What Do We Want To See in Brain Imaging?

Web Sites

Nature NeuroPod
The December 2007 edition of the Nature Neuroscience podcast featured a report on this conference. Click here to download the podcast. (The segment on the conference begins at 14:30.)

Introduction to fMRI
This site from the FMRIB Centre at the University of Oxford contains a tutorial on functional magnetic resonance imaging.

Learn more about Brain Imaging Technologies
Part of the Genetic Science Learning Center at the University of Utah, this site aims to explain imaging technologies for the public.

LONI: Laboratory of Neuro Imaging
The Web site of the laboratory of Neuro Imaging at UCLA contains a resources section and links to imaging consortia including the International Consortium for Brain Mapping and the Alzheimer's Disease Neuroimaging Initiative.

Statistical Parametric Mapping
This site, put together by members and collaborators of the Wellcome Department of Imaging Neuroscience, contains software for the analysis of brain imaging sequences along with instructions on how to use it.

Wolfson Brain Imaging Center
The Wolfson Brain Imaging Centre (WBIC) is a research facility attached directly to the Addenbrooke's Hospital Neuro Critical Care Unit and dedicated to imaging function in the injured human brain using Positron Emission Tomography and Magnetic Resonance. Contains links to additional resources on PET, MRI, and the brain.

Journal Articles

Overview

Agid Y, Buzsäki G, Diamond DM et al. 2007. How can drug discovery for psychiatric disorders be improved? Nat. Rev. Drug Discov. 6: 189-201.

Matthews PM, Honey GD, Bullmore ET. 2006. Applications of fMRI in translational medicine and clinical practice. Nat. Rev. Neurosci. 7: 732-744.

Paul Grasby

David SP, Murthy NV, Rabiner EA et al. A functional genetic variation of the serotonin (5-HT) transporter affects 5-HT1A receptor binding in humans. J. Neurosci. 25: 2586-2590. Full Text

Grasby PM. 2002. Imaging the neurochemical brain in health and disease. Clin. Med. 2: 67-73.

Munafò MR, Brown SM, Hariri AR. 2007. Serotonin transporter (5-HTTLPR) genotype and amygdala activation: a meta-analysis. Biol. Psychiatry. Epub ahead of print.

Thomas Klausberger

Jinno S, Klausberger T, Marton LF et al. 2007. Neuronal diversity in GABAergic long-range projections from the hippocampus. J. Neurosci. 27: 8790-8804.

Somogyi P, Klausberger T. 2005. Defined types of cortical interneurone structure space and spike timing in the hippocampus. J. Physiol. 1: 9-26. Full Text

Tukker JJ, Fuentealba P, Hartwich K et al. 2007. Cell type-specific tuning of hippocampal interneuron firing during gamma oscillations in vivo. J. Neurosci. 27: 8184-8189.

Christine Holt

Leung KM, van Horck FP, Lin AC et al. 2006. Asymmetrical beta-actin mRNA translation in growth cones mediates attractive turning to netrin-1. Nat. Neurosci. 9: 1247-1256. Full Text

Lin AC, Holt CE. 2007. Local translation and directional steering in axons. EMBO J. 26: 3729-3736.

Piper M, Anderson R, Dwivedy A, et al. 2006. Signaling mechanisms underlying Slit2-induced collapse of Xenopus retinal growth cones. Neuron 49: 215-228. Full Text

Strochlic L, Dwivedy A, van Horck FP et al. 2007. A role for S1P signalling in axon guidance in the Xenopus visual system. Development. Epub ahead of print.

Karl Friston

Stephan KE, Weiskopf N, Drysdale PM et al. 2007. Comparing hemodynamic models with DCM. Neuroimage. 38: 387-401.

Stephan KE, Penny WD, Marshall JC et al. 2005. Investigating the functional role of callosal connections with dynamic causal models. Ann. N. Y. Acad. Sci. 1064: 16-36.

Stephan KE. 2004. On the role of general system theory for functional neuroimaging. J. Anat. 205: 443-470. Full Text

Eve Johnstone

Hall J, Whalley HC, Job DE et al. 2006. A neuregulin 1 variant associated with abnormal cortical function and psychotic symptoms. Nat. Neurosci. 9:1477-1478.

McIntosh AM, Baig BJ, Hall J et al. 2007. Relationship of catechol-O-methyltransferase variants to brain structure and function in a population at high risk of psychosis. Biol. Psychiatry 15: 1127-1134.

Owens DG, Johnstone EC. 2006. Precursors and prodromata of schizophrenia: findings from the Edinburgh High Risk Study and their literature context. Psychol. Med. 36: 1501-1514.

Hugh Gurling

Gurling HM, Critchley H, Datta SR et al. 2006. Genetic association and brain morphology studies and the chromosome 8p22 pericentriolar material 1 (PCM1) gene in susceptibility to schizophrenia. Arch. Gen Psychiatry 63: 844-854. Full Text

Gurling HM, Kalsi G, Brynjolfson J et al. 2001. Genomewide genetic linkage analysis confirms the presence of susceptibility loci for schizophrenia, on chromosomes 1q32.2, 5q33.2, and 8p21-22 and provides support for linkage to schizophrenia, on chromosomes 11q23.3-24 and 20q12.1-11.23. Am. J. Hum. Genet. 68: 661-673. Full Text

Bhismadev Chakrabarti

Chakrabarti B, Baron-Cohen S. 2006. Empathizing: neurocognitive developmental mechanisms and individual differences. Prog. Brain Res. 156: 403-417.

Chakrabarti B, Kent L, Suckling J et al. 2006. Variations in the human cannabinoid receptor (CNR1) gene modulate striatal responses to happy faces. Eur. J. Neurosci. 23: 1944-1948.

Eric Taylor and Katya Rubia

Rubia K, Smith AB, Taylor E, Brammer M. 2007. Linear age-correlated functional development of right inferior fronto-striato-cerebellar networks during response inhibition and anterior cingulate during error-related processes. Hum. Brain Mapp. 28: 1163-1177.

Rubia K, Overmeyer S, Taylor E et al. 2000. Functional frontalisation with age: mapping neurodevelopmental trajectories with fMRI. Neurosci. Biobehav. Rev. 24: 13-19.

Shaw P, Eckstrand K, Sharp W et al. 2007. Attention-deficit/hyperactivity disorder is characterized by a delay in cortical maturation. Proc. Natl. Acad. Sci. U.S.A. 104: 19649-19654.

David Brooks

Brooks DJ. 2006. PET markers of amyloid deposition and inflammation. Eur. J. Neurol. 13: 305.

Tai YF, Pavese N, Gerhard A et al. 2007. Imaging microglial activation in Huntington's disease. Brain Res. Bull. 72: 148-151.

Paul French

Soloviev VY, Tahir KB, McGinty J et al. 2007. Fluorescence lifetime imaging by using time-gated data acquisition. Appl. Opt. 46: 7384-7391.

Suhling K, French PM, Phillips D. 2005. Time-resolved fluorescence microscopy. Photochem. Photobiol. Sci. 4: 13-22.

Treanor B, Lanigan PM, Kumar S et al. 2006. Microclusters of inhibitory killer immunoglobulin-like receptor signaling at natural killer cell immunological synapses. J. Cell. Biol. 174: 153-161. Full Text

Christopher Johnson

Ahmed M, Briggs MA, Bromidge SM et al. 2005. Bicyclic heteroarylpiperazines as selective brain penetrant 5-HT6 receptor antagonists. Bioorg. Med. Chem. Lett. 15: 4867-4871.

Antoine Triller

Bannai H, Lévi S, Schweizer C et al. 2006. Imaging the lateral diffusion of membrane molecules with quantum dots. Nat. Protoc. 1: 2628-2634.

Dahan M, Lévi S, Luccardini C et al. 2003. Diffusion dynamics of glycine receptors revealed by single-quantum dot tracking. Science 302: 442-445.

Gero Miesenböck

Deisseroth K, Feng G, Majewska AK et al. 2006. Next-generation optical technologies for illuminating genetically targeted brain circuits. J. Neurosci. 26: 10380-10386. Full Text

Shang Y, Claridge-Chang A, Sjulson L et al. 2007. Excitatory local circuits and their implications for olfactory processing in the fly antennal lobe. Cell 128: 601-612.

David Attwell

Attwell D, Laughlin SB. 2001. An energy budget for signaling in the grey matter of the brain. J. Cereb. Blood Flow Metab. 21: 1133-1145. Full Text

Attwell D, Gibb A. 2005. Neuroenergetics and the kinetic design of excitatory synapses. Nat. Rev. Neurosci. 6: 841-849.

Nigel Emptage

Emptage NJ. 2001. Fluorescent imaging in living systems. Curr. Opin. Pharmacol. 1: 521-525.

Emptage NJ, Reid CA, Fine A, Bliss TV. 2003. Optical quantal analysis reveals a presynaptic component of LTP at hippocampal Schaffer-associational synapses. Neuron 38: 797-804. Full Text

Emptage NJ, Bliss TVP, Fine A. 1999. Single synaptic events evoke NMDA receptor-mediated release of calcium from internal stores in hippocampal dendritic spines. Neuron 22: 115-124. Full Text

Tony Ng

Ameer-Beg SM, Peter M, Keppler MD et al. 2005. Dynamic imaging of protein–protein interactions by MP-FLIM. Proc. S.P.I.E. 5700: 152-161.

Festy F, Ameer-Beg SM, Ng T, Suhling K. 2007. Imaging proteins in vivo using fluorescence lifetime microscopy. Mol. Biosyst. 3: 381-391.

Parsons M, Monypenny J, Ameer-Beg SM et al. 2005. Spatially distinct binding of Cdc42 to PAK1 and N-WASP in breast carcinoma cells. Mol. Cell Biol. 25: 1680-1695. Full Text

Ray Dolan

Coricelli G, Dolan RJ, Sirigu A. 2007. Brain, emotion and decision making: the paradigmatic example of regret. Trends Cogn. Sci. 11: 258-265.

O'Doherty JP, Dayan P, Friston K et al. 2003. Temporal difference models and reward-related learning in the human brain. Neuron 38: 329-337. Full Text

Pessiglione M, Seymour B, Flandin G et al. 2006. Dopamine-dependent prediction errors underpin reward-seeking behaviour in humans. Nature 442: 1042-1045.

B. J. Casey

Casey BJ, Galvan A, Hare TA. 2005. Changes in cerebral functional organization during cognitive development. Curr. Opin. Neurobiol. 15: 239-244.

Hare TA, Tottenham N, Davidson MC et al. 2005. Contributions of amygdala and striatal activity in emotion regulation. Biol. Psychiatry 57: 624-632.

Matthew Rushworth

Walton ME, Croxson PL, Behrens TE et al. 2007. Adaptive decision making and value in the anterior cingulate cortex. Neuroimage 36 Suppl 2: T142-T154.

Croxson PL, Johansen-Berg H, Behrens TE et al. 2005. Quantitative investigation of connections of the prefrontal cortex in the human and macaque using probabilistic diffusion tractography. J. Neurosci. 25: 8854-8866. Full Text

Richard Wise

Awad M, Warren JE, Scott SK et al. 2007. A common system for the comprehension and production of narrative speech. J. Neurosci. 27: 11455-11464.

Spitsyna G, Warren JE, Scott SK et al. 2006. Converging language streams in the human temporal lobe. J. Neurosci. 26: 7328-7336. Full Text

Wise RJ. 2003. Language systems in normal and aphasic human subjects: functional imaging studies and inferences from animal studies. Br. Med. Bull. 65: 95-119. Full Text

Sponsors

AstraZeneca Neuroscience

Canon

GlaxoSmithKline

The Gatsby Charitable Foundation

Imperial College London

Intra-Cellular Therapies

Johnson & Johnson

King's College London

Medical Research Council

Siemens

University College London

Wellcome Trust

Wyeth

Paul Broca, the father of modern neurology, was among the first to understand the value of brain lesion studies, which attempt to correlate a neurological disorder with damage to a specific brain region. Based on his studies in the 1860s of aphasic patients, people who had speech or language disorders as a result of brain injury, he was able to identify an area of the brain that is involved in speech production. His most famous patient was nicknamed "Tan" due to his inability to utter any other word, and the area is now known as Broca's area. Other researchers followed his lead, expanding the approach to look for regions that correlate with specific changes in behavior, personality, or cognitive ability. Today, their tools are far more sophisticated: using brain imaging, they can look at how single genes or molecules influence those same human traits.

A personalized approach to neuromedicine is on the horizon, but what will it take to get there?

The potential of these highly sensitive imaging techniques is huge, not only for unravelling the mechanisms of behavior, personality, and cognition, but also for homing in on the causes of disease. Neuroimagers dream of taking their techniques into the clinic and using them to identify healthy individuals who will go on to develop a brain disease, for making more reliable diagnoses and prognoses, or for tailoring therapies to individual patients. "The excitement is back in neuroscience," says Paul Matthews of GlaxoSmithKline and Imperial College London, "For the first time since the nineteenth century, we can look at individual variation again."

This kind of personalized medicine is close, but it is not here yet, which is why scientists in the UK and New York considered it timely to ask, What do we want to see in brain imaging? Researchers working with imaging at the molecular, cellular, organ, and systems levels met to discuss this question at the Royal Institute of British Architects in London on December 3–4, 2007. The meeting, which was organized by the New York Academy of Sciences with Imperial College London, King's College London, University College London, the Royal Institution and GlaxoSmithKline, also marked the launch of the South East England Global Medical Excellence Cluster (GMEC), a collaboration between the three London universities involved in the meeting, the Maudsley and Royal Marsden National Health Service (NHS) Hospital Trusts, GlaxoSmithKline, and more recently, Oxford and Cambridge Universities.

At the meeting, researchers presented their findings, debated the merits of various techniques, and discussed the future of brain imaging research. Research highlights of the conference included the following:

• Early studies on how genes affect the brain show promise but methodological issues remain to be resolved.
• Neuronal growth cones make decisions independently of the cell body.
• At least 21 types of interneuron regulate the storage and retrieval of memories.
• State-of-the-art, biologically informed modelling allows more information to be extracted from neuro-imaging data.
• Fluorescence imaging can be used to detect early cancer.
• Neurotransmitter receptors move laterally through neuronal membranes, allowing for rapid tuning of synapses.
• Some neuronal networks enhance the transmission of information by reducing, rather than increasing, the signal-to-noise ratio.
• Emotional circuits in the adolescent brain may mature faster than those underlying self-control.
• Combined imaging and genetics studies reveal different subtypes of schizophrenia.
• Speech production and speech recognition areas may be linked by multiple nerve tracts in the brain.

Navigate through this eBriefing for a detailed meeting report, a selection of speakers' slides and audio on our multimedia page, and links to related resources.

An array of tools

"Five years ago, I would have said that brain imaging is taking a computerized tomography (CT) scan or a magnetic resonance imaging (MRI) scan of someone's head and looking at the shape of their brain," says Paul Matthews, who helped organize the conference. Today, he says, brain researchers can visualize many levels of brain organization, from molecular events taking place at a synapse or neuronal junction, to the coordinated activity of small neuronal populations, to imaging the whole human brain using the tools of modern neuroradiology: CT, positron emission tomography (PET), and MRI.

If the field is to continue evolving, however, people working at these different scales, with different imaging devices, need to share their expertise, which is why the conference was organized to span scales: genes to molecules, molecules to cells, cells to organs, and organs to systems. The therapeutic application of imaging, for example, can only proceed with a better understanding of the underlying mechanisms of disease.

One way to improve that understanding is to combine genetics and imaging, a field called imaging genomics. While behavioral geneticists study the effects of genetic manipulations on behavior, researchers in imaging genomics look at the effects of such manipulations on brain activity, working on the assumption that this is a more sensitive measure of gene function than behavior.

Gene associations

The last five years have seen an explosion of studies exploring how genes affect brain structure and function, spurring others to employ molecular imaging techniques to look at the proteins that mediate that relationship. Paul Grasby of Imperial College London described this as a very interesting approach, but he sounded a note of caution: many of the associations found to date have vanished or shrunk on re-examination, revealing a number of methodological issues that need to be resolved before any real progress can be made in this field.

It is difficult to draw plausible biological connections between neuroimaging datasets and single genes.

Grasby said that it was sometimes difficult to draw plausible biological connections between neuroimaging datasets and single genes. Even where a connection was relatively clear, as in the link between the val and met variants or alleles of the catechol-o-methyltransferase gene (COMT), dopamine levels in the prefrontal cortex (PFC), and levels of PFC activation, imaging data could be hard to replicate, and there was a risk of false positives. Recent work on the serotonin transporter gene linked polymorphic region, 5-HTTLPR, was a good illustration of this problem, he said.

Theory predicts that the short (s) allele of the 5-HTTLPR polymorphism should be associated with reduced levels of serotonin transporters in the brain, compared to the long (l) allele, and studies have also linked this polymorphism to enhanced activation of a brain structure that plays an important role in emotion, the amygdala, and enhanced responses to emotional facial expressions—links that have led some researchers to propose a role for 5-HTTLPR in depression and anxiety. However, a recent meta-analysis of 14 such studies shows that most of them have lacked statistical power, and that although there does appear to be a genuine effect of the 5-HTTLPR polymorphism on brain function, its size has probably been overestimated.

Grasby's group has recently used PET to explore the mechanisms underlying this effect in more detail. PET can be used to visualize molecular interactions using radioactively labelled tracers that bind to the molecule of interest. In this case, the group used ¹¹C-DASB, a ligand that binds to serotonin transporters. In 81 subjects, they found no effect of variation in 5-HTTLPR on the density of serotonin transporters in the brain, suggesting that if the gene is indeed having an effect on brain function, it is probably not by altering the expression of the transporter itself, but via some other mechanism. Sure enough, they found that variation in 5-HTTLPR affects the binding of 5-HT1A receptors—a finding which they have since replicated. "This is one of the few instances where a PET receptor-genetic association has been replicated," says Grasby.

Discovering new cells

One brain structure that is of particular significance to researchers interested in memory is the hippocampus. Pyramidal cells in this structure, named for their shape, are thought to store information in their electrical firing patterns. More specifically, the coordination in time of the activity of pyramidal cell networks is now considered to be critical to the efficient storage and retrieval of information. In an elegant series of experiments, Thomas Klausberger of Oxford University has used electrical recordings followed by microscopic visualization to probe how those networks operate in time, and he has identified new classes of cells in the process.

The functions of many kinds of interneurons are unknown.

He talked about interneurons, neurons whose own firing patterns contain no memory information, but whose function is to regulate the firing of the pyramidal cells, behaving like "traffic lights" that coordinate activity across pyramidal cell networks. Klausberger said that research to date had focused on a few well-characterized classes of interneuron—basket cells and axo-axonic cells, to name but two—but that there were many more whose role remained unknown, that could nevertheless be characterized by their distinctive molecular machinery.

Klausberger and colleagues recorded the activity of individual interneurons in anaesthetized rats, and after labelling the cells, analyzed the molecules expressed by these cells. Neuro-imaging alone cannot distinguish between different classes of interneuron, he said, even if, as with PET, it were possible to trace individual molecules in the brain. That is because no molecular marker is specific to a given type of interneuron; rather, it is the combination of markers that each carries that identifies them. Another potential problem for imaging studies of the hippocampus is that the axons or main extensions of GABAergic interneurons can extend over long distances. By injecting those axons with tracers, he hopes to be able to discover where they project to, and hence what other brain regions are influenced by hippocampal activity.

Klausberger said that at least 21 types of interneuron were known to date, and he gave an example of one of the most recent to be identified: "ivy cells." Using electrodes to record from these cells, researchers have described their firing patterns in freely moving as well as in anaesthetized rats. In experiments in which the activity of ivy cells and pyramidal cells have been recorded simultaneously, ivy cells have been found to evoke slowly decaying input to pyramidal cells—a very different type of effect from the fast spiking of basket cells, for example. "Ivy cells probably take care of homeostasis and overall excitability of pyramidal cells," Klausberger concluded.

Action in real time

For 20 years Christine Holt of Cambridge University has been creating time-lapse movies of the brain wiring itself. The ability of newborn neurons to grow and reach the correct target in the brain is critical to the function and survival of any organism that has a brain. Holt combines imaging and genetic engineering in a simple vertebrate, the Xenopus tadpole, to tease apart the molecular cues that guide the exploratory tip of a neuron—the growth cone—to its destination, and the protein-manufacturing machinery that gets it there.

One of the most startling findings in this area in the last two decades has been that the growth cone makes its own navigational "decisions," and does not need to consult the cell body, which was previously thought to contain the cell's only equipment for translating messenger RNA (mRNA) into protein. Why should it be advantageous for the cell to replicate this equipment at the tip of the axon, to allow for local protein synthesis? Holt suggests that it provides adaptability, so that the growth cone is ready to respond appropriately and rapidly to any new encounter with a guidance cue.

Preexisting β-actin 3′UTR-Kaede is converted from the green form to the red form after exposure to UV light. Addition of Netrin-1 stimulates translation in the growth cone, as seen by the appearance of new green protein after 10 minutes (top right panel).

Recently, with Kimmy Leung in her lab, she has used a protein called Kaede to visualize new protein synthesis in the growth cone. Kaede was originally derived from stony coral and named after the Japanese maple, whose leaves turn from green to red in autumn, because it can be persuaded to undergo the same color change when exposed to ultraviolet (UV) light. By engineering Xenopus neurons to express Kaede when β-actin—an important element of their protein infrastructure—is made, then zapping it with UV to turn the Kaede red, they have shown that the cells produce new, green, Kaede-tagged β-actin when exposed to attractive guidance cues.

Their current model of axon guidance proposes that both attraction and repulsion involve the asymmetrical translation of mRNAs that regulate the actin cytoskeleton or infrastructure. However, while attractive cues elicit local synthesis of actin (including β-actin) on the near-stimulus side of the growth cone, leading to growth toward that cue, repulsive cues elicit the synthesis of proteins that break down the cytoskeleton, on the near-stimulus side, leading to local collapse and turning away of the growth cone from the cue.

Same data, more information

At higher levels of brain organization, methods for the statistical analysis of imaging data have become more sophisticated in the last decade, allowing researchers to study brain circuits underlying cognition and emotion in terms of their components and interactions, and to visualize widespread disease processes more easily and objectively. Karl Friston of University College London discussed one way of making the most of imaging data, using a method called dynamic causal modelling (DCM).

Dynamic causal modelling is a way of inferring the properties of networks from fMRI data.

He gave the example of functional MRI (fMRI). The signal measured by fMRI, a combination of blood flow, blood volume, and oxygen use, is caused by perturbations in widely distributed neuronal networks—the competing interactions of excitatory and inhibitory inputs to an area, for example, and the impact of a third variable, such as attention, on those interactions. DCM is a way of inferring the properties of those networks from fMRI data.

As Friston explained, the idea is that, as with any experiment, you begin with a series of brain models that you wish to test and select the one that best fits the data. This allows you to make inferences about the brain, by comparing models that do and do not have a key biological component, such as a connection between areas that gets stronger with learning. Depending on how well the model fits the data, it can be improved as more components are added to it, and more information can in turn be gleaned from the data. As this finessing proceeds, it becomes possible to identify novel variables that define the behaviour of the system—variables that would have remained invisible to anyone working within the constraints of the original model.

"Once we've understood the underlying dynamics, we can go and reconstruct them beyond the acuity of the original imaging device," says Friston. The beauty of DCM, he adds, lies in the fact that data derived from different imaging devices can be explained or predicted by the same model, thereby contributing to an increasingly accurate representation of the brain at all levels, from genes to systems.

Linking genes to behavior

The title of this workshop, as the chair Peter McGuffin, professor of psychiatric genetics at King's College London pointed out, was a little misleading. It turned out to be more about genes to people than genes to molecules, focusing as it did on the contribution of genes to neuropsychiatric disorders or traits that might potentially be related to those disorders.

Hugh Gurling of University College London kicked off with an overview of the genetic analysis of one of the most devastating neuropsychiatric disorders, schizophrenia. Although this disease is highly heritable, affected members of the same family can show different symptom patterns, suggesting that there are unlikely to be single genes for single symptoms—a gene for auditory hallucinations, for example. These facts have led to a prevailing theory of schizophrenia, according to which many genes make small contributions to the disease.

A variant of PCM1 is associated with high risk for schizophrenia.

Recent genome-wide association studies have revealed a handful of these genes. Gurling gave the example of one in particular, called pericentriolar material 1 (PCM1). His group sampled some of the largest affected families in the world—some of which have as many as 15 affected members—to look for associations between disease status and variants of this gene. They also compared the DNA of more than 600 unrelated people with schizophrenia in London, with that of a similar number of healthy controls. They found a variant of PCM1 that was associated with high risk for the disease.

In collaboration with Richard Frackowiak's group at University College London, who scanned the brains of two groups of schizophrenic patients—one which had the high risk PCM1 variant, and one which did not—Gurling's team went on to show that the two groups showed tissue loss in different parts of the brain. Those with the high risk PCM1 variant had damage in the orbitofrontal cortex, an area above and behind the eyes, while in the second group the damage was located further back in the brain, at the temporal poles. Gurling said this pointed to the existence of different subtypes of schizophrenia, which should ultimately be treated using different pharmacological strategies.

All in the family

Since 1994, Eve Johnstone, a clinical psychiatrist at the University of Edinburgh, has been following a cohort of people who are considered at high risk of developing schizophrenia because they are related to someone who is affected. The objective of the Edinburgh High Risk Study (EHRS), as the project is called, is to address the neurodevelopmental hypothesis of the disease—that is, that some abnormality of brain development gives rise to the symptoms, which nevertheless do not manifest themselves until early adulthood. The average age of onset is 23 in men, and 27 in women.

The Edinburgh team has shown that brain differences as revealed by imaging are much greater between unaffected members of psychotic and non-psychotic families, than between affected and unaffected individuals within psychotic families—though in terms of clinical symptoms, the situation is reversed. If many genes contribute to schizophrenia in small ways, as many researchers believe, large samples may be needed in order to identify those multiple, small effects. However, Johnstone thinks that relatively small samples can also be revealing about the ways in which genes combine their effects to produce schizophrenia, particularly if those samples consist of at-risk individuals.

"We believe that what people inherit is not schizophrenia, but a state of vulnerability."

Of 146 high-risk individuals in the study, she reported that 20 had been diagnosed as schizophrenic while 66 still had no psychotic symptoms. Most intriguingly, the remaining 60 had reported mild psychotic symptoms. These were not enough to warrant a diagnosis of schizophrenia, nor to stop them living relatively normal lives, yet scans revealed brain abnormalities in this group. "We believe that what people inherit is not schizophrenia per se, but a state of vulnerability," Johnstone concludes.

The researchers have now identified those participants of the EHRS who are carriers of high risk variants of candidate genes for the disease, including neuregulin-1 (NRG1) and catechol-o-methyltransferase (COMT). Their findings suggest that by highlighting brain changes associated with those high risk alleles, imaging could potentially provide a marker of preclinical disease—in other words, a way of identifying people in that vulnerable state, who with the right care could be prevented from developing the full-blown disease.

A span of attentions

Eric Taylor and Katya Rubia of King's College London suspect that attention-deficit/hyperactivity disorder (ADHD) represents one extreme of a continuum. According to Taylor, twin studies have revealed that the disorder is highly heritable, but what is inherited is not ADHD as such, but a sliding scale of genetic predisposition to ADHD, depending on the number and/or combination of risk gene variants a person carries. This predisposition interacts with the child's environment to produce a variable trait. Although genome-wide scans have revealed no obvious genetic associations with ADHD, candidate gene studies have implicated several genes related to the neurotransmitter dopamine—notably, the dopamine transporter DAT1 and the dopamine receptor DRD4.

Rubia said that studies are now combining imaging and genetics to look at the functions of these genes in the brain, and that a new model of the disorder is emerging, according to which ADHD is explained in terms of maturational delay. She pointed to a recent U.S. study that found that in children diagnosed with ADHD, the age at which the brain's cortex reaches its peak thickness is delayed by about three years, compared to normal children. Rubia explained how this phenomenon could be mediated by the dopamine receptor DRD4. In the future, she said, children with ADHD could potentially be treated with the drug that the combination of their genetic make-up (genotype) and pattern of brain abnormality, as revealed by scans, indicates will be most effective for them.

The eyes have it

Bhismadev Chakrabarti of Cambridge University discussed studies of emotion perception from facial expressions using fMRI and a technique that is particularly relevant to autism: gaze tracking. This relates to the observation that healthy people tend to look at happy faces for longer than they look at sad ones, and in general they focus particularly on the eye area. Autistic people, by contrast, spend relatively little time looking at the eyes, and perhaps because the eyes are so expressive of emotion, it has been suggested that this may be linked with their poor social skills and relative lack of empathy (the ability to feel another person's emotion). By studying gaze tracking behavior with imaging and genetic analysis, Chakrabarti hopes to discover the molecular underpinnings of empathy.

He is particularly interested in a gene called CNR1, which encodes a cannabinoid receptor in the brain. The human brain produces its own cannabinoids, which are recognized by cannabinoid receptors, and this system is known to work in concert with the brain's reward circuit, which in turn is mediated by the neurotransmitter dopamine. A happy face can be considered a form of social reward, hence Chakrabarti's interest in CNR1. In a series of small studies of healthy volunteers, he and his colleagues found that variants of CNR1 are associated with differences in the brain's response to happy faces, as measured by fMRI, as well as to the time people spend gazing at those faces and to trait measures of empathy. Chakrabarti stressed that CNR1 "is not the gene for happy face perception"; however the convergent findings suggest that, in principle, it should be possible to parse out the genetic architecture of complex traits.

Summarizing the workshop, McGuffin said that a problem that has to be solved in the relatively young field of imaging genomics is the lack of standardization of experimental protocols and imaging techniques, making studies addressing the same question hard to compare. Another potential obstacle is the relatively small sample sizes used in such studies, which typically involve tens or hundreds of volunteers, as compared to genetic association studies which tend to recruit thousands. He also raised the issue of the clinical relevance of some of these findings, pointing to Rubia's intriguing suggestion that though conduct disorder and ADHD often occur in the same individuals, imaging could be used to tease them apart, and that pharmacological therapies could be tailored to individuals on the basis of their imaging/genotype interactions.

How to see a single molecule

This workshop, which was chaired by Shitij Kapur, a psychiatrist and schizophrenia expert at King's College London, focused on three types of imaging: PET, fluorescence, and the lesser known technology of quantum dots. All three methods allow researchers to visualize events taking place in the brain at the scale of single molecules.

PET, fluorescence, and quantum dots offer ways to visualize events at the molecular scale.

PET provides a means of measuring three-dimensional distributions of radioactively labelled molecules in the body. It can be used to image the time course of those molecules within tissues, enabling researchers to probe molecular interactions and pathways, and it is sensitive to picomolar (10⁻¹² moles) concentrations. For these reasons, David Brooks of Imperial College London uses PET to try to understand the role of inflammation in brain disease. Inflammation is part of the body's defense against injury and infection, but if it gets out of hand it can exacerbate the underlying problem. Scientists now know that it contributes not only to the prototypic central nervous system (CNS) inflammatory disease, multiple sclerosis, but also to many others, including stroke and Alzheimer's disease.

Brooks is interested in a type of immune cell found in the brain called microglia. The normal function of microglia is not well-understood, but their presence in an active form seems to reflect ongoing injury or disease. When microglia are activated, the expression of a brain receptor called peripheral benzodiazepine-binding site (PBBS) increases. Using a PET ligand or molecule that binds to PBBS as a marker of microglial activity, Brooks's team has found that microglia are active in a wide range of neurological diseases. The PET data unfortunately cannot distinguish between various theories as to what the active microglia are doing in those diseases—for example, whether they are mediating the inflammation or scavenging inflammatory signals—but they do at least suggest a new therapeutic avenue, since antibiotics already exist that suppress microglial activity.

Imaging drug-target interactions

One of the problems the pharmaceutical industry faces is the high risk of failure of drug candidates that act on novel targets. According to medicinal chemist Chris Johnson of GlaxoSmithKline (GSK), the first thing you would like to know before embarking on a potentially long and expensive drug development program is whether your putative drug actually hits its presumed target, and in suitably high concentrations—a particular problem for CNS drugs, which must cross the blood-brain barrier. Once that has been ascertained, you would like to confirm that the target is relevant to the disease in question. PET can be useful for investigating such drug–target interactions at the molecular level. However, it is only as good as the radioactive ligands it employs, and suitable ligands are not always available.

In these pig brain PET images, the specific saturable signal in the striatum is due to binding of PET ligand to 5-HT6 receptors. The cortical signal blocked by ketanserin: 5HT2A receptors.

To illustrate how the pharmaceutical industry gets around this problem, Johnson gave the example of a drug GSK recently developed, which blocks a receptor for the neurotransmitter serotonin, called 5-HT6 (14 subtypes of serotonin or 5-HT receptor have been identified to date). Studies in rodents have shown that this 5-HT6 antagonist has pro-cognitive effects, but at the time that GSK started work on it, 5-HT6 was a novel clinical target. Thus the drug had to proceed through the entire development process, from preclinical or animal tests to a clinical testing phase. By the time the team was ready to enter human trials with their drug candidate, they still had no suitable PET ligand for it, so instead they used a ligand for another 5-HT receptor, 5-HT2A.

The animal studies had suggested that the drug must bind to 80% of the 5-HT₆ receptors to produce an improvement in learning and memory. The researchers knew that it had a lower affinity for 5-HT2A than for 5-HT6, so when they found that it occupied 60% of 5-HT2 receptors in humans, they were able to predict that it would probably bind to around 95% of 5-HT6 receptors—in other words, enough to produce a clinical effect. This gave them the confidence to proceed with the programme until they were able to confirm their findings with a specific 5-HT6 ligand that was developed later. Commenting on the case, Kapur said it demonstrated that in the real world, pharmaceutical companies could not wait for the best ligands to come along, and that the use of surrogate markers could be justified.

FLIM and FRET

In fluorescence imaging, the label is a fluorophore that emits light when stimulated with a laser pulse. Its simplest and probably most common application to date is the localization of cells or subcellular structures genetically engineered to express that fluorescent label. But as physicist Paul French of Imperial College London pointed out, there are many other things you can do with it. His group is developing technologies that exploit different properties of fluorescence to extract more information from biological tissues. One example is the half-life of fluorescence decay, which he says is often a more sensitive and robust measure than intensity or wavelength.

FLIM senses the biological environment by detecting the autofluorescence of molecules in the body.

Decay is the principle behind fluorescence lifetime imaging (FLIM). Since its half-life is fixed, differences in the visualization of the decaying fluorescent signal must arise from environmental perturbations. FLIM can therefore provide a kind of sensor of a biological environment and, since the body's own molecules have the capacity to fluoresce (this is autofluorescence), it can do so without the need to introduce chemicals into the body. Using an ultra-fast laser, a short pulse of light is shone on a sample. A series of timed images is then taken of that sample, using a microscope, and each pixel of each image is analyzed to construct a decay profile. The profile of an experimental sample can then be compared to that of a control.

Another technique known as FRET, for Förster Resonant Energy Transfer, exploits the observation that two fluorescent molecules coming close to one another can transfer energy between them, amplifying the fluorescence signal. These techniques are now enabling researchers to look at molecular interactions that take place in under a second. One way in which FLIM and FRET have been applied is in the detection of immune cells signalling to one another. Another is in the detection of neoplasia or potentially cancerous changes in human tissue biopsies. However, French warned that for biological imaging to be useful in vivo, significant advances need to be made in the speed and sensitivity of these techniques, as well as in data processing.

Quantum dots

Neurobiologist Antoine Triller of the French medical research agency INSERM and the Ecole Normale Supérieure in Paris is at the forefront of research into the trafficking of neurotransmitter receptors. Where two neurons form a synapse or junction, the two communicate via the transfer of neurotransmitter molecules across the synapse, which bind to receptors in the membrane of the post-synaptic neuron. Synaptic plasticity, or the modification of the strength of that connection—thought to be the basis of learning and memory—is partly the result of changing receptor numbers in the post-synaptic nerve terminal, due to receptor trafficking.

Receptors temporarily become involved in synaptic transmission before moving away from the synapse.

Using quantum dot technology, which allows him to track single molecules in cell membranes in close to real time, Triller has made some novel insights into these processes. He finds that receptors are constantly moving laterally through the membrane of the post-synaptic neuron, temporarily becoming involved in synaptic transmission before moving away from the synapse again, and he suggests that this might be the mechanism by which the brain is able to rapidly tune the number of receptors at a synapse. Receptors may even be able to serve more than one synapse in this way.

His findings have potentially important implications, suggesting as they do that the variability of receptor number at synapses depends not only on the recycling of receptors between the cell interior and its outer membrane, but also on their diffusion within the membrane and interactions with proteins which give that membrane its structure. That in turn has implications for future therapies, since much of modern neuropharmacology depends on altering the number or affinity of receptors for the neurotransmitters that bind to them. As Kapur said, Triller's work indicates that if the lateral diffusion of receptors could be harnessed, it might be possible to alter synaptic efficiency by changing only the distribution of receptors, rather than their overall number. "Would this provide a sufficiently wide therapeutic window?" Kapur wondered. "These and other questions still need to be answered, nevertheless the clinical possibilites are exciting."

Summarizing the workshop, Kapur said there is a real sense emerging that basic researchers need imaging techniques with greater sensitivity and resolution than are currently available. He also thought there was scope for better chemistry, to provide PET users with the ligands they need, at the time they need them and in sufficient quantities, and that this situation could be improved by better communication between industry and academia. PET will remain a valuable imaging tool, he says, since it is the only one available today that is sensitive to picomolar concentrations of neurotransmitters—the sensitivity needed to watch a synapse in action. However it is expensive, and costs might prevent it become widely available in a clinical context.

The meaning of noise

Gero Miesenböck of Oxford University studies the olfactory system of a simple animal model, the fruit fly Drosophila melanogaster. Specifically, he is interested in how different populations of olfactory neurons contribute to the processing of different odors, and he measures this using light and genetically encoded sensors that are expressed only in target subpopulations. Cochair Nigel Emptage of Oxford University said that Miesenböck's approach reveals different things from an electrophysiological study, in which the outputs of single neurons are recorded, because it enables one to see how groups of neurons work together.

The image shows a circuit diagram of the three principal populations of neurons—olfactory receptor neurons, projection neurons, and local neurons in the olfactory system of flies. The expression of the genetically encoded activity sensor synapto-pHluorin in each of the three classes of neurons opens a window on how olfactory information flows through this circuit in the living brain.

Miesenböck's research has indeed thrown up some surprising findings. Counter-intuitively, since one might expect that the various parts of the nervous system would work to reduce noise in the system as a whole, he has found that certain neuronal circuits actually add noise to the sensory signal. This phenomenon, known as "stochastic resonance," is well known to physicists and engineers. It turns out that the addition of background noise can actually enhance the transmission of a signal from one neuron to the next. Though people have suggested stochastic resonance may play a role in the brain, Miesenböck says, "This may be the first empirical example of a circuit that is noisy by design."

"The idea seems to translate across systems," says Paul Matthews of GlaxoSmithKline and Imperial College London, who was impressed by Miesenböck's work. "It struck me that this ... might be a clue as to why there's so much high background activity in the brain, and why we pay the energy price for this."

How to interpret an MRI

Ever since fMRI revolutionized the field of noninvasive brain imaging in the early 1990s, researchers have wondered about the exact relationship between the blood-oxygen-level-dependent (BOLD) signal that it actually measures—a combination of blood flow, blood volume, and oxygen use—and the underlying neural activity that interests them. Most have assumed, not unreasonably, that the two are closely related, and that the greater the blood flow and oxygen consumption, the harder a region is working. But there have been dissenters who argue that the colorful spots that show up on a brain scan are far from being a reliable gauge of brain activity.

About half of the brain's energy consumption supports the generation of action potentials.

Among them is David Attwell of University College London. He has calculated the energy that is required for the generation of action potentials—the "firing" of neurons—in different parts of the nervous system, in different species and during different tasks. One of his most striking findings is that in rodents, about half of the brain's energy consumption supports the generation of action potentials, a figure which falls to only 13% in primates. So what other activities consume so much energy in the primate brain? Attwell believes a lot of it goes into reversing the ion movements that produce currents at the synapse (the junction between two neurons) and give rise to action potentials, in order to bring the system back to its ground state. If he is right, then BOLD may be telling researchers more about the synaptic input to a cortical area than about the firing of the principal neurons within that area.

Nigel Emptage presented "a new form of heresy." He studies long-term potentiation (LTP), the prolonged activation of synapses that is widely believed to be the mechanism of synaptic plasticity and hence, the basis of learning and memory. The basic principle of synaptic transmission is that neurotransmitter molecules released on the presynaptic side of the junction bind to receptors on the post-synaptic side, triggering a release of ions in that neuron, including calcium (Ca²⁺) ions. Emptage measures synaptic transmission by using fluorescence microscopy to image the flow of Ca²⁺ ions into pre- and post-synaptic neurons, in living slices of hippocampus (a brain structure that plays a critical role in memory). He now has evidence that a receptor which binds the excitatory neurotransmitter NMDA (N-methyl-D-aspartic acid) is also located on the presynaptic terminal, rather than exclusively on the post-synaptic side, as is widely believed. This suggests that NMDA receptors act to modulate neurotransmitter release. Emptage suspects that their function may be to augment synaptic activity at "the all-important theta frequency," at which LTP is readily induced.

Fluorescent indicators

Tony Ng of King's College London gave an elegant demonstration of the power of fluorescence imaging. Like Paul French [see Workshop 2], he uses fluorescence lifetime imaging (FLIM) and Förster resonant energy transfer (FRET) to visualize rapid molecular events, but in his case he uses them to image protein–protein interactions. The idea behind FLIM is that fluorescence decays over a predictable timescale, and monitoring that decay provides a sensor of a biological environment. FRET exploits the fact that two fluorescent molecules coming close to one another—within a "nano-proximity"—can transfer energy between them, amplifying the fluorescence signal.

Ng and colleagues have used these techniques in tissue biopsies taken from cancer patients, to detect molecular-level changes that are associated with increased risk of metastases or secondary tumors and shorter survival. Emptage described the implications of their work as exciting, since it becomes theoretically possible to investigate tissue-specific and patient-specific protein–protein interactions. That in turn means that pharmacological therapy could be tailored to patients, on the basis of the particular molecular interactions observed in their tissues. The effects of those therapies could also be monitored at the molecular level.

Summing up the workshop, Emptage said that researchers would like to be able to look at the consequences of behaviorally relevant stimuli on a cell, but that imaging techniques rarely allow that. In the context of his own work, he would like to be able to image many synapses simultaneously, rather than just one, and to be able to obtain as good quality images in vivo as in vitro. For these kinds of advances to be made, he said, a "sea change" in fluorescence technology is needed.

The basis of decision-making

How the brain weighs different options and comes to a decision is one of the hottest topics in neuroscience at the moment, and the basis of the new field of neuroeconomics. Ray Dolan of University College London described how his team, in collaboration with researchers at UCL's Gatsby Computational Neuroscience Unit, has been studying the brain circuits that contribute to decision-making, using mathematical algorithms and fMRI in tasks that require people to discover a hidden value which governs the choice that will deliver the greatest reward.

When making choices, the human brain seems to follow certain basic rules. Having learned to associate a certain cue with a certain reward, for example, it monitors discrepancies between the predicted reward and the actual reward, or lack of it, given in response to that cue. That discrepancy, or prediction error, manifests itself as a brain signal that is measurable with fMRI, and that alters in size and in time as the brain adapts to an evolving situation. The theory of reward learning is based on this capacity of the brain to continually update the value of an action and the probability that it will bring a reward.

The above images show how brain activity differs when subjects make choice (a) driven primarily by an imperative to exploit what currently seems to be the best option (b) versus sampling an alternative option that is less certain but potentially more valuable (i.e., exploration).

Research has shown that the neurotransmitter dopamine is involved in the brain's response to reward, and Dolan's group set out to test its role in decision-making. They scanned the brains of people engaged in a choice task, but with an extra twist: some of them had received a low dose of a dopamine-boosting agent, levodopa (L-DOPA), while others had received haloperidol, a dopamine-blocking agent. They found that dopamine-rich brain circuits responded differently to the task in the two groups, and that the L-DOPA group learned better than those who received haloperidol. Dolan concludes that dopamine provides a "teaching signal" that prompts the brain to respond appropriately to a prediction error.

A multimodal approach

Continuing on the theme of choice and action selection, Matthew Rushworth of Oxford University took a more methodological line, making the point that much could be learned by combining imaging modalities. For example, fMRI could be combined with tractography, which highlights white matter nerve fibre tracts, to reveal anatomically distinct regions that interact to give rise to a certain behavior or cognitive skill. Alternatively, fMRI could be combined with a so-called interference methodology such as transcranial magnetic stimulation (TMS)−in which rapidly changing magnetic fields temporarily and reversibly modulate activity in a targeted brain region−to look at how that region interacts with remote regions.

Rushworth described how this combined technique had been used to show that changes in activity in the frontal eye field, an area at the top of the brain involved in generating eye movements, brought about changes in early parts of the pathway for processing visual signals. He also said that deep brain stimulation−a technique that involves the surgical implantation of electrodes, which is currently used to treat the symptoms of Parkinson's and other diseases−could provide a useful interference methodology in combination with imaging, as long as researchers could overcome the methodological problems of mixing electrodes, the human brain, and the strong magnetic fields that circulate in a MRI scanner.

The terrible teens

B. J. Casey, a developmental psychobiologist at Weill-Cornell Medical College, raised an important point which prompted much discussion: adolescent brains differ from the brains of adults and children. This is one area in which neuroscience has already had an impact on public policy, because teenagers are more likely than other age groups to be involved in violence, substance abuse, and other problematic behaviors, raising the question of how many rights and responsibilities they should reasonably be accorded. Casey said that functional neuroimaging is throwing light on these brain differences, but that researchers need to be aware of potential pitfalls if they are to come to meaningful conclusions.

Different parts of the adolescent brain may be at different stages of development.

The first is a technical issue: the fMRI signal may differ across developmental periods. Differences in heart rate and respiration between childhood and adulthood could create biological noise, for example. Second, the adolescent brain is not merely halfway between a child's and an adult's brain; different parts of the adolescent brain may be at different stages of development. The limbic system, for example, which mediates emotion, may be more developed than control circuits which inhibit emotional responses in the adult. This could explain why studies have shown a heightened response to emotional stimuli in teenagers, compared to adults and children. To date, Casey says, the imaging community has been too focused on the cortex, the outer layer of the brain in which those control circuits lie, rather than deeper, limbic structures, to see these differences. There may also be individual differences in brain maturation, across adolescents of the same age. She has evidence, for example, that the limbic response to fearful faces habituates or lessens on repeated exposure in some teenagers, but not in others. These and other findings have implications for our understanding of impulse control in adolescents, and on the way that understanding is translated into policy: "Adolescents know what's right and wrong, but they can't help themselves," she says.

Finally, language expert Richard Wise of Imperial College London described how understanding of the organization of language in the brain had moved on since the 19th century, when Broca and Wernicke gave their names to the brain areas thought to mediate production and comprehension of language respectively. In particular, he outlined how functional imaging and animal studies had contributed to the field in the past three decades. In the past, for example, Broca's and Wernicke's areas were thought to be connected exclusively by a nerve fibre tract called the arcuate fasciculus. Wise said this theory should now be abandoned, since studies based on tractography are revealing new routes that connect the two language areas.

He also warned that researchers using imaging to understand cognitive components of language should be aware of potentially confounding signals generated by the motor and somatosensory components—that is, the laryngeal and other movements needed to actually generate language, and the feedback sensations they elicit. He encouraged researchers to limit themselves to a functional and anatomical understanding of language in the brain, arguing that there were now so many competing theoretical models of language coming from computational neuroscience and other disciplines, that none of these could yet be helpful in interpreting fMRI data.

Will imaging become a practical tool in the clinic?

To what extent is researchers' understanding of the human brain constrained by limitations on the resolution and sensitivity of imaging devices?

To what extent can good statistical modelling improve the data that imaging devices are currently capable of delivering?

In studies of the interaction between brain function and genetics, what is the optimal sample size?

Can researchers agree on experimental protocols that will permit comparisons across species, genders and age groups?

Is gender a factor in the relationship between genes and brain function?

How important is it to know the developmental history of a brain, when imaging that brain?

How should imagers allow for the tremendous variability, not only in the brains of patients with neuropsychiatric diseases, but also in those of healthy volunteers?

What will understanding how pyramidal cells work in the hippocampus reveal about how that structure stores memories?

How many classes of interneurons are there in all?