You are on page 1of 14

Nature Reviews Drug Discovery | AOP, published online 7 April 2006; doi:10.

1038/nrd2027 REVIEWS

A role for fMRI in optimizing CNS


drug development
David Borsook*, Lino Becerra* and Richard Hargreaves‡
Abstract | Drug development today needs to balance agility, speed and risk in defining the
probability of success for molecules, mechanisms and therapeutic concepts. New techniques
in functional magnetic resonance imaging (fMRI) promise to be part of a sequence that could
transform drug development for disorders of the central nervous system (CNS) by examining
brain systems and their functional activation dynamically. The brain is complex and multiple
transmitters and intersecting brain circuits are implicated in many CNS disorders. CNS
therapeutics are designed against specific CNS targets, many of which are unprecedented.
The challenge is to reveal the functional consequences of these interactions to assess
therapeutic potential. fMRI can help optimize CNS drug discovery by providing a key metric
that can increase confidence in early decision-making, thereby improving success rates and
reducing risk, development times and costs of drug development.

Functional magnetic
There is a clear need to fundamentally change drug fMRI of the brain measures changes in the blood
resonance imaging discovery paradigms to meet today’s healthcare oxygenation levels in microcirculation. This provides
An MRI technique that uses needs1–3. The past 20 years, which included the ‘decade an indirect measure of neural activity (BOXES 1 , 2) .
metabolic-induced capillary of the brain’ (the 1990s), has seen the development of It can be performed in awake animals and humans, and
blood flow changes caused by
few novel therapeutic agents for CNS disorders. Indeed, provides information about neural circuit activity in
neuronal activity to produce
images reflecting such activity. many of the drugs coming to market have been derived response to specific, reproducible and well-character-
Activation can be correlated from the pharmacology of older agents with proven ized stimuli that can serve as a ‘fingerprint’ of specific
with brain structures as clinical utility with the aim of improving potency, function. Such fingerprinting can be applied to neural
determined by these images. tolerability or ease of administration. Drug develop- circuits involved in drug action or disease states. The
ment in the CNS domain has to overcome a number big question is whether fMRI can help provide the
of challenging barriers, not the least of which is the focus and level of specific functionality to guide CNS
inaccessibility of the brain itself, when evaluating the drug development. This review critically evaluates the
consequences of drug–target interactions. The lack of potential of fMRI at different stages during the pre-
facile surrogate endpoints for potential CNS therapeu- clinical-to-clinical drug development process (BOX 3)
tics means that making the right choice of targets and to assess its potential to engender a paradigm shift in
molecules for development is crucial4. For the CNS, the CNS drug discovery by providing early knowledge that
*Imaging Center for problem is also confounded by the paucity of processes validates therapeutic concepts and rationally endorses
Drug Development (ICD), and technologies that can provide a true translational candidate molecule choices.
Mclean Hospital Department
of Psychiatry, Massachusetts
bridge between preclinical studies and subsequent clin-
General Hospital Department ical testing and evaluation5. A functional approach such Brain circuits — functional arrays
of Radiology, Harvard as functional magnetic resonance imaging (fMRI) provides a Gene-expression signatures have helped identify patterns
Medical School. ‘systems neuroscience’ evaluation of the circuitry that of cellular pathway control and dysregulation, thereby

Imaging, Merck & Co. Inc.,
underlies the behavioural effects of a drug, independ- opening the possibility of designing targeted thera-
West Point, Pennsylvania
19486, USA. ent of its specific biochemical mechanism of action. pies that match disease mechanisms more precisely6.
e-mails: dborsook@mclean. Indeed, many CNS drugs have multiple mechanisms of Similarly, an understanding of the functional neural
harvard.edu; lbecerra@ action and can vary in efficacy across the CNS targets circuits and the interactions between them that drive
mclean.harvard.edu; with which they interact. fMRI monitors the combined CNS behaviour in health and disease could lead to the
richard_hargreaves@
merck.com
or integrated effect of these interactions across multiple development of novel circuit-specific drugs, rather than
doi:10.1038/nrd2027 brain systems and thereby reflects activity of the true today’s pharmacologically selective agents. This could
Published online 7 April 2006 neural circuitry that drives behaviour. provide a rationale for the selection of combinations of

NATURE REVIEWS | DRUG DISCOVERY ADVANCE ONLINE PUBLICATION | 1


REVIEWS

neural circuits, including regions involved in emotion


Box 1 | Functional MRI — the basics
or cognition, have been harder to evaluate. However,
Neuronal activity triggers an increase in blood supply to the surrounding capillary beds novel insights into these systems that were previously
that overcompensates for neuronal oxygen extraction, causing a relative change in not possible in humans have been gained using fMRI.
their oxy/deoxyhaemoglobin status; oxyhaemoglobin concentration increases while Chemical circuits (glutamate, GABA (γ-aminobutyric
deoxyhaemoglobin decreases (see FIGURE). Because oxyhaemoglobin is less acid), dopaminergic, cholinergic, serotonergic and
paramagnetic than deoxyhaemoglobin the MR signal intensities in the area around
noradrenergic) can also be construed as complex because
these capillary beds change; increased concentrations of oxyhaemoglobin result in
higher MR intensities. A net increase in MR signal intensity is therefore detected in areas
they generally project to many neuroanatomical brain
that become active and depends on the magnitude of the blood-flow changes evoked. regions8–10. Structural systems (such as neurons and glia)
For this reason, it has been defined as blood oxygenated level dependent (BOLD) fMRI. define functional brain activity by determining the net
The increased blood flow (which correlates with increased neural activity152) produces balance of numerous neurotransmitter/neuromodulators
fMRI signal changes of the order of 0.5–5%. that are found on cell bodies and terminals. Functional
neurological circuits can also be sub-categorized because
O2 they often differ in their activity in health and disease.
In the early stages of CNS drug development, the deliv-
Normal
activity ery of potential therapeutics and their actions are often
Normal flow studied in healthy individuals with normal neurocir-
cuitry. Although this can have great value in ensuring
O2 target engagement (using technologies such as nuclear
molecular imaging, positron-emission tomography (PET) and
Change
in single-photon-emission computed tomography (SPECT)) and
activity proof of biochemical mechanism, these studies usually
Increased flow = increased MRI signal
create baselines for the study of disease states and their
O2 Oxygen Oxyhaemoglobin therapeutic modulation — the holy grail of targeted CNS
Relative signal intensity Deoxyhaemoglobin therapies. In some cases, it is difficult to recapitulate key
aspects of psychiatric or neurological disease in healthy
therapies, which could improve the treatment of com- individuals, but their use is a rapid step towards the
plex CNS disorders. Understanding functional circuits development of paradigms for patients.
as ‘arrays’ or units that contribute to behaviour might Activation of specific neuroanatomical circuitry
provide novel signatures that can be used as endpoints within the brain defines the behavioural consequences
to define CNS disease processes or the effects of drugs of a drug effect or disease process. The patterns of
on the brain in health and disease. neuroactivation and their changes are therefore poten-
Functional circuits can be defined as simple or com- tial markers of disease state and therapeutic efficacy.
plex. It has been relatively easy to image simple circuits Targeting the pathological alterations in CNS process-
such as sensory systems (for example, vision and pain ing involved in CNS dysfunction in humans allows for
Molecular imaging
detection), and to define their functionality7. Complex a top-down approach to the evaluation of drug effects
An imaging technique in
which cellular/molecular
processes have been tagged
in such a way that they can
Box 2 | Functional MRI — the experiment
be non-invasively imaged. In an fMRI experiment, subjects are
inserted into the MRI scanner and a
Positron-emission
tomography
series of stimuli applied. Typically, the Statistical
(PET). A dual-photon nuclear stimuli are presented as ‘on’ and ‘off’ map
imaging technique in which periods (see FIGURE). During the
radioactive tracers are stimulus presentation several images of
administered in non- the brain are acquired. One approach to
pharmacologically active detecting activation consists of
doses to subjects and images correlating the temporal profile of the Time
are created that reveal brain stimulus with the temporal evolution of
blood flow, glucose
Pseudo-
each ‘point’ (voxel) in the image. Voxels colour
metabolism or fractional
that are highly correlated with the statistical
receptor binding by drugs. map over
stimulus receive a high statistical score.
fMRI anatomical
Single-photon-emission An image is created that reflects the signal image
computed tomography statistical score for each voxel, a
(SPECT). A nuclear imaging statistical map. The statistical map,
technique in which radioactive however, lacks any anatomical Stimulus
tracers generating single information that could define the Off On Off On Off On Off
photons of a specific energy structures with significant activation.
are administered to subjects to
Statistical maps are colour coded and overlaid on anatomical images (usually, statistical maps are displayed in such a
produce images. SPECT can
give information about blood
way that only voxels with a significance larger than a threshold are included and the rest are not considered/displayed).
flow to tissues, molecular The resulting combined image allows the identification of voxels that are significantly activated and the central nervous
targets, and chemical reactions system substrate that they belong to. fMRI resolution for standard clinical 3-Tesla scanners is typically about 10 mm3
(metabolism) in the body. (and could be as low as 1 mm3) and for 7-Tesla animal scanners as low as 0.05 mm3.

2 | ADVANCE ONLINE PUBLICATION www.nature.com/reviews/drugdisc


REVIEWS

Box 3 | Potential application of fMRI across drug development

Lead optimization Life-cycle management


Off-target CNS effects Novel CNS indications
CNS ‘therapeutic window’ Expanded use

Approved drugs
Preclinical Clinical
Drug discovery
development development
Failed, safe
CNS drugs

Target validation Early proof of CNS activity


Experimental assays Early support for concept Novel uses for safe molecules
Disease models Therapeutic index Recover development costs
Transgenics Optimized dosing Mechanistically
Responder identification unprecedented targets
Patient stratification Breakthrough neuroscience

The figure shows the classic progress of the components of drug development (ovals) and the potential applications of
fMRI (boxes) at each point along this development line.

on disease states11–13. The circuits that control complex of fMRI in drug development22–24. Moreover, it is now
behaviours have been approached using large-scale widely accepted that pain is not a unitary sensation
genomic profiling approaches to define phenotypes but, in its chronic form especially, has cognitive, moti-
associated with altered neural circuits14. Clearly the vational and emotional valences. Chronic pain states
reverse process — defining circuits and then linking therefore provide a clinical model with co-morbid
back to genomics — is an alternative approach (FIG. 1). psychophysical aspects. These can be studied to pro-
Detailed mapping of brain circuits allows for the sys- vide insight into other CNS diseases, such as anxiety
tematic evaluation of functional reorganization in, for and depression, because there is a high incidence of
example, the neural plasticity seen following stroke15, pain following the onset of depression 25 and a high
dyslexia16,17, alterations induced by drugs (such as tha- incidence of depression in chronic pain26. It is impor-
lidomide-induced dysmelia18), reorganization of visual tant when considering fMRI as a top-down transla-
processing in the condition macular degeneration19, and tional technology, that it can be used to evaluate and
chronic pain20,21. Functional neuroimaging provides an focus selection of the most appropriate surrogate27
objective, differential readout of CNS function (neuro- animal models of chronic pain28–30 to those most rel-
informatics) that can inform neurobehavioural studies of evant to the human state31,32. It should be noted that,
CNS disorders and provide a novel framework in which depending on the disease model, it might be necessary
to evaluate therapeutic hypotheses rapidly. to consider imaging in gyrencephalic species, especially
The approach we propose would develop standard- if the predicted changes are predominantly cortical.
ized methods that define the functional fingerprints or The concept of functional neuropathomics is shown in
fMRI signatures in animals and humans for drugs with FIG. 2. fMRI can be used to integrate across core areas
proven clinical efficacy. This undertaking would require of CNS research that are currently segregated bioinfor-
a long-term commitment and sustained investment, but is matically. These encompass normal systems (functional
potentially transforming as it would enable a true ‘systems neuromics ), disease states (functional pathology),
Neuroinformatics
neuroscience’ approach to be applied to CNS drug dis- molecular-profiling genomic-based disease markers,
A field that deals with data
structure and software tools covery. We envisage that these activities lie initially in the pharmacotherapy or molecular mechanisms directed
devoted to the analysis and ‘pre-competitive space’, such that the resources needed to at novel targets. For example, in Parkinson’s disease33,
integration of neuroscience. establish reference databases could be shared by consor- for which genetic variation has been evaluated in the
tia formed between government, academia and industry. context of possible differences in functional neuro-
Gyrencephalic species
Mammalian species that have
An excellent example of this approach is the Alzheimer’s circuitry34, fMRI can be used to differentiate between
developed cerebrums in which Disease Neuroimaging Initiative (see Further informa- healthy and disease states and to evaluate the impact
gyri and sulci can be defined. tion), a US$60-million programme. Once established, of therapeutics.
the reference databases could be used by the biopharma-
Functional neuropathomics
ceutical industry in a proprietary confidential setting to fMRI: drug discovery to clinical trials
Intended to define the
underlying pathophysiology evaluate new molecules and hypotheses. The path of traditional outcomes-based drug discovery
in neural conditions at a The simple and complex systems that are involved and development can be long, expensive and uncer-
systems level. in normal and pathological acute and chronic pain tain35–37. Early knowledge that speeds decision-making
states, and the response to analgesic drugs, have been about candidate molecules and therapeutic concepts
Functional neuromics
Defining the functional
well defined by clinical and basic science research. will facilitate the development process by ensuring that
components of normal neural Pain and analgesia therefore provide a rational start- if we are to fail, we fail fast. This will enable resources to
function at a systems level. ing point for exploring the current and potential uses be focused on more promising molecules or alternative

NATURE REVIEWS | DRUG DISCOVERY ADVANCE ONLINE PUBLICATION | 3


REVIEWS

Molecular imaging (PET) Molecular diagnostics


Functional imaging (fMRI) Genomic profiling

Disease Function Targets Pathways Structure

Function Organ Tissue Receptors Cellular Genomic

Figure 1 | Locus of imaging in drug development — from disease to genomics and back again. The figure
shows the spectrum that imaging covers from molecular (for example, positron-emission tomography (PET)) and
functional (for example, fMRI) imaging to its use in molecular diagnostics and a link to genomic profiling in the
clinical and preclinical domains. Potential targets (circuits or specific regions) can be evaluated in the context of
the role they play in functional pathways, and can be studied in the preclinical domains using more invasive
processes, including molecular and cellular imaging, before detailed evaluation using high-field and high-resolution
functional, neuroanatomical and microimaging techniques. (Thanks to K. Moldoff for permission to use some of
these graphics)

hypotheses. This is becoming increasingly important, as The fMRI approach is circuit-based, and might therefore
there are burgeoning numbers of molecules to choose also reveal activation that is cautionary for potential CNS
between and no shortage of potential therapeutic tar- side effects, including mechanism-based or off-target
gets that are being revealed by genomics in health and activities that could influence compound selection for
disease. Indeed, it can be argued that the more novel the clinical trials.
target chosen, the lower the probability of successfully fMRI has now been used to study rodents and primates
using the target to treat a particular disease. in both the anaesthetized and awake states42–44. Activation
Below we discuss how fMRI can enhance every level in specific circuitry of the brain produces a dynamic
of drug development from drug discovery through signature that can be used to define behaviours and the
preclinical and clinical development. Indeed, fMRI functional consequences of a drug effect. Identification
might provide novel insights suggesting new uses for and mapping of the specific brain structures that are
marketed drugs, and could help rescue safe and well- involved directly and indirectly in behavioural responses
tolerated CNS agents that have failed in their primary (such as activation of sensory pathways) can be used to
endpoint by revealing hitherto unknown therapeutic reveal the involvement of previously unsuspected brain
potential. An overview of the incorporation of fMRI into regions. Mapping the effects of drugs with known clinical
drug discovery is shown in BOX 3. A number of reviews properties can be used to generate an fMRI ‘fingerprint’
have focused on pharmacological fMRI38,39 of the CNS that is characteristic of their activity. Knowledge from
in humans40 and animals in drug discovery41, and we experiments such as these could provide a rational basis
recommend these as additional reading. for initiating novel research into the pharmacology of
the newly identified brain regions and a means to select
fMRI in preclinical development between novel drug molecules by comparing their
In preclinical phases, fMRI can be used to focus the selec- fingerprints against those with known clinical effects.
tion of clinically relevant CNS animal assays, enhance
target validation based on objective assessment of behav- fMRI as a functional bioassay. Preclinical animal assays
iour, increase the speed of lead optimization based on have great mechanistic value for examining in vivo the
functional CNS responses, and define circuits of action pharmacology and neurochemistry of CNS active agents,
based on the neuroanatomy of the functional signature. but are often not true surrogates of human disease. There

4 | ADVANCE ONLINE PUBLICATION www.nature.com/reviews/drugdisc


REVIEWS

Integrative
neuroinformatics

Markers
Specific brain Disease Drug Novel
for disease
regions state development targets
and drug

Functional Functional Genetic Pharmaco- Molecular


neuromics pathology variations therapy mechanisms

Brain function

Figure 2 | Functional neuropathomics — integrating brain function. The figure summarizes the potential value of
the use of fMRI when integrated (integrative neuroinformatics) into other neuroscience core disciplines such as;
systems biology (functional neuromics), to contribute to basic understanding of how the brain or specific brain
regions are involved in a particular function; the evaluation of central nervous system (CNS) disease states (functional
pathology; for example, chronic pain and depression and so on), to define how these areas malfunction at a
functional level, including how these evolve over time; the contribution of genetic markers (for example, genes or
single-nucleotide polymorphism markers), to define specific functional markers for CNS disease or effects of specific
drugs on disease; drug evaluation (drug development) or the clinical evaluation of the efficacy of a drug; and finally,
the evaluation of mechanisms (molecular mechanisms) of potential novel targets or the early evaluation of drugs that
could have follow-on uses.

is a general lack of understanding of what preclinical CNS applications51. They have failed comprehensively
behavioural responses are predictive of, as the func- in extensive Phase III trials for the treatment of depres-
tional neuroanatomy of the response is unknown (FIG. 3). sion52 but provide protection against chemotherapy-
Animal assays, especially those used in neuropsychiatric induced nausea and vomiting through their actions in
drug discovery, can tend to be mechanism specific the brainstem53,54. fMRI might be useful in assessing the
and sometimes rediscover what was known already. A potential utility of NK1 receptor antagonists in other
consequence of the fact that fMRI measures functional CNS disease domains, and could reveal potentially novel
circuits that underlie behaviour is that it can be used to indications for follow-on drug development worthy of
probe the fundamental neuroanatomical basis of CNS more extensive clinical trials55,56.
behavioural assays. This is an important consideration We propose that the preclinical step in the early
because pharmacological manipulation of subconscious evaluation of potential therapeutics can be improved
processing can be of great significance in CNS drug by using an objective fMRI circuit-based approach
development, and this is difficult to objectively define to evaluate preclinical models instead of interpreting
or evaluate in behavioural assays, many of which have ‘binary’ type motor behavioural responses. Measuring
motor endpoints. Sometimes, however, fMRI might be complex phenomena with single-focus behaviours
more sensitive than the behavioural assays it reports45. (for example, paw withdrawal) clearly has conceptual
Model selection and validation remains a key chal- limitations, as well as potential practical complications
lenge in understanding or predicting drug efficacy46. based on interference with motor, motivational and
Current animal models have been relatively successful in peripheral nociceptor systems. By using CNS circuits
the discovery of anti-inflammatory analgesics but have as our measure, we can map laboratory models to
had limited success in predicting the activity of novel early clinical studies in healthy volunteers by examin-
agents for the pharmacotherapy of chronic pain. For ing the effects of therapy in activation paradigms that
example, nerve growth factor (NGF) showed interesting produce similar global activation patterns in animals
preclinical efficacy47, as did substance P neurokinin 1 and humans before more complex studies in patients
(NK1) receptor antagonists, but failed in clinical with chronic disease (FIG. 3). This is an iterative pro-
trials 48–50. Interestingly, substance P NK1 receptor cess that allows functional bioassays of physiological
antagonists have been suggested to have multiple other pain, inflammatory pain and neuropathic pain to be

NATURE REVIEWS | DRUG DISCOVERY ADVANCE ONLINE PUBLICATION | 5


REVIEWS

Species behaviour fMRI-based drug profiling: fMRI fingerprinting. The


identification of specific areas of the brain that are
active in disease processes in combination with, or
through the actions of, a drug enables mechanistic
inferences to be drawn regarding how these processes
are mediated. Once a drug/disease profile is identi-
? fied, novel animal assays with greater predictive value
can be developed to test novel drug candidates. The
development of a preclinical fMRI profile database is
a prerequisite for this approach. For example, such a
database could be used to functionally define areas of
Functional circuits the brain that are implicated in neuropathic pain and
their interaction with drugs known to be clinically
effective and those known to have failed as analgesics
Somatosensory cortex
(that is, functional fingerprinting). These databases
must be able to be interrogated alongside parallel clini-
Sensory cal data-sets to establish the equivalence or lack thereof
of the animal and human data based on circuit function
Thalamus (FIG. 4). Once defined, this behavioural and functional
‘efficacy fingerprint’ can be used translationally by drug
development groups to guide selection for clinical test-
ing of potentially effective drug candidates that have a
higher probability of success. fMRI evaluation of drug
Cingulate cortex effects on neuronal circuits might provide a unique way
to separate lead compound finalists that are equivalent
Emotional in simplified behavioural systems. There is emerging
preclinical and clinical fMRI literature on the direct
Nucleus accumbens effects of drugs on CNS activity across a range of phar-
macological agents and targets (TABLE 2). It is notewor-
thy too that activation maps of the brain generated by
Figure 3 | The language of translation — circuits not behaviour. fMRI brings fMRI might be the only way to provide proof of target
integrative processing to central nervous system (CNS) function and becomes the and target engagement for full agonist, partial agonist
‘language of translation’ between the human condition and preclinical models by and inverse agonist drugs. Potent CNS agonists that
focusing on functional circuits and not behavioural assays. The top part of the act at low-density receptor sites in small brain regions
figure indicates that behaviours measured in humans and in animals are hard to cannot be tracked using PET neuroreceptor imaging
compare and difficult to interpret in terms of producing animal models of a disease
because the regional anatomy and fractional occu-
state (for example, chronic pain) that are truly predictive of the human condition in
terms of drug evaluation. Numerous candidate drugs fail in clinical trials with very pancy required for pharmacological effects is beyond
strong preclinical/behavioural results. The lower part of the figure shows fMRI data the resolution of the technology. The use of fMRI in
from rats and humans exposed to the same thermal stimulus (46°C) that activates combination with transgenic and knockout mouse
nociceptors in a similar manner across mammalian species. Note that activations in genetics could increase the value of both technolo-
sensory and emotional systems are not only similar for structures and within gies in defining the links between genetics and CNS
structures (medial and lateral thalamic activations) but also for strength of signal function, and also provide a platform for CNS target
(decreased signal in the nucleus accumbens). This approach requires animal and validation and drug candidate selection. It should
human databases. Such databases would converge on a specific signature for a drug be noted that, although this approach might identify
or drug family. The specific signature or functional fingerprint is directly related to drugs with a common mechanism of action (MOA), no
physiological activity of the drug or class of drugs.
single, specific MOA is required because the fingerprint
predicts a therapeutic category of drug effect.

defined by their neurocircuitry, and for drug activity Integration: top-down approach. Non-invasive imaging
to be defined by agonist/antagonist-induced changes can provide specific insights into drug and disease phe-
in these pathways rather than by behaviour alone57,58. notypes in humans that can be used to align preclinical
TABLE 1 summarizes reports on animal and human CNS models and to evaluate effective and ineffective
fMRI of pain conditions. It is noteworthy that because drug therapies, thereby improving early decision-
fMRI is non-invasive it can be repeated longitudinally making on novel candidates (FIG. 5). Many CNS drugs
to follow the evolution of changes that are known to have been found serendipitously59 from astute clinical
occur temporally in the brain in many chronic pain observations in humans that have spawned drug devel-
models. This improves the quality of data with intra- opment programmes. A more rational approach is to use
subject design, and also reduces the number of subjects imaging techniques such as fMRI to define the condition
needed for a study relative to discontinuous cross- in humans first. Indeed, the role of brain regions such as
sectional terminal techniques that are often used to the nucleus accumbens in pain processing was identified
probe brain function. in humans by fMRI prior to confirmation of their role in

6 | ADVANCE ONLINE PUBLICATION www.nature.com/reviews/drugdisc


REVIEWS

Table 1 | fMRI pain studies in animals and humans increasingly important in such commercially competi-
tive times in which there is considerable overlap across
Model Condition Refs the pharmaceutical industry in the therapeutic targets
Animal being pursued.
Formalin Tonic pain 104
fMRI in clinical drug development
Allodynia Capsaicin 105
The optimal use of fMRI in clinical drug development
Physiological pain Electrical stimulation 106,107 is early in Phase I or II trials (FIG. 6). Using relatively
Human small populations of healthy subjects, data relating
Allodynia Neuropathic pain 81,108,109 to efficacy (targeted circuits) and side effects can be
Capsaicin 110–112 determined. Indeed, it might be possible to develop
experimental models of some clinical conditions in
Pressure Back pain 113
Fibromyalgia 114,115 healthy subjects to facilitate the evaluation of drug effi-
cacy (FIG. 5)63. fMRI responses could also be useful for
Physiological Electrical stimulation 102,116 optimizing drug-dosing regimes. The accumulation of
Noxious heat 60,117–122
Noxious cold 122,123 objective fMRI data alongside traditional early clinical
data on novel drug candidates could provide a unique
Visceral pain Experimental 124–128
Irritable bowel 129,130 metric that affects go/no-go decisions. Indeed, early in
drug discovery and development it is essential to have
several orthogonal sources of information to increase
experimental models60, making evaluation of their roles confidence in decision-making. fMRI can contribute
in animal models more credible and useful61. It has been unique information about the potential success of a
argued that a hypothesis-driven MOA approach to drug drug candidate or therapeutic hypothesis that, when
development has limited utility because many drugs interwoven with traditional metrics such as safety and
that are effective have no clear MOA (for example, anti- tolerability, can provide an early matrix within which
epileptics62) and others have an unpredicted rich and to judge probability of success.
complex pharmacology that provides benefit through
multiple actions. In these cases, looking at the end of Therapeutic versus side effects. fMRI interweaved
the process — the functional consequences of CNS drug into early clinical trials, particularly safety and toler-
effects — to match key components of the brain fMRI ability studies involving dose escalation, could provide
fingerprints provides a unique opportunity to identify an early indicator of both therapeutic and side-effect
‘outside the box’ novel therapeutic approaches. This is potential. Indeed, incorporation of fMRI into early

a Anatomical scan b Functional scan c Brain segmentation d Whole-brain parcellation

e ‘Brain array’ f Array fingerprint g Analysis/efficacy h Molecule

Figure 4 | Functional brain arrays. Anatomical scans (a) provide high-resolution details of the brain onto which
functional activation patterns can be mapped (b). By compartmentalizing anatomically defined brain regions
(parcellation/segmentation) (c), it is possible to obtain defined regional anatomy for the whole brain (d). e | A brain array
based on activations within anatomically defined regions of the brain can then be generated. Such an approach can
allow for differentiation of drugs or diseases by so-called ‘array fingerprinting’ (f) that allows for network reconstruction
and analysis steps to define neuroanatomically linked groups or neurocircuits in different ways, including efficacy (g).
h | Array fingerprinting can also be used to match different chemical compounds, because novel therapeutic compounds
can be recognized by their fMRI signature expression profiles.

NATURE REVIEWS | DRUG DISCOVERY ADVANCE ONLINE PUBLICATION | 7


REVIEWS

Table 2 | Pharmacological MRI: drug-induced CNS circuit activation It should also be noted that fMRI can detect unantici-
pated central effects of peripherally targeted analgesic
Drug(s) Mechanism Refs
drugs that could have side-effect liability. This is espe-
Animal cially important when selectivity is achieved through
d-Amphetamine Decrease dopamine reuptake 131 differential rather than absolute differences in receptor
Amphetamine Modulation of dopamine D2/D3 132 distribution between the peripheral and central compart-
function ments, or through manipulation of the physico-chemical
properties of a molecule to exclude it from the CNS. An
Dopamine Antagonism of amphetamine 133
example that preclinically had well-defined antinocic-
Bromocriptine and haloperidol D2 receptor agonist 134 eptive effects is the κ-opioid agonist ADL 10-010168,69,
Quinelorane D2/D3 receptor agonist 135 which was designed to be confined to peripheral recep-
Cocaine Dopamine reuptake inhibitor 136–138 tor sites. Experience with ADL-10-0101 is instructional
because it was found to produce dysphoria similar to
Bicuculine GABAA antagonist 139
other κ-agonists in later-stage clinical development70.
LY404187 AMPA receptor potentiator 140 Such unexpected surprises are to be avoided at all costs
GR-73632 NK1 receptor antagonist 141 as the failure is expensive. Had this drug been evalu-
Baroreceptor activation α-Adrenergic agonist 142 ated early using fMRI, perhaps preclinically at doses
Phenylephrine or sodium nitroprusside Nitric oxide release active in the pain assays, but almost certainly in early
clinical evaluation, then this unwanted and unexpected
Rivastigmine ACE inhibitor 143
side effect, which the drug was designed to avoid, might
Nicotine withdrawal Cholinergic agonist 144 have been predicted. Finally, it is also clear that fMRI
Heroin Opioid agonist 145 could be used to evaluate the central effects of therapies
Morphine withdrawal Naloxone reversal 146 that are targeted to diseases outside the CNS, because, if
present, they might prevent or limit the clinical utility of
HU210 Cannabinoid receptor agonist 147
potentially useful agents.
Human
Naloxone (N) µ-Opioid antagonist 148 fMRI to optimize drug dosing. Can drugs initially
Methamphetamine (N) Dopamine release 149 evaluated in healthy subjects in early clinical trials be
used as ‘smart’ indicators of drug efficacy in patient
Morphine (N) µ-Opioid agonist 58
populations, and will this index change with disease
Cocaine (P) Dopamine reuptake inhibitor 150 progression? In chronic pain, changes are now known to
Nicotine 151 take place along the entire neural axis, from peripheral
Note: these references are measures of the direct effect of a drug on CNS processing; for nerves to cortical regions. Insights show changes related
measures of effects of drugs on behavioural or other paradigms in humans see a recent review to pathological centralization of pain (hypersensitive
on human pharmacological MRI40. ACE, acetylcholinesterase; AMPA, α-amino-3-hydroxy-5- circuits), apparent loss of gray matter21 and altered func-
methyl-4-isoxazolepropionic acid; CNS, central nervous system; GABA, γ-aminobutyric acid;
NK1, neurokinin 1; N, normal/healthy volunteers; P, patients. tional connectivity71. It is also well known that there are
significant differences in the tolerability of several drug
classes (for example, antipsychotic, antidepressant and
low-dose clinical studies authorized under the experi- analgesic) used for the treatment of chronic undulating
mental Investigational New Drug (eIND) application CNS diseases at effective therapeutic doses when given
process could increase attrition of unsuitable drug can- to healthy volunteers rather than patients. Indeed, in
didates and mechanisms. Moreover, by focusing early some individuals the therapeutic dose makes no sense
on the best candidates and hypotheses there is greater in terms of simple receptor biology. How, then, do we
safety for human subjects in clinical trials. An example judge dosing to optimize benefits over side effects? We
case is an analgesic (for example, gabapentin, morphine suggest that fMRI has the potential to fulfill this need by
and amitryptiline) in which the desirable CNS effects providing an objective readout of CNS function during
are on sensory systems, descending pain control sys- drug dosing that could be correlated to drug exposure
tems and limbic structures including the amygdala64. and disease stage, and perhaps also used in experimen-
By contrast, major CNS side effects include sedation, tal paradigms conducted in the context of disease (for
dysphoria, possible addiction, and nausea and vomit- example, in analgesia and attention deficits72).
ing65,66 that can vary depending on the subject’s age,
sex or genetic make-up. fMRI can define the circuits Neurotransmitter status. fMRI has been used to evalu-
involved in these responses in a rational objective ate a number of neurotransmitter systems to study neu-
manner. For example, the neurocircuitry involved in ral processing in humans (TABLE 2), and this has been
sedation is known to involve intralaminar nuclei of the subject of a recent review40. Most of these studies
the thalamus and cortical regions. Reward circuitry, have, however, not evaluated direct effects of drugs on
particularly the accumbens and orbitofrontal lobe, are neural pathways but have instead used behavioural
involved in euphoria and dysphoria, and brainstem interventions. Nevertheless, clever study design allows
nuclei (nucleus tractus solitarius) are known emetic for the functional dissection or modulation of neural
centres67. If these are activated, the decision not to pathways specific to a particular neurotransmitter
pursue further development might be taken. system (for example, dopamine73). In addition, there

8 | ADVANCE ONLINE PUBLICATION www.nature.com/reviews/drugdisc


REVIEWS

Drugs Surrogate model fMRI Signal example, the antimalarial mefloquine (Larium) produces
differentiates psychosis as a side effect75 with an incidence rate of 33%
drug efficacy
in a randomized controlled trial, even though only 6%
sought medical attention76. fMRI evaluation of this type
Drug 1
of drug can be used for a number of ends: to understand
circuitry activated during psychosis in otherwise healthy
Drug 2 subjects; as a model to discover drugs that ameliorate
this effect (compared with fMRI data on activation of
Drug 1 Healthy subjects regions such as the hippocampus in schizophrenia77);
Drug 2 +
Drug 3 stressor Placebo and to target investigations of drugs in small populations
Drug 4 + of patients with schizophrenia.
Placebo drug
Drug 3 Phenotyping for focused Phase II and III CNS clinical
trials. Mechanistic approaches to drug development
have been hampered by the lack of objective phenotypes
Drug 4 for diseases that are polygenetic or do not have specific
biomarkers. Studies to define a correlation between
Figure 5 | Surrogate models in drug evaluation. Because fMRI reflects neural activity, genotype and phenotype have now begun using fMRI
and brain systems are by their very nature complex and interconnected, activation maps in small focused groups of patients with known genetic
provide a ‘picture’ of activation patterns in response to a perturbation (BOX 1) at a subtypes (for example, Parkinson’s disease) and in
particular time. For example, in response to pain stimuli, sensory, emotional, autonomic
larger cohorts using single-nucleotide polymorphism
and endogenous analgesic systems all process this information to evaluate and respond
to the perturbation. The complexity of the system is such that overall maps of the
(SNP) or related genotype mapping78–80. Phenotyping to
patterns of activation might prove useful for interpreting drug action. This can be carried produce functional classifiers using a circuit-based fMRI
out in healthy subjects to differentiate drugs that could have utility in a particular fingerprinting approach could yield novel experimental
disease, such as neuropathic pain63. Even in the diseased state, drugs also act on ‘normal’ imaging paradigms81 in which these markers are used
or unaffected brain regions which may contribute to the distributed response. The to evaluate novel drug candidates in defined patient
figure shows that, following different drug treatments, fMRI measures of a surrogate populations that are enriched and stratified on the basis
model of stress produce different activation maps for each drug compared with placebo, of brain function.
which might differentiate drug efficacy (that is, drugs 1 and 2 being useful for, and drugs
3 and 4 having no utility for, the condition). fMRI as an integrator. fMRI provides a hub for integrat-
ing the technologies and stages of CNS drug develop-
ment. fMRI can act as a key translational technology
are a few studies using fMRI with correlative plasma for CNS research that provides a bridge from human to
levels72,74 to define a relationship between brain activation, animal and back again. Within the preclinical and clini-
drug plasma levels and effect. cal arenas it can interface with other target validation
and MOA strategies to provide functional correlates of
fMRI for marketed or failed yet safe drug candidates. drug action. fMRI signal change is thought to reflect
There are two broad areas of use of fMRI in this domain: changes in field potentials82, and can therefore direct
extending a known compound into a new therapeutic drug discovery using in vitro slice electrophysiology
area, and rescuing a CNS active compound that has failed and in vivo electrophysiology to the correct brain regions
for its primary CNS indication by defining a new indica- involved in a drug’s action and focus evaluation of drug
tion for which it was not originally designed51,55,56. hits that come from mechanistic screens. fMRI-defined
Few marketed CNS drugs have a well-defined MOA. pathways can inform the interpretation of clinical symp-
They are categorized simply on the basis of the cell sur- toms and drug effects, potentially helping to optimize
face proteins with which they interact rather than the dose selection and therapeutic windows. fMRI can also
neurobiological consequences of the interaction. Drugs be leveraged by integration with genomic analyses at a
that were traditionally used in different domains have human level (for example, SNP mapping) to describe
serendipitously found uses for new indications55. This the complex functional changes that occur in different
has been obvious in the anti-epileptic domain, in which psychiatric and neurological diseases. Finally, fMRI
many drugs have found off-label use in neuropathic pain, can be used with mouse genetic approaches to provide
post-traumatic stress disorder or other CNS diseases. novel mechanistic insights that link CNS dysfunction to
Functional classifiers A more defined method of evaluating these drugs is to altered genes, thereby providing a segue from human
Drugs can act on neural identify circuit activation that is predictive of potential disease to animal models and back83,84.
networks or systems in a
use against a database established in this domain. This
particular fashion — for
example, produce sedation, could use the advanced bioinformatics approaches that A working example. Defining an objective model of
euphoria or analgesia. have been developed to define molecular signatures in neural systems in humans allows for hypothesis-driven
Activation patterns that show genomic profiling (FIG. 4). evaluation of pain systems (physiological, inflammatory
specific regions of the brain and neuropathic conditions); the objective definition
known to be involved in a
particular function allow for
fMRI studies of drug side effects. Some drug-induced of surrogate models for disease states or drug effects
the segregation or functional side effects can be used as models for understanding (see below); and the evaluation of these effects over
classification of drug action. disease symptoms and therapeutic drug effects. For time. Our group has begun to use this approach by

NATURE REVIEWS | DRUG DISCOVERY ADVANCE ONLINE PUBLICATION | 9


REVIEWS

Human drug database

a d
Objective markers Objective markers
of drug action of drug efficacy
GO/NO GO and disease

Clinical drug
IND Phase I Phase II Phase III NDA Approval
development

Clinical
Repeat dose
pharmacology Safety
toxicology
&ADME/PK
b
Initial clinical Dose Manufacture
evaluation ranging QA methods

c
Repeat dose
Efficacy
toxicology

Long-term
toxicology

Manufacture Manufacture Manufacture


QA methods scaleup scaleup

Figure 6 | Clinical applications in Phase I studies: enhanced information for go/no-go decisions. fMRI has
applications in early-phase drug development in three principal domains. a | fMRI can provide objective markers of
drug action for go/no-go decisions. b | fMRI can provide additional information for central nervous system dose
ranging. c | fMRI can provide insights into efficacy of drug action in the healthy population normally used in Phase I
studies. d | fMRI can provide an objective marker of efficacy in Phase II and III trials once tools and standards are
developed to deal with the diversity of disease within clinical populations with a common diagnosis. ADME,
absorption, distribution, metabolism and excretion; IND, Investigational New Drug; NDA, New Drug Application;
PK, pharmacokinetics; QA, quality assurance.

defining activation in trigeminal pathways7,85,86 (FIG. 7). alter afferent inputs to the brain or affect cardiovascular
This has allowed us to evaluate changes that occur in changes or nonspecific functional responses must be
clinical conditions affecting the trigeminal system, such investigated, and attention to experimental design and
as neuropathic pain and migraine. With respect to the data analysis is required41,88.
latter, preclinical and clinical work has defined a model The on/off approach to stimuli for disease state or drug
of progressive central sensitization of the trigeminal evaluation is currently standard. This approach provides
system that underlies allodynia associated with migraine provocative stimuli that can be used to evaluate differences
attacks87. Specifically, fMRI might provide insights into in the disease versus non-disease state with or without
inter-individual differences in responsiveness to triptan pharmacological interventions. Novel approaches that use
5-hydroxytryptamine (5-HT1B/1D) agonist antimigraine resting fMRI have provided useful insights into process-
agents, as well as inter-attack differences within a patient, ing and connectivity in CNS disease89–91. These and other
and reveal their true clinical sites of action. technical developments, such as real-time imaging92,
could enhance the utility of magnetic resonance tech-
Challenges. fMRI is a relatively new field and its appli- nologies by adding new dimensions to the study of brain
cations to drug development are still in their infancy. function in health and disease.
fMRI holds great promise because it adds objective With all its potential to elucidate new paths and appli-
methods for the measurement of CNS function in awake cations in drug discovery, we need to recognize that fMRI
animals and humans. However, like any technology, is a relatively new technique and as such has yet to be
there are limits and caveats that are worth noting. How fully characterized. Several studies have looked at fMRI
valid is the use of fMRI in CNS disease and drug evalu- reproducibility in terms of size (intensity) and focus of
ation? How sensitive and reproducible is this? Is it a activation93,94. Studies involving sensorimotor tasks,
population-based approach or an individual approach? working memory and other tasks have looked at how well
Global versus local changes produced by agents that the signal intensity can be reproduced95–97. In general, for

10 | ADVANCE ONLINE PUBLICATION www.nature.com/reviews/drugdisc


REVIEWS

Define functional system Apply functional system has been a significant contribution to the pain-imaging
d
literature from numerous laboratories around the world.
Indeed, the notion that other domains are using pain in
Physiological model imaging studies speaks for the robustness of the stimulus
in imaging paradigms. As a result, the field has advanced
to a point that offers the exciting possibility of using this
c technology both in drug development and in the clinic.
For the technology to become valid as a diagnostic tool
Migraine
(for drug/therapeutic effects of disease state), it needs
to satisfy the standard set of criteria for validity: face
validity (a reflection of the problem), construct valid-
ity (robustness/reproducibility of assay), and predictive
b Neuropathic pain validity (defining or predicting a result in the context of
false-positive/-negative outcomes)101.
The revolution in imaging from a technological
and analytical point of view offers some exciting
prospects for clinical use. If one critically evaluates
Surrogate models
progress being made in fulfilling the standards of
validity, based on the enormous breakthroughs that
the technology has provided within the past couple
of decades, then the future looks bright in terms of
Drug/analgesic effects a transition from basic research in functional imag-
a ing of pain into clinical utility. Given that scanners
Figure 7 | Application of a defined systems approach to models and disease in are available at most academic institutions, access to
humans. This is an example of the use of functional imaging to define a neural system instruments to perform studies is not likely to be a
(that is, trigeminal sensory pathway) that includes first-, second- and third-order neural limiting factor. The main issue is the heterogeneity
projections to the cortex. Activation can be measured in first-order (trigeminal of tools, technologies and techniques across imag-
ganglion (a)); second-order (spinal nucleus of the trigeminal nerve (b)); third-order ing centres that make fMRI studies, even in the
(thalamus (c)); and fourth-order (primary sensory cortex (d)). Activations shown as same domain, difficult to compare. Clearly what is
cutouts of the head (small white boxes) are shown in enlarged boxes. An understanding now needed are common standards that define the
of the connectivity and anatomical regions involved under baseline conditions
populations included in imaging studies, ensure the
(physiological pain) can then be used to identify changes in these pathways in clinical
specificity of stimuli, data collection and analysis to
conditions such as migraine or neuropathic pain, applied to the development of
surrogate models such as capsacin induced sensitization of the system or used to enable the field to advance.
measure the effects of drugs on this pathway in health and disease. Modified, with
permission, from REF. 85 (C) Wiley (2004). Realizing the potential
The ability to realize the potential of fMRI in drug
development will be challenged by practical issues
the same subjects, signals are found to be highly repro- including costs, validation and the need to define
ducible. The variability among subjects, however, is sig- new analytical mathematical approaches to evalu-
nificantly larger. For population studies, a cohort size of ate complex neurological and psychiatric diseases.
about 15 subjects is necessary to draw conclusions98. One Issues include current evaluation of the disease at
of the main obstacles for fMRI is the high-fidelity stand- baseline, medication use (including potential taper-
ardization of different brain anatomies across subjects ing paradigms), appropriate ethical approaches to
to a common anatomical template99. New approaches human experimentation in these difficult populations,
to reduce variability use calibration techniques to adjust and evaluation of novel clinically unproven drugs in
each subject’s response to a standard obtained from the these conditions that could require withdrawal of best
same subject100. fMRI enjoys high specificity in activa- practice usual care.
tion patterns across tasks and subjects. fMRI sensitivity The benefit–cost relationship is difficult to predict,
varies with the brain structure under study, the specific but what is clear is that the traditional mechanism- and
paradigm used and the scanner characteristics (such as target-based framework for CNS drug discovery has
magnetic field and imaging coils). For cortical structures recently not delivered truly novel approaches to com-
involved in visual or sensorimotor tasks in a 3-Tesla scan- plex CNS diseases. Functional systems and circuit-based
ner, it is possible to detect activation in volumes as small approaches have the potential to provide a paradigm
as 1 mm3. For subcortical structures the detectable vol- shift that could advance the field of CNS drug discovery
ume is somewhat larger. The wide diversity of techniques beyond the roadblock that we have today.
and instruments from different manufacturers used in
fMRI drives the need for common standards for routine Conclusions
studies across imaging centres. Non-invasive functional imaging has fundamentally
The example of pain and analgesia serves to validate changed the way that we can question brain systems
the use of fMRI in disease evaluation and drug develop- because we can see them in action. fMRI brings
ment101. Since the early fMRI studies in pain102,103 there systems neuroscience to life by showing how genes,

NATURE REVIEWS | DRUG DISCOVERY ADVANCE ONLINE PUBLICATION | 11


REVIEWS

proteins, transmitters, neurons and neural networks species achieved by measuring neural circuit function
work togther to create brain responses in health and as a correlate of behaviour, might transform our cur-
disease. fMRI can be used in drug development in rent thinking in neuroscience.
new and exciting ways that provide valuable informa- Ultimately, novel technologies need to prove their
tion relating to drug and disease phenotype, using a value. The potential future for fMRI technology is to
circuit-based approach to defining behaviours that contribute to decreasing risks and costs of CNS drug
makes translational processes between the preclini- development by increasing the ability to make focused
cal and clinical domains more useful. Many of these rational choices that help decision-making and so speed
insights, such as ‘the language of translation’ across the discovery of safe and cost-effective therapies.

1. Cohen, C. M. A path to improved pharmaceutical 21. Apkarian, A. V. et al. Chronic back pain is associated 40. Honey, G. & Bullmore, E. Human pharmacological
productivity. Nature Rev. Drug Discov. 2, 751–753 with decreased prefrontal and thalamic gray matter MRI. Trends Pharmacol. Sci. 25, 366–374 (2004).
(2003). density. J. Neurosci. 24, 10410–10415 (2004). An outstanding review of the state of the art of
A critical article on the evaluation of performance A revolutionary paper and quantum leap in pharmacological fMRI.
metrics in the pharmaceutical industry. understanding that chronic pain is a degenerative 41. Shah, Y. B. & Marsden, C. A. The application of
2. Kola, I. & Landis, J. Can the pharmaceutical industry disease affecting the CNS. functional magnetic resonance imaging to neurophar-
reduce attrition rates? Nature Rev. Drug Discov. 3, 22. Tracey, I. Prospects for human pharmacological macology. Curr. Opin. Pharmacol. 4, 517–521 (2004).
711–715 (2004). functional magnetic resonance imaging (phMRI). 42. Vanduffel, W. et al. Visual motion processing
An evaluation of fiscal issues facing R&D and how to J. Clin. Pharmacol. (Suppl.), 21S–28S (2001). investigated using contrast agent-enhanced fMRI in
improve this domain in the pharmaceutical industry. A good review of the potential of fMRI in awake behaving monkeys. Neuron 32, 565–577
3. Frank, R. & Hargreaves, R. Clinical biomarkers in drug pharmacology. (2001).
discovery and development. Nature Rev. Drug Discov. 23. Borsook, D., Ploghaus, A. & Becerra, L. Utilizing brain 43. Febo, M. et al. Imaging cocaine-induced changes in the
2, 566–580 (2003). imaging for analgesic drug development. Curr. Opin. mesocorticolimbic dopaminergic system of conscious
Biomarkers will become highly valued processes to Investig. Drugs. 3, 1342–1347 (2002). rats. J. Neurosci. Methods 139, 167–176 (2004).
characterize drugs and disease states. 24. Borsook, D. & Becerra, L. Pain imaging: future 44. Brewer, A. A., Press, W. A., Logothetis, N. K. &
4. Pritchard, J. F. et al. Making better drugs: Decision applications to integrative clinical and basic Wandell, B. A. Visual areas in macaque cortex
gates in non-clinical drug development. Nature Rev. neurobiology. Adv. Drug. Deliv. Rev. 55, 967–986 measured using functional magnetic resonance
Drug Discov. 2, 542–553 (2003). (2003). imaging. J. Neurosci. 22, 10416–10426 (2002).
5. Bleicher, K. H., Bohm, H. J., Muller, K. & Alanine, A. I. 25. Fava, M., Mallinckrodt, C. H., Detke, M. J., Watkin, J. G. 45. Wilkinson, D. & Halligan, P. The relevance of
Hit and lead generation: beyond high-throughput & Wohlreich, M. M. The effect of duloxetine on behavioural measures for functional-imaging studies of
screening. Nature Rev. Drug Discov. 2, 369–378 painful physical symptoms in depressed patients: cognition. Nature Rev. Neurosci. 5, 67–73 (2004).
(2003). do improvements in these symptoms result in higher 46. Brunner, D., Nestler, E. & Leahy, E. In need of high-
6. Bild, A. H. et al. Oncogenic pathway signatures in remission rates? J. Clin. Psychiatry 65, 521–530 throughput behavioral systems. Drug Discov. Today 7,
human cancers as a guide to targeted therapies. (2004). S107–S112 (2002).
Nature 439, 353–357 (2006). 26. Ohayon, M. M. Specific characteristics of the pain/ 47. Ren, K., Thomas, D. A. & Dubner, R. Nerve growth
7. DaSilva, A. F. et al. Somatotopic activation in the depression association in the general population. factor alleviates a painful peripheral neuropathy in
human trigeminal pain pathway. J. Neurosci. 22, J. Clin. Psychiatry 65 (Suppl. 12), 5–9 (2004). rats. Brain Res. 699, 286–292 (1995).
8183–8192 (2002). 27. Klein, T., Magerl, W., Rolke, R. & Treede, R. D. Human 48. Apfel, S. C. et al. Efficacy and safety of recombinant
One of the first papers to describe pain activation surrogate models of neuropathic pain. Pain 115, human nerve growth factor in patients with diabetic
in the spinal cord (brainstem) in humans. Imaging a 227–233 (2005). polyneuropathy: A randomized controlled trial. rhNGF
complete pathway provides an approach to 28. Bennett, G. J. & Xie, Y. K. A peripheral Clinical Investigator Group. JAMA 284, 2215–2221
evaluate surrogate models and disease states mononeuropathy in rat that produces disorders of (2000).
affecting the trigeminal system. pain sensation like those seen in man. Pain 33, 49. Goldstein, D. J., Wang, O., Gitter, B. D. & Iyengar, S.
8. Hornung, J. P. The human raphe nuclei and the 87–107 (1988). Dose-response study of the analgesic effect of
serotonergic system. J. Chem. Neuroanat. 26, 29. Decosterd, I. & Woolf, C. J. Spared nerve injury: an lanepitant in patients with painful diabetic neuropathy.
331–343 (2003). animal model of persistent peripheral neuropathic Clin. Neuropharmacol. 24, 16–22 (2001).
9. Grzanna, R. & Fritschy, J. M. Efferent projections of pain. Pain 87, 149–158 (2000). 50. Hill, R. NK1 (substance P) receptor antagonists--why
different subpopulations of central noradrenaline 30. Mantyh, P. W. A mechanism-based understanding of are they not analgesic in humans? Trends Pharmacol.
neurons. Prog. Brain. Res. 88, 89–101 (1991). bone cancer pain. Novartis Found. Symp. 261, Sci. 21, 244–246 (2000).
10. Everitt, B. J. & Robbins, T. W. Central cholinergic 194–214 (2004). 51. Duffy, R. A. Potential therapeutic targets for
systems and cognition. Annu. Rev. Psychol. 48, 31. Blackburn-Munro, G. Pain-like behaviours in animals- neurokinin-1 receptor antagonists. Expert Opin.
649–684 (1997). how human are they? Trends Pharmacol. Sci. 25, Emerg. Drugs 9, 9–21 (2004).
11. Johansson, B. B. Brain plasticity in health and disease. 299–305 (2004). 52. Keller, M. et al. Lack of efficacy of the substance P
Keio. J. Med. 53, 231–246 (2004). 32. Mogil, J. S. & Crager, S. E. What should we be (neurokinin(1) receptor) antagonist aprepitant in the
12. Raineteau, O. & Schwab, M. E. Plasticity of motor measuring in behavioral studies of chronic pain in treatment of major depressive disorder. Biol.
systems after incomplete spinal cord injury. Nature animals? Pain 112, 12–15 (2004). Psychiatry 59, 216–223 (2005).
Rev. Neurosci. 2, 263–273 (2001). 33. Kachergus, J. et al. Identification of a novel LRRK2 53. Diemunsch, P. & Grelot, L. Potential of substance P
13. Chen, R., Cohen, L. G. & Hallett, M. Nervous system mutation linked to autosomal dominant parkinsonism: antagonists as antiemetics. Drugs 60, 533–546
reorganization following injury. Neuroscience 111, evidence of a common founder across European (2000).
761–773 (2002). populations. Am. J. Hum. Genet. 76, 672–680 54. Pendergrass, K. et al. Aprepitant: an oral NK1
14. Hatten, M. E. & Heintz, N. Large-scale genomic (2005). antagonist for the prevention of nausea and vomiting
approaches to brain development and circuitry. Annu. 34. Grafton, S. T. Contributions of functional imaging to induced by highly emetogenic chemotherapy. Drugs
Rev. Neurosci. 28, 89–108 (2005). understanding parkinsonian symptoms. Curr. Opin. Today (Barc.) 40, 853–863 (2004).
15. Seitz, R. J., Kleiser, R. & Butefisch, C. M. Neurobiol. 14, 715–719 (2004). 55. Ashburn, T. T. & Thor, K. B. Drug repositioning:
Reorganization of cerebral circuits in human brain A critical review on the use of imaging to evaluate identifying and developing new uses for existing drugs.
lesion. Acta Neurochir. Suppl. 93, 65–70 (2005). specific circuits involved in this disease that clearly Nature Rev. Drug Discov. 3, 673–683 (2004).
16. Eden, G. F. et al. Neural changes following remediation has applications to other CNS disease states. 56. Dimasi, J. A. & Paquette, C. The economics of follow-
in adult developmental dyslexia. Neuron 44, 411–422 35. DiMasi, J. A., Hansen, R. W. & Grabowski, H. G. on drug research and development: trends in entry
(2004). The price of innovation: new estimates of drug rates and the timing of development.
17. Temple, E. et al. Neural deficits in children with development costs. J. Health Econ. 22, 151–185 Pharmacoeconomics 22, 1–14 (2004).
dyslexia ameliorated by behavioral remediation: (2003). 57. Borras, M. C. et al. fMRI measurement of CNS
evidence from functional MRI. Proc. Natl Acad. Sci. 36. Rawlins, M. D. Cutting the cost of drug development? responses to naloxone infusion and subsequent mild
USA 100, 2860–2865 (2003). Nature Rev. Drug Discov. 3, 360–364 (2004). noxious thermal stimuli in healthy volunteers.
18. Stoeckel, M. C., Jorgens, S., Witte, O. W. & 37. Littman, B. H. & Williams, S. A. The ultimate model J. Neurophysiol. 91, 2723–2733 (2004).
Seitz, R. J. Reduced somatosensory hand organism: progress in experimental medicine. Nature One of the first studies evaluating CNS effects of
representation in thalidomide-induced dysmelia as Rev. Drug Discov. 4, 631–638 (2005). a drug. Interestingly, there were no behavioural
revealed by fMRI. Eur. J. Neurosci. 21, 556–562 38. Leslie, R. A. & James, M. F. Pharmacological effects associated with the drug.
(2005). magnetic resonance imaging: a new application for 58. Becerra, L., Harter, K., Gonalez, R. G. & Borsook, D.
19. Baker, C. I., Peli, E., Knouf, N. & Kanwisher, N. G. functional MRI. Trends Pharmacol. Sci. 21, 314–318 Morphine activates neural circuits involved in reward,
Reorganization of visual processing in macular (2000). sedation, and endogenous analgesia in drug-naive
degeneration. J. Neurosci. 25, 614–618 (2005). 39. Braus, D. F., Brassen, S., Weimer, E. & Tost, H. humans. Anesthesia And Analgesia (in the press).
20. Borsook, D. et al. Acute plasticity in the human [Functional magnetic resonance imaging of 59. Borsook, D. & Edwards, A. D. Antineuropathic effects
somatosensory cortex following amputation. psychopharmacological brain effects: an update]. of the antibiotic derivative spicamycin KRN5500. Pain
Neuroreport 9, 1013–1017 (1998). Fortschr. Neurol. Psychiatr. 71, 72–83 (2003). Med. 5, 104–108 (2004).

12 | ADVANCE ONLINE PUBLICATION www.nature.com/reviews/drugdisc


REVIEWS

A paper that defines emotional circuitry in 83. Prathikanti, S. & Weinberger, D. R. Psychiatric 107. Tuor, U. I. et al. Functional magnetic resonance
response to pain and has opened the door to genetics- the new era: genetic research and some imaging in rats subjected to intense electrical and
understanding the potential role of pain in clinical implications. Br. Med. Bull. 73–74, 107–122 noxious chemical stimulation of the forepaw. Pain 87,
affecting circuitry involved in the placebo response (2005). 315–324 (2000).
and in functional illness. 84. Winterer, G., Hariri, A. R., Goldman, D. & Weinberger, 108. Lebel, A. et al. fMRI of mechanical allodynia in
60. Becerra, L., Breiter, H. C., Wise, R., Gonzalez, R. G. & D. R. Neuroimaging and Human Genetics. Int Rev children with complex regional (leg) pain syndrome
Borsook, D. Reward circuitry activation by noxious Neurobiol 67PB, 325–383 (2005). (CRPS). Soc. Neurosci. Conf. Washington DC, A53.17
thermal stimuli. Neuron 32, 927–946 (2001). 85. Borsook, D., Burstein, R. & Becerra, L. Functional (2005).
61. Gear, R. W., Aley, K. O. & Levine, J. D. Pain-induced imaging of the human trigeminal system: An important model that demonstrates the use of
analgesia mediated by mesolimbic reward circuits. opportunities for new insights into pain processing in fMRI in assessing drug effects.
J. Neurosci. 19, 7175–7181 (1999). health and disease. J. Neurobiol. 61, 107–125 109. Peyron, R. et al. An fMRI study of cortical
62. Perucca, E. An introduction to antiepileptic drugs. (2004). representation of mechanical allodynia in patients
Epilepsia 46 (Suppl. 4), 31–37 (2005). 86. Borsook, D., DaSilva, A. F., Ploghaus, A. & Becerra, L. with neuropathic pain. Neurology 63, 1838–1846
This paper provides evidence that fMRI signal has Specific and somatotopic functional magnetic (2004).
a neural basis. resonance imaging activation in the trigeminal 110. Maihofner, C. & Handwerker, H. O. Differential coding
63. Becerra, L. et al. Dissecting drug efficacy for ganglion by brush and noxious heat. J. Neurosci. 23, of hyperalgesia in the human brain: a functional MRI
neuropathic pain in healthy subjects. Annu. Meeting 7897–7903 (2003). study. Neuroimage 28, 996–1006 (2005).
Soc. Neurosci. Washington DC, A50.10 (2005). 87. Burstein, R. Deconstructing migraine headache into 111. Baron, R., Baron, Y., Disbrow, E. & Roberts, T. P. Brain
64. Apkarian, A. V., Bushnell, M. C., Treede, R. D. & peripheral and central sensitization. Pain 89, processing of capsaicin-induced secondary
Zubieta, J. K. Human brain mechanisms of pain 107–110 (2001). hyperalgesia: a functional MRI study. Neurology 53,
perception and regulation in health and disease. 88. Salmeron, B. J. & Stein, E. A. Pharmacological 548–557 (1999).
Eur. J. Pain 9, 463–484 (2005). applications of magnetic resonance imaging. 112. Zambreanu, L., Wise, R. G., Brooks, J. C., Iannetti, G. D.
65. Ramaekers, J. G. Behavioural toxicity of medicinal Psychopharmacol. Bull. 36, 102–129 (2002). & Tracey, I. A role for the brainstem in central
drugs. Practical consequences, incidence, 89. De Luca, M., Beckmann, C. F., De Stefano, N., sensitisation in humans. Evidence from functional
management and avoidance. Drug Saf. 18, 189–208 Matthews, P. M. & Smith, S. M. fMRI resting state magnetic resonance imaging. Pain 114, 397–407
(1998). networks define distinct modes of long-distance (2005).
66. Bazil, C. W. Effects of antiepileptic drugs on sleep interactions in the human brain. Neuroimage 29, 113. Giesecke, T. et al. Evidence of augmented central pain
structure: are all drugs equal? CNS Drugs 17, 1359–1367 (2005). processing in idiopathic chronic low back pain.
719–728 (2003). 90. Foss, J. M., Apkarian, A. V. & Chialvo, D. R. Dynamics Arthritis Rheum. 50, 613–623 (2004).
67. Hornby, P. J. Central neurocircuitry associated with of pain: fractal dimension of temporal variability of 114. Gracely, R. H. et al. Pain catastrophizing and neural
emesis. Am. J. Med. 111 (Suppl. 8A), 106S–112S spontaneous pain differentiates between pain States. responses to pain among persons with fibromyalgia.
(2001). J. Neurophysiol. 95, 730–736 (2006). Brain 127, 835–843 (2004).
68. Eisenach, J. C., Carpenter, R. & Curry, R. Analgesia 91. Salvador, R., Suckling, J., Schwarzbauer, C. & This study provided novel insights into CNS
from a peripherally active kappa-opioid receptor Bullmore, E. Undirected graphs of frequency- processing in a disease about which we have had
agonist in patients with chronic pancreatitis. Pain 101, dependent functional connectivity in whole brain little understanding.
89–95 (2003). networks. Philos. Trans. R. Soc. B Biol. Sci. 360, 115. Cook, D. B. et al. Functional imaging of pain in
69. DeHaven-Hudkins, D. L. & Dolle, R. E. Peripherally 937–946 (2005). patients with primary fibromyalgia. J. Rheumatol. 31,
restricted opioid agonists as novel analgesic agents. 92. deCharms, R. C. et al. Control over brain activation 364–378 (2004).
Curr. Pharm. Des. 10, 743–757 (2004). and pain learned by using real-time functional MRI. 116. Alkire, M. T., White, N. S., Hsieh, R. & Haier, R. J.
70. Dortch-Carnes, J. & Potter, D. E. Bremazocine: Proc. Natl Acad. Sci. USA 102, 18626–18631 Dissociable brain activation responses to 5-Hz
a kappa-opioid agonist with potent analgesic and (2005). electrical pain stimulation: a high-field functional
other pharmacologic properties. CNS Drug Rev. 11, 93. Smith, S. M. et al. Variability in fMRI: a re-examination magnetic resonance imaging study. Anesthesiology
195–212 (2005). of inter-session differences. Hum. Brain Mapp. 24, 100, 939–946 (2004).
71. Flor, H. Remapping somatosensory cortex after injury. 248–257 (2005). 117. DaSilva, A. F. et al. Somatotopic activation in the
Adv. Neurol. 93, 195–204 (2003). 94. Casey, B. J. et al. Reproducibility of fMRI results human trigeminal pain pathway. J. Neurosci. 22,
72. Muller, U. et al. Plasma level-dependent effects of across four institutions using a spatial working 8183–8192 (2002).
methylphenidate on task-related functional magnetic memory task. Neuroimage 8, 249–261 (1998). 118. Brooks, J. C., Zambreanu, L., Godinez, A., Craig, A. D.
resonance imaging signal changes. 95. Marshall, I. et al. Repeatability of motor and working- & Tracey, I. Somatotopic organisation of the human
Psychopharmacology (Berl) 180, 624–633 (2005). memory tasks in healthy older volunteers: assessment insula to painful heat studied with high resolution
73. Honey, G. D. et al. Dopaminergic drug effects on at functional MR imaging. Radiology 233, 868–877 functional imaging. Neuroimage 27, 201–209
physiological connectivity in a human cortico-striato- (2004). (2005).
thalamic system. Brain 126, 1767–81 (2003). 96. Wei, X. et al. Functional MRI of auditory verbal 119. Moulton, E. A., Keaser, M. L., Gullapalli, R. P. &
fMRI demonstrates functional connectivity of the working memory: long-term reproducibility analysis. Greenspan, J. D. Regional intensive and temporal
human caudate nucleus using the pharmacological Neuroimage 21, 1000–1008 (2004). patterns of functional MRI activation distinguishing
agent sulpiride. 97. Yoo, S. S., Wei, X., Dickey, C. C., Guttmann, C. R. & noxious and innocuous contact heat. J. Neurophysiol.
74. Wise, R. G. et al. Combining fMRI with a Panych, L. P. Long-term reproducibility analysis of 93, 2183–2193 (2005).
pharmacokinetic model to determine which brain fMRI using hand motor task. Int J. Neurosci. 115, 120. Bornhovd, K. et al. Painful stimuli evoke different
areas activated by painful stimulation are specifically 55–77 (2005). stimulus-response functions in the amygdala,
modulated by remifentanil. Neuroimage 16, 98. Friston, K. J., Holmes, A. P., Price, C. J., Buchel, C. & prefrontal, insula and somatosensory cortex: a single-
999–1014 (2002). Worsley, K. J. Multisubject fMRI studies and trial fMRI study. Brain 125, 1326–1336 (2002).
75. Meier, C. R., Wilcock, K. & Jick, S. S. The risk of severe conjunction analyses. Neuroimage 10, 385–396 121. Apkarian, A. V., Darbar, A., Krauss, B. R., Gelnar, P. A.
depression, psychosis or panic attacks with prophylactic (1999). & Szeverenyi, N. M. Differentiating cortical areas
antimalarials. Drug Saf. 27, 203–213 (2004). 99. Brett, M., Johnsrude, I. S. & Owen, A. M. The related to pain perception from stimulus identification:
76. Overbosch, D. et al. Atovaquone-proguanil versus problem of functional localization in the human brain. temporal analysis of fMRI activity. J. Neurophysiol.
mefloquine for malaria prophylaxis in nonimmune Nature Rev. Neurosci. 3, 243–249 (2002). 81, 2956–2963 (1999).
travelers: results from a randomized, double-blind 100. Foland, L. C., Thomason, M. E. & Glover, G. H. 122. Davis, K. D., Kwan, C. L., Crawley, A. P. & Mikulis, D.
study. Clin. Infect. Dis. 33, 1015–1021 (2001). Calibrating functional MRI data across subjects and J. Functional MRI study of thalamic and cortical
77. Heckers, S. The hippocampus in schizophrenia. Am. J. scan sites. Annu. Meeting Soc. Neurosci. San Diego, activations evoked by cutaneous heat, cold, and tactile
Psychiatry 161, 2138–2139 (2004). A693.9 (2004). stimuli. J. Neurophysiol. 80, 1533–1546 (1998).
78. Zubieta, J. K. et al. COMT val158met genotype 101. Borsook, D. & Becerra, L. Functional imaging of pain 123. Tracey, I. et al. Noxious hot and cold stimulation
affects mu-opioid neurotransmitter responses to a and analgesia — a valid diagnostic tool? Pain 117, produce common patterns of brain activation in
pain stressor. Science 299, 1240–1243 (2003). 247–250 (2005). humans: a functional magnetic resonance imaging
A paper that has defined what is yet to come: 102. Davis, K. D., Wood, M. L., Crawley, A. P. & Mikulis, D. study. Neurosci. Lett. 288, 159–162 (2000).
mapping phenotype (using brain imaging) with J. fMRI of human somatosensory and cingulate cortex 124. Strigo, I. A., Albanese, M. C., Bushnell, M. C. &
genotype/SNP mapping. during painful electrical nerve stimulation. Duncan, G. H. Visceral and cutaneous pain
79. Momose, Y. et al. Association studies of multiple Neuroreport 7, 321–325 (1995). representation in parasylvian cortex. Neurosci. Lett.
candidate genes for Parkinson’s disease using single 103. Becerra, L. R. et al. Human brain activation under 384, 54–59 (2005).
nucleotide polymorphisms. Ann. Neurol. 51, controlled thermal stimulation and habituation to 125. Dunckley, P. et al. Cortical processing of visceral and
133–136 (2002). noxious heat: an fMRI study. Magn. Reson. Med. 41, somatic stimulation: differentiating pain intensity from
80. Diatchenko, L. et al. Genetic basis for individual 1044–1057 (1999). unpleasantness. Neuroscience 133, 533–542
variations in pain perception and the development of 104. Tuor, U. I., McKenzie, E. & Tomanek, B. Functional (2005).
a chronic pain condition. Hum. Mol. Genet. 14, magnetic resonance imaging of tonic pain and 126. Lu, C. L. et al. Neuronal correlates of gastric pain
135–143 (2005). vasopressor effects in rats. Magn. Reson. Imaging 20, induced by fundus distension: a 3T-fMRI study.
81. Becerra, L., Morris, S., Chizh, B. & Borsook, D. fMRI 707–712 (2002). Neurogastroenterol. Motil. 16, 575–587 (2004).
Activation of the trigeminal sensory pathway in 105. Malisza, K. L. & Docherty, J. C. Capsaicin as a source 127. Yaguez, L. et al. Brain response to visceral aversive
trigeminal neuropathy. Annu. Meeting Soc. Neurosci. for painful stimulation in functional MRI. J. Magn. conditioning: a functional magnetic resonance
Washington DC, A53.16 (2005). Reson. Imaging 14, 341–347 (2001). imaging study. Gastroenterology 128, 1819–1829
82. Logothetis, N. K., Pauls, J., Augath, M., Trinath, T. & 106. Chang, C. & Shyu, B. C. A fMRI study of brain (2005).
Oeltermann, A. Neurophysiological investigation of activations during non-noxious and noxious electrical 128. Hobson, A. R. et al. Real-time imaging of human
the basis of the fMRI signal. Nature 412, 150–157 stimulation of the sciatic nerve of rats. Brain Res. 897, cortical activity evoked by painful esophageal
(2001). 71–81 (2001). stimulation. Gastroenterology 128, 610–619 (2005).

NATURE REVIEWS | DRUG DISCOVERY ADVANCE ONLINE PUBLICATION | 13


REVIEWS

129. Verne, G. N. et al. Central representation of visceral 139. Reese, T. et al. Impaired functionality of reperfused 149. Vollm, B. A. et al. Methamphetamine activates
and cutaneous hypersensitivity in the irritable bowel brain tissue following short transient focal ischemia in reward circuitry in drug naive human subjects.
syndrome. Pain 103, 99–110 (2003). rats. Magn. Reson. Imaging 20, 447–454 (2002). Neuropsychopharmacology 29, 1715–1722
130. Bonaz, B. et al. Central processing of rectal pain in 140. Jones, N., O’Neill, M. J., Tricklebank, M., Libri, V. & (2004).
patients with irritable bowel syndrome: an fMRI study. Williams, S. C. Examining the neural targets of the 150. Breiter, H. C. et al. Acute effects of cocaine on human
Am. J. Gastroenterol. 97, 654–661 (2002). AMPA receptor potentiator LY404187 in the rat brain brain activity and emotion. Neuron 19, 591–611
131. Chen, Y. C. et al. Detection of dopaminergic using pharmacological magnetic resonance imaging. (1997).
neurotransmitter activity using pharmacologic MRI: Psychopharmacology (Berl) 180, 743–751 (2005). A pivotal study in drug addiction and reward.
correlation with PET, microdialysis, and behavioral 141. Gozzi, A. et al. Functional magnetic resonance 151. Stein, E. A. et al. Nicotine-induced limbic cortical
data. Magn. Reson. Med. 38, 389–398 (1997). mapping of intracerebroventricular infusion of a activation in the human brain: a functional MRI
132. Chen, Y. C., Choi, J. K., Andersen, S. L., Rosen, B. R. & neuroactive peptide in the anaesthetised rat. study. Am. J. Psychiatry 155, 1009–1015 (1998).
Jenkins, B. G. Mapping dopamine D2/D3 receptor J. Neurosci. Methods 142, 115–124 (2005). 152. Logothetis, N. K., Pauls, J., Augath, M., Trinath, T. &
function using pharmacological magnetic resonance 142. Henderson, L. A. et al. Functional magnetic resonance Oeltermann, A. Neurophysiological investigation of
imaging. Psychopharmacology (Berl) 180, 705–715 signal changes in neural structures to baroreceptor the basis of the fMRI signal. Nature 412, 150–157
(2005). reflex activation. J. Appl. Physiol. 96, 693–703 (2004). (2001).
133. Dixon, A. L. et al. Dopamine antagonist modulation 143. Rausch, M., Gentsch, C., Enz, A., Baumann, D. &
of amphetamine response as detected using Rudin, M. A study paradigm allowing comparison of Acknowledgements
pharmacological MRI. Neuropharmacology 48, multiple high-resolution rCBV-maps for the examination We would like to thank K. Moldoff (www.galeriekirk.com)
236–245 (2005). of drug effects. NMR Biomed 18, 260–268 (2005). for kind permission to reproduce some of his graphics for
134. Hagino, H. et al. Effects of D2 dopamine receptor 144. Shoaib, M., Lowe, A. S. & Williams, S. C. Imaging this review.
agonist and antagonist on brain activity in the rat localised dynamic changes in the nucleus accumbens
assessed by functional magnetic resonance imaging. following nicotine withdrawal in rats. Neuroimage 22, Competing interests statement
Brain Res. 813, 367–373 (1998). 847–854 (2004). The authors declare competing financial interests: see Web
135. Ireland, M. D. et al. Mapping the effects of the 145. Xu, H. et al. Heroin-induced neuronal activation in rat version for details.
selective dopamine D2/D3 receptor agonist quinelorane brain assessed by functional MRI. Neuroreport 11,
using pharmacological magnetic resonance imaging. 1085–1092 (2000). DATABASES
Neuroscience 133, 315–326 (2005). 146. Lowe, A. S., Williams, S. C., Symms, M. R., Stolerman, The following terms in this article are linked online to:
136. Febo, M. et al. The neural consequences of repeated I. P. & Shoaib, M. Functional magnetic resonance neuro- Entrez Gene:
cocaine exposure revealed by functional MRI in awake imaging of drug dependence: naloxone-precipitated http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=gene
rats. Neuropsychopharmacology 30, 936–943 morphine withdrawal. Neuroimage 17, 902–910 (2002). NGF | NK1
(2005). 147. Shah, Y. B., Prior, M. J., Dixon, A. L., Morris, P. G. & OMIM:
137. Schwarz, A. J. et al. Concurrent pharmacological MRI Marsden, C. A. Detection of cannabinoid agonist http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=OMIM
and in situ microdialysis of cocaine reveal a complex evoked increase in BOLD contrast in rats using Parkinson’s disease
relationship between the central hemodynamic functional magnetic resonance imaging.
response and local dopamine concentration. Neuropharmacology 46, 379–387 (2004). FURTHER INFORMATION
Neuroimage 23, 296–304 (2004). 148. Borras, M. C. et al. fMRI measurement of CNS Alzheimer’s Disease Neuroimaging Initiative:
138. Marota, J. J. et al. Cocaine activation discriminates responses to naloxone infusion and subsequent mild http://www.loni.ucla.edu/ADNI/
dopaminergic projections by temporal response: an noxious thermal stimuli in healthy volunteers. Access to this interactive links box is free online.
fMRI study in Rat. Neuroimage 11, 13–23 (2000). J. Neurophysiol. 91, 2723–2733 (2004).

14 | ADVANCE ONLINE PUBLICATION www.nature.com/reviews/drugdisc

You might also like