55 Kendell, R. and Jablensky, A. (2003). Distinguishing between validity and utility of psychiatric diagnoses. American Journal of Psychiatry 160, 4–12.
56 Shendure, J. and Ji, H. (2008). Next-generation DNA sequencing. Nature Biotechnology 26, 1135–45.
57 Green, R. E., Krause, J., Briggs, A. W. et al. (2010) A draft sequence of the neandertal genome. Science 328, 710–22.
58 Talkowski, M. E., Rosenfeld, J. A., Blumenthal, I. et al. (2012). Sequencing chromosomal abnormalities reveals neurodevelopmental loci that confer risk across diagnostic boundaries. Cell 149, 525–37.
59 Major Depressive Disorder Working Group of the Psychiatric GWAS Consortium (2013). A mega-analysis of genome-wide association studies for major depressive disorder. Molecular Psychiatry 18, 497–511.
60 Collins, F. S., Morgan, M., and Patrinos A. (2003). The Human Genome Project: lessons from large-scale biology. Science 300, 286–90.
61 Lichtman, J. W. and Sanes J. R. (2008). Ome sweet ome: What can the genome tell us about the connectome? Current Opinions in Neurobiology 18, 346–53.
62 Meaney, M. J. and Szyf, M. (2005). Maternal care as a model for experience-dependent chromatin plasticity. Trends in Neuroscience 28, 456–65.
63 Brennand, K. J., Simone, A., Jou, J. et al. (2011). Modelling schizophrenia using human induced pluripotent stem cells. Nature 473, 221–5.
64 Deisseroth, K. (2012). Optogenetics and psychiatry: Applications, challenges, and opportunities. Biological Psychiatry 71, 1030–2.