Can Functional Magnetic Resonance Imaging Improve Success Rates in CNS Drug Discovery?

Abstract

Introduction

The bar for developing new treatments for CNS disease is getting progressively higher and fewer novel mechanisms are being discovered, validated and developed. The high costs of drug discovery necessitate early decisions to ensure the best molecules and hypotheses are tested in expensive late stage clinical trials. The discovery of brain imaging biomarkers that can bridge preclinical to clinical CNS drug discovery and provide a ‘language of translation’ affords the opportunity to improve the objectivity of decision-making.

Areas Covered

This review discusses the benefits, challenges and potential issues of using a science based biomarker strategy to change the paradigm of CNS drug development and increase success rates in the discovery of new medicines. The authors have summarized PubMed and Google Scholar based publication searches to identify recent advances in functional, structural and chemical brain imaging and have discussed how these techniques may be useful in defining CNS disease state and drug effects during drug development.

Expert opinion

The use of novel brain imaging biomarkers holds the bold promise of making neuroscience drug discovery smarter by increasing the objectivity of decision making thereby improving the probability of success of identifying useful drugs to treat CNS diseases. Functional imaging holds the promise to: (1) define pharmacodynamic markers as an index of target engagement (2) improve translational medicine paradigms to predict efficacy; (3) evaluate CNS efficacy and safety based on brain activation; (4) determine brain activity drug dose-response relationships and (5) provide an objective evaluation of symptom response and disease modification.

1. Introduction

1.1. Challenges in CNS Drug Discovery

There is a significant unmet medical need for therapeutics to improve the treatment of brain disorders that drives discovery programs to discover novel, safe and effective medicines (Table 1). The validation and translation of new neuroscience targets to clinical efficacy is becoming rare and the bar for the acceptance of new CNS pharmacotherapies set so high that clinical development is very long and expensive and success rates in drug registration are amongst the lowest of all disease areas [1]. As a consequence several large pharmaceutical companies have decreased their investments in the CNS therapeutic area [2] since it gives a poor return on research investment. Indeed, it has been estimated that the total discovery costs of a new medicine to the completion of phase 3 clinical trials can now take up to 15 years and cost up to $1.5BN. CNS drug discovery is particularly challenging for many reasons. Many CNS disorders have (i) slowly progressive yet ill-defined pathophysiology [3]; (ii) the clinical assessment of signs and symptoms of disease including individual differences in the responses to the disease and therapy are subjective [4]; (iii) the relative inaccessibility of the brain compared with other organ systems to biochemical investigations of disease and the effects of drug treatment [5]; (iv) the paucity of translational preclinical models [6] and (v) the uncertain relationship between early discovery studies in healthy volunteers and drug effects in patients. Currently there is a high failure rate of CNS drugs entered into clinical trials (61% of drugs still fail compared to placebo and 11% fail because of lack of efficacy or differentiation [7]; Many of these trials fail because the wrong molecules, doses or patients are studied and this could be avoided by the use of objective markers for drug target engagement and functional pharmacodynamics (PD) to ensure early proof of concept trials for a therapeutic hypothesis are valid. Recently Paul Matthews, Head of Imaging at GSK [8], noted that the drug industry has looked to major changes in structure as a solution to the problems of productivity. He suggested that the systematic use of CNS biomarkers to drive early go no go decision making is a fundamentally optimistic alternative future scenario. Embracing the idea that drug discovery and development can be made better through the use of “smart” technologies is certainly another way to think about bringing useful medicines to the right patients sooner. The broad potential of using functional imaging for drug discovery in CNS disorders (see Figure 1, fMRI in Drug Discovery) has been covered in a number of recent papers [9–11]. In this review we summarize functional, morphological and chemical imaging techniques and discuss the usefulness and current limitations of applying these techniques in CNS drug development.

Table 1.

Incidence of Functional Brain Diseases in USA

Disease	Numbers	Reference
Degenerative Disorders
Alzheimer’s Disease	4.5 million 5.3 million	[124] [125]
Parkinson’s Disease	>500,000 121/100,000 3–4 million undiagnosed	[126] [127] [126]
System Disorders
Narcolepsy	25–50/100,00	[128, 129]
Mood Disorders	20.9 million	[130]
Depression	17.1 million	[131]
Bipolar Disorder	5.7 million	[130]
Dysthymic Disorder	3.3 million	[130]
Anxiety Disorders
PTSD	7.7 million	[130]
Gen. Anxiety Disorder	40 million	[130]
Panic Disorder	6 million	[130]
Schizophrenia	2.4 million	[132]
Pain Disorders
Chronic Pain	50–70 million
Neuropathic Pain	17%	[133]
Fibromyalgia	0.5% in men 3.4% in women	[134]
Chronic Back Pain	10.2%	[135]
Osteoarthritis	16%	[136]
Migraine	12/100 in adults 11.3 million	[137] [138]
Developmental Disorders
Autism		[139]
Attention Deficit Hyperactivity Disorder	4.1 million	[130]

Figure 1. fMRI in CNS Drug Discovery.

The figure shows fMRI at the hub of research that has contributed to 4 major domains that can be integrated or interrelated (arrows) in Drug Discovery: (1) CNS Neurobiology: fMRI has contributed to understanding of CNS Pathways in humans and animals taking into account new and exciting information relating to genetics (functional genomics) brain function. (2) Applied Biology and Pharmacology: Exciting developments in using drugs in exploring neural system responses placebo. (3) Disease Process: In the domain of human Surrogate Models for disease, Disease Plasticity, Co-morbid Disease (e.g., depression or anxiety), novel insights are being reported using fMRI to evaluate these processes. (4) New Pharmacotherapies: Large investments have recently been made in academia and industry to use fMRI in Drug Discovery looking at functional effects of drugs on neural systems (see [9]). These developments will hopefully lead to the ability to evaluate drugs in early phase of development, new uses for current drugs, and perhaps contribute to a significant problem as it relates to Animal-Human Translation, where many drugs that do well preclinically, fail in the clinic. (Adapted from [161], with permission from Springer Science+Business Media).

1.2. The Need for CNS Biomarkers

The considerable time to produce new CNS drugs is largely driven by attrition during the drug discovery process making fast early disciplined decision making around the selection and validation of targets, molecules and therapeutic indications the key to any efficient research effort. Biomarkers can be used to validate targets, select molecules, specify active dose ranges, stratify patients into subpopulations and provide early evidence for supporting the mechanism of drug action, biology and clinical proof of concept. Biomarkers are central to the discovery of treatments for CNS disorders since there is often no objective diagnosis, no objective measure of treatment efficacy, and sadly no cure (Table 1).

Many different approaches have been attempted to define non-imaging based CNS biomarkers. These techniques are summarized in Table 2. Overall, about 20% of current biomarker effort is directed at CNS systems [12] to define brain diseases, characterize patient populations and assess drug efficacy and safety to help determine therapeutic windows [13–15]. To date, EEG has been the most commonly used pharmacodynamic biomarker as it has potential to reflect both CNS disease and drug effects [16]. Other biomarkers that could define the causal path between disease, drug administration and therapeutic outcome [17] are: (i) genotype/phenotype to determine disease susceptibility, drug response or potential for adverse effects, (ii) target engagement markers for receptor occupancy by drugs using receptor specific nuclear radiotracers; (iii) brain responsivity markers such as evoked potentials; and (iv) biochemical markers of disease pathophysiology and drug effects. It is noteworthy that multimodal biomarker approaches that provide orthogonal readouts on disease and drug effects are gaining interest amongst researchers [18, 19]since no one read-out may alone have sufficient predictive value.

Table 2.

Examples of Biomarker Approaches being used for CNS Disease

Potential Biomarker for CNS Disease	Example	Method of Measurement	Level of Complexity	Reference
Biochemical (CSF)	Alzheimer’s Disease Aβ(42) Tau/Aβ (42) pTau(181)/Aβ (42)	Antibody	Easily accessed via lumbar puncture	[140]
Biochemical (Plasma)	Fabry’s Disease Alpha-galactosidase A	Fluorometric method using cultured skin fibroblasts	Easily accessed; Well-defined assays	[141]
Genotyping	Familial Hemiplegic Migraine Notch3 Gene mutations of Notch3 receptor	Restriction enzyme analysis of specific mutations or by sequence analysis.	Easily accessed	[142]
Electrophysiology
EEG	Depression	Scalp electrodes; measures of brain electrical activity	Easy to use	[143]
Evoked Potentials	Alzheimer’s (P300)	Scalp electrodes; measures of response to specific stimuli including erroneous responses (p300)	Standard utilization of well-defined technique	[144]

Perhaps the most valuable biomarkers are those that can be used for the practice of medicine in new treatment paradigms for progressive disorders to enable prediction, prevention, treatment and tracking of disease driving a paradigm shift from today’s approaches that have to see disease before treating it. Obviously the most simple, least invasive and most inexpensive biomarkers that reflect the clinical rating of disease are most desirable in healthcare management but in drug discovery the value proposition is different. Here biomarkers are used to focus research activities on the patients and molecules most likely to test therapeutic hypotheses and achieve beneficial clinical outcomes [20]. The hope is that the use of biomarkers in early discovery and development, despite adding cost to early trials, would lead to fewer more expensive late stage failures by quickly eliminating the approaches that are most likely to fail. Figure 2 shows the standard Drug Discovery Flow, with potential applications of functional imaging.

Figure 2. Drug Discovery Flow.

The drug discovery process and the potential points at which fMRI could be applied to help development.(Adapted from (Borsook et al., 2006 [9]) with permission from Nature Publishing Group).

CNS drug development has many unique challenges including, the inaccessibility of the brain to simple biochemical measurements, the need for drugs to penetrate the blood brain barrier; the existence of patient heterogeneity manifest as disease subtypes; unreliability of subjective methods together with a paucity of true objective measures of the presence of CNS disease and its progression in chronic illness progression; choosing from a myriad of poorly validated targets, selecting lead drug candidates that have the best benefit-risk profiles [21]; and the prediction and early detection of efficacy in the clinic (indeed CNS drugs have the highest failure rate) without incurring significant expense to fail in late stage clinical trials; indeed, the financial costs of developing even a single drug limits the extent to which novel targets and molecules can be explored The high risk of clinical CNS drug development (see Prichard, 2008 [22]) often based on preclinical evidence and subjective measures of efficacy is such that many pharmaceutical companies are now reducing or stopping efforts in this important area of medicine that has high unmet medical need.

How then might we improve success rates in CNS drug development? What is missing from our current approaches? The brain is about behaviors and behavior is about neuronal circuits and systems and until recently with the advent of non-invasive functional imaging it has been impossible to observe the brain in action in health disease and therapy. The explosion of functional MRI techniques that now allows us to “look into functional aspects of living brain” and correlate these observations with our understanding of brain neurobiology and the effects of drugs on brain systems can add to our understanding of where and how drugs may act to produce their therapeutic effects and perhaps also give insight into the etiology of the disease itself.

PET neuroreceptor imaging has been extensively used in CNS drug development to prove brain target-engagement at safe and well-tolerated doses before clinical efficacy testing. PET has been most useful for tracking orthosteric antagonist drugs where occupancy has relationship to efficacy but is more difficult to use with agonist, partial agonists and positive and negative allosteric modulators where fractional occupancies may be low and undetectable by PET or binding sites are specific to individual chemical series such that tracers detect both on and off target binding. For CNS drugs with multiple targets PET can be used as an index of target engagement but may not reflect occupancy at all receptor subtypes involved in a clinical response. PET using [18F] Fluorodeoxyglucose (FDG) has been used as a functional marker for brain glucose metabolism that is thought to reflect neuronal activity, particularly at nerve terminals where glucose metabolism is highest, but this has had limited success due to the long time frame for data acquisition such that fast events are lost during the accumulation of [18F] FDG-phosphate. PET imaging also suffers from radiation exposure such limiting the number of times that subjects can be studied safely especially when tracers incorporate high energy radionuclides such as [18F]. These limitations of PET imaging leave significant opportunities to improve our understanding of CNS disease and therapy-using NMR approaches to understand brain function. Uniquely using functional MRI of the brain (1) subjects can undergo repeated imaging safely; (2) functional correlates of drug action on brain circuits (phMRI) can be evaluated using rapid data acquisition techniques to provide insights into drug dosing; and off-target effects (3) the effects of a drugs on brain activity evoked deliberately by specific tasks and stimuli activity can be studied and (4) the activity of brain networks in health and disease can be defined. Functional MR imaging techniques can also be correlated with NMR structural investigations to detect altered brain morphology and MRS studies of brain neurochemistry to provide orthogonal yet complimentary data to define disease and treatment effects. Today MRI is strictly not validated for drug development yet has significant potential to provide biomarkers that can change the way we discover and develop new drug therapies. We now consider where best to focus our efforts and the hurdles that have to be overcome to make this a reality.

1.3 Imaging for CNS Biomarker Discovery

There are a number of neuroimaging read-outs available for CNS biomarker discovery including molecular, functional, morphometric/anatomical, and chemical measures (Figure 3). Each technique has potential to span biomarker domains but each has its own distinct advantages and disadvantages (Table 3, adapted from Borsook and Becerra, 2010 [23]). Proof of target engagement is a key step in CNS drug discovery that provides information on receptor occupancy that is essential to guide proof of concept clinical efficacy testing. Reliable estimates of dose occupancy relationships requires knowledge of the specificity of human receptor interaction, the CNS receptor distribution of the drug target and, most importantly, the availability of receptor imaging agents. However, when PET or SPECT molecular imaging tracers that measure target engagement are lacking, functional pharmacodynamic measures, such as fMRI BOLD and arterial spin labeling (ASL), that provide an index of target engagement may provide pharmacodynamic readouts can be used to guide dosing in early clinical trials. In addition to target engagement, structural, activity (regional and circuit based networks) and chemical changes induced by drugs in the brain can be evaluated using magnetic imaging research approaches. The recent advent of PET/MR fusion imaging enables simultaneous determination of PET and MRI measures in the same subject giving the opportunity to monitor events with molecular and anatomical specificity in a single imaging session.

Figure 3. Functional, Structural and Chemical Approaches to Measures of Brain Function.

The figure shows the different approaches. Examples of data obtained for functional imaging method for BOLD fMRI, resting state networks (RSN) and pharmacological MRI (phMRI) are shown in the first three panels on the left. Morphological or anatomical methods that measure changes in gray matter (voxel-based methods or VBM) include measure of sub-cortical or cortical (cortical thickness) gray matter changes. Measure of chemical changes using magnetic resonance spectroscopy (MRS) in neural and glial systems in the brain is shown in the right panel for changes derived from specific region of interest.

Table 3.

Advantages and Disadvantages of CNS Imaging Modalities in Drug Discovery (adapted from Borsook and Becerra, 2010, with permission of Elsevier)[23]

Imaging Technology	Imaging Measures	Biological Measures	Advantages	Disadvantages	References
Positron Emission Tomography (PET)	Radioligand concentration, displacement, blood volume changes	Brain activation following or not stimulation,	Absolute changes in blood volume, receptor displacement	Requires development of specific ligand for receptor studies	[145, 146]
PET/MR	Combination of fMRI, phMRI, MRS, Structural MRI and PET	As described above	Simultaneous detection of PET and MRI measures in the same subject	Under development	[147]
Functional MRI	BOLD, CBV, CBF	Brain Activation following stimulation Brain networks in resting state	Non-invasive, repeatable, spatial, temporal resolution	Indirect brain activation measure.	[9, 50]
Structural MRI	Volumetric, fiber integrity, fiber connectivity	Atrophy, fiber degeneration	Non-invasive, repeatable, high spatial resolution	Time consuming, may require very high magnetic field scanners	[148, 149]
Magnetic Resonance Spectroscopy (MRS)	Metabolic concentrations	Alterations in brain chemistry	Can perform absolute measures, repeatable	Low spatial, temporal resolution, might be limited in certain brain structures	[149, 150]

An example of the wide range of neuroimaging biomarkers that can be used to investigate a neurodegenerative condition such as Alzheimer’s Disease is shown in Table 4. In this case, each of the measures has the potential to act a biomarker of the disease progression but they have most power when used together to test therapeutic hypotheses. For example for drug trials of potential amyloid modifying therapies amyloid PET imaging to select patients with high amyloid burden could be combined with structural MRI to detect slowing of cerebral atrophy and functional MRI to examine reversal or preservation of brain function.

Table 4.

Examples of Biomarker Potential in one CNS Disease (Alzheimer’s) using Functional Imaging

Imaging Approach	Possible Imaging Biomarker	Main Finding	Reference
Functional
fMRI	Default mode network (DMN)	Distinguishes AD from dementia with Lewy bodies	[151]
Anatomical
Volumetric	Hippocampal atrophy (HiA)	Regional measures of HiA predicted AD progression	[152, 153]
	Brain atrophy	Volumes of hippocampus entorhinal cortex, ventricle and whole brain	[154]
	Hippocampal atrophy	Smaller hippocampal and entorhinal volumes predict conversion to AD	[155]
Diffusion Tensor Imaging (DTI)	White matter integrity (Decreased FA)	Wallerian degeneration secondary to cortical atrophy	[84]
	White matter integrity	Isolation of hippocampal formation with deterioration of medial temporal pole	[156]
Chemical
Magnetic Resonance Spectroscopy	1-H Spectroscopy	Metabolic MRS profile in AD differs to Parkinson’s dementia	[157]
		Metabolic changes detectable in presymptomatic mutation carriers of AD	[158]
Ligand Binding
Positron Emission Tomography	11C-PIB PET (Amyloid ligand)	Mapping Amyloid Toxicity – Temporal pole susceptibility	[159]
		Reductions in cerebral metabolic rate for glucose after onset of AD	[160]
	18F-Florbetapir PET	Defines β amyloid plaques	[26]

2. Imaging Technologies with Potential for CNS Biomarker Discovery

2.1 PET and SPECT nuclear imaging – benefits and limitations

Radionuclide imaging modalities, particularly PET and SPECT, have become powerful tools for CNS drug development. Positron emission tomography (PET) and single photon emission computed tomography (SPECT) are nuclear imaging techniques that can be used to visualize the distribution of radiolabeled drug – like molecules (radiotracers) and their interaction with protein targets on or within cells [24]. Nuclear imaging uses tracers labeled with positron emitting radionuclides (PET: most commonly 15O, 11C and 18F), or with gamma-emitting radioactive isotopes (SPECT: e.g., 123I). Both can be used to track small molecules and biologic therapeutics. Radiotracers are versatile; they can be designed to track the drug itself, image the drug target or monitor key biochemical and physiological processes. Radiotracers are highly sensitive and today are the only way to measure receptor populations and pharmacology (at picomolar to nanomolar densities) quantitatively in vivo in both animals and humans. The development of small animal tomographic cameras (microPET or microSPECT with computer tomography) has facilitated translational bridging between preclinical and clinical central nervous system (CNS) research. PET based mapping of neurotransmitter receptors in the brain can provide novel molecular information on regional brain receptor densities that can be correlated with functional data and cytoarchitectural or anatomical maps [25] to generate hypotheses on neurotransmission in health and disease. PET neuroreceptor imaging however has to be grounded in measures of brain function and ultimately requires therapeutic intervention to verify the role of specific receptor populations in the presentation of any CNS disease. PET can image neuropathology (e.g., amyloid in Alzheimer’s disease) enabling enrichment of trials of amyloid modifying therapies with patients with high CNS amyloid loads but its true value as a diagnostic in medicine requires an effective therapy that is lacking for many neurodegenerative diseases today [26]. In drug discovery research the synthesis of novel PET tracers relies upon parallel discovery efforts in medicinal chemistry and radiochemistry to bring drug candidates and tracers forward together. Early discovery of a PET tracer helps selection of orthosteric antagonist lead molecules in the preclinical lead optimization phase by focusing research on those drug candidates that achieve the highest target engagement off the lowest exposure at safe and well tolerated doses thereby maximizing the potential therapeutic safety window. In some instances the ability to test therapeutic hypotheses at high occupancies may be limited by unwanted off-target effects [27] or have a significant beneficial contribution from receptors other than those being imaged [28]. An additional complication for some mechanisms is the existence of a complex dose response relationship often as a result of dose limiting side effects [29] that can differ between animals and man such that the occupancy targets set for optimal efficacy from preclinical experiments do not translate to humans. To have maximum value, PET tracers need to be clinically validated in advance of phase I clinical studies so that they can be incorporated into early pharmacokinetic, safety and tolerability study paradigms. Data that determine how hard and how long a drug must hit its target to produce the desired behavioral/pharmacological effect are critical to later clinical trial designs. If a drug candidate sufficiently engages its target in vivo at safe and well-tolerated doses but does not produce the expected biological or clinical effects, the therapeutic concept is flawed and clinical development may be stopped. The drawbacks of neuroreceptor PET are threefold: (i) occupancy of a target by a drug isn’t efficacy – it just allows you to test your therapeutic hypothesis; (ii) that it is not always possible to create a radiotracer for CNS receptor targets; and iii) occupancy does not indicate effects on brain function especially for agonists, partial agonists and allosteric modulators. In all these cases magnetic resonance brain imaging can play a powerful role in providing a functional read out that reflects disease and health and a pharmacodynamic readout that can serve as an index of drug receptor interactions.

2.2 Functional Brain Imaging Biomarkers

2.2.1 BOLD fMRI

Blood Oxygen Level Dependent (BOLD) fMRI provides an index of neuronal activity [30, 31]. BOLD fMRI imaging detects changes in the concentration ratio of oxy-and deoxy-hemoglobin as a result of neuronal activity-induced changes in blood flow and volume [32, 33]. BOLD does not provide an absolute value of blood flow but a relative change from baseline and so the approach does have some limitations [34]. BOLD imaging can nevertheless be used to produce signatures of brain activity that reflect disease state and drug effects triggered by a wide variety of stimuli [30, 35]. fMRI BOLD imaging has been used to assess brain responses in cognitive, affective and neurological disorders [36]; Its use in CNS research has been particularly effective because of its ability to evaluate well-characterized stimuli in block “on-off” experimental designs contrasting the effects of a challenge to a control activation state. BOLD fMRI has been used to define brain function in cognitive processing in schizophrenia [37] and sensitization during migraine [38].

2.2.2 Resting State Networks (RSN) and Functional Connectivity

RSN’s measurements evaluate low frequency BOLD signal fluctuations among specific neuronal networks that contribute to different brain functions [39]. An example of a resting state network is the so-called default mode network (DMN). DMNs are a set of regional activation networks that reflect resting brain functions related to orientation and interpretation of environment, monitoring and reporting state of the self. Changes in these networks can be used to differentiate brain states. RSN’s can differentiate cohesive brain networks (DMNs) in healthy subjects (that are consistent across healthy subjects [36, 40]and in disease states [41, 42]. It is believed that in disease and in the presence of a drug, some correlations between brain areas are either suppressed or enhanced [43]. Evaluating brain conditions in terms of networks seems more rational than evaluating individual brain regions for biomarker potential, given the complexity of the brain’s structural and functional connections that underpin behavior. Functional connectivity analyses give the opportunity to interpret behaviors associated with drugs, disease, or provocative testing of healthy volunteers and patients by monitoring up or down modulation of the strength of activity in defined neural circuits associated with the behavior. Functional connectivity approaches can help understand how specific brain structures link to each other and how this connectivity may be modulated by disease states or the actions of drugs [44]. Functional connectivity analyses may therefore provide the best avenue to discover CNS biomarkers of brain activity [45–47]. Connectivity changes have for example been reported in diabetic neuropathic pain, where thalamocortical activity is altered in patients vs. controls [48].

2.2.3 Pharmacological MRI (phMRI)

Two types of fMRI approaches are being used in the functional evaluation of drugs (phMRI). These are BOLD (see above) and arterial spin labeling (ASL). In contrast to BOLD, ASL measures only blood flow changes through magnetic spin labeling or tagging of the iron in red blood cells in the proximal blood flow which for the brain is the carotid artery [49, 50]. A control and spin labeled image is subtracted to produce a perfusion-imaging map. The main advantages of ASL over BOLD is its improved contrast and signal to noise ratio [50] and its ability to quantify blood flow [51]but it has lower sensitivity.

BOLD and ASL phMRI can evaluate the direct effects of a drug on resting state brain circuit activity [52, 53] as well as the effects a drug may have on regional or circuit based activation evoked by specific stimuli (see Bifone and Gozzi, 2011 [54]). For both these phMRI approaches, brain activity changes can be correlated with the distribution and occupancy of specific CNS receptor populations by drugs. Taken together these measurements can be used as a pharmacodynamic marker of the functional and anatomical consequences of drug target engagement. phMRI also enables evaluation of drug dose response relationships in the brain – and this may serve as a better index of overall target engagement than that provided by molecular receptor imaging approaches such as PET especially when drugs have more than one site of action within the brain.

Functional imaging approaches can be used in preclinical species. Understanding the effects of drugs on brain circuits could provide a “language of translation” [9], that could improve the predictability of the transition between preclinical to clinical models. A number of recent reviews address this issue [23, 55–57]. Preclinical imaging in drug discovery has now advanced to include: (i) an ability to image awake, trained animals, for phMRI experiments and activation experiments using CNS stimuli [58, 59]; (ii) defining new insights into CNS processing in novel pharmacologically based models of disease [60]; (iii) evaluation of the similarity of responses (e.g., drug effects, sensory stimuli) across species [61]; and (v) improved selection of candidate CNS drugs in development by mapping the central activation patterns they produce [62].

Figure 4 provides examples of how fMRI may be used to optimize drug evaluation. (1) Real time imaging of drug agonist antagonist interactions can be evaluated (Figure 4A) as shown using the μ opioid agonist morphine [52] and naloxone [53]. This ability to detect pharmacological events with high spatial resolution within a specific region is an example of the unique specificity of the approach. Activation along known CNS pathways such as the sensory system from ganglion to cortex [63] can also be imaged with high sensitivity and specificity. (2) The translational value of using MRI to help discover drugs is shown in Figure 4B. In this study significantly fewer subjects were required to differentiate drug responses in humans compared with subjective ratings and similarities of drug action were observed across species [61] supporting its value as a translational measure. (3) In Figure 4C, we show a proof of concept example of how fMRI could be used to image drug receptor interactions when PET ligands are unavailable or occupancies too low to image. The images show how well the phMRI activation map for morphine [52] corresponds with a PET receptor evaluation using carfentanil [64]. (4) in Figure 4D we illustrate the translational value of fMRI across species particularly for discrete physiological stimuli such as pain responses to noxious stimuli. The panels show how well the increases or decreases in fMRI signal to a noxious stimulus applied to the dorsum of the foot in humans and rats correspond across species [61].

Figure 4. Examples of Applications of Functional Imaging Approaches.

A: Opioid Agonist-Antagonist phMRI Activation. The top portion of the figure shows increased (red) or decreased (blue) activation patterns in the brain following morphine 4 mg/70 kg (adapted from [52] with permission of Wolters Kluwers Health) vs. naloxone 4 mg (adapted from [53] with permission of the American Physiological Society). An example of actual significant activation is shown in the coronal sections below with increased activation for morphine and decreased in naloxone in the orvitofrontal gyrus. Schematic BOLD responses are noted next to each figure. Figures derived from original data sets.

B: Differentiating Analgesic Drugs with fMRI. The figure shows 3 components to the effects of various analgesics (imipramine (50 mg), gabapentin (600 mg), clonazepam (0.5 mg), ketorolac (10 mg) and rofecoxib (25 mg), and vs. placebo) to a thermal stimulus (stressor) (Adapted from [61], with permission of John Wiley and Sons):

Subjective Ratings – Subjective rating of pain (VAS 0–10 where 0 is no pain and 10 is the maximal pain they could imagine) shown in the graph cannot differentiate between drugs following noxious heat. The data is presented as mean rating ± SEM.

fMRI Response – Sample axial slices depicting activation maps for two drugs (imipramine and clonazepam) and placebo. Note that there is an overall decrease in activation for imipramine vs. placebo and an overall increase in activation for clonazepam vs. placebo.

Quantitative Ratings - Voxel count for 5 drugs vs. placebo for whole brain (WHB) activation. Note that for imipramine and gabapentin more voxels are activated in drug vs. placebo while for clonazepam, rofecoxib and ketorolac more voxels are activated in the drug vs. placebo. Topiramate has an intermediate or mixed effect.

C: Opioid phMRI and PET studies. The figure shows phMRI maps (z > 2.3) following infusion of morphine (4 mg/70 kg) in healthy adult male subjects [52]. Note the commonality of activations for morphine [52]and [11C] carfentanil[64].

Key: 1 – anterior cingulate; 2 – caudate nucleus; 3 – frontal cortex; 4 – putamen; and 5 – thalamus for phMRI morphine [52] vs. carfentanil [64]. Dotted white lines indicate limits of brain acquisition for scans. This is adapted from [52] with permission of Wolters Kluwer Health and from [64] with permission of Nature Publishing Group.

D: Translational Measures across Species. The figure shows the thermal response of rats and humans to a 46°C stimulus applied to the dorsum of the foot. Note that the activation patterns in the regions of interest (primary somatosensory cortex (SI), thalamus (Th), insula (I), anterior cingulate cortex (aCG) and amygdala (A)) is similar to that shown for specific regions known to be involved in pain. Note the signal sign (i.e., increase or decrease in BOLD signal) is similar in both species. Furthermore, two peaks are noted in the thalamus (insert) in both species. This provides some evidence for at least partial equivalence of response in across species. (From [61], with permission from John Wiley and Sons).

2.3 Anatomical and Chemical CNS Biomarkers

2.3.1 Volumetric Imaging

The utility of MRI measures of anatomical variation in disease states is currently being evaluated for Alzheimer’s Disease through the fNIH Biomarker Consortium in the Alzheimer’s Disease Neuroimaging Initiative [65, 66]. Measurements of hippocampal and entorhinal cortex, ventricular volume and cortical thickness are central to biomarkers of this neurodegenerative disease. The ADNI will also yield some generalizable operational observations about the reliability and reproducibility of MRI for determining the usefulness of longitudinal volumetric brain biomarkers. The diverse range of computational and brain segmentation techniques used to measure regional gray matter volume using voxel-based morphometry or cortical thickness has been reviewed in detail elsewhere [67]. It is noteworthy that high levels of consistency have been shown using automated segmentation algorithms to estimate regional brain volumes [68]. Of particular interest is a technique of “ballooning” out images of the brain computationally to enable measurement of very small changes in regional cortical thickness and volume (http://surfer.nmr.mgh.harvard.edu). To date voxel-based morphometry approaches have been used mostly to measure changes in the volume of subcortical structures in health and disease [69–71] and has revealed changes in amygdala and hippocampal volumes in chronic depressive illness [72]. Chronic pain also clearly changes brain morphometry [73–75], making structural MRI a promising candidate biomarker measure for pain disease progression and modification by analgesics.

Morphometric biomarkers therefore have potential to (i) follow disease progression in neurodegenerative disease; (ii) differentiate subtypes of disease; (iv) assess therapeutic efficacy (e.g., slowing down or reversal of gray matter loss) and (v) indicate duration of disease [76]. A relatively new and exciting use of MR is postmortem volumetric and anatomical imaging of the brain at very high spatial resolution to pinpoint abnormal brain regions for further intensive biochemical, anatomical and immunohistochemical investigations [77].

2.3.2 Diffusion Tensor Imaging

DTI is a MRI technique that detects microstructural changes in white matter tract integrity that cause changes in the diffusion properties of water within axonal projections throughout the brain (anisotrophy) [78–80]. The technique has been used in a number of clinical disorders including multiple sclerosis [81], depression [82], drug abuse [83] and Alzheimer’s Disease [84]. Although there are some limitations to DTI [85], when combined with fMRI, it may help provide new insights into brain signal processing in health and disease [86].

2.4 Chemical CNS Measures for Biomarkers

2.4.1 Magnetic Resonance Spectroscopy (MRS)

Changes in brain chemistry as a result of disease [87, 88] or drug action [88, 89] can be measured non-invasively using MRS. Several CNS neurotransmitters (e.g., glutamate, GABA, glycine) and brain metabolites (e.g., NAA, choline) that reflect excitatory and inhibitory neuronal activity and neuronal health status can be measured simultaneously. MRS can also measure flux within metabolic and neurotransmitter pathways by the incorporation of MR sensitive nuclei in biochemical substrates used in brain function. These include: (i) carbon (13C) [90]; to monitor brain metabolism and neural transmission during functional activation in the brain[91, 92]. 13C MRS spectroscopy has been used to detection of metabolism at a high spatial and/or temporal resolution and to measure glutamatergic (excitatory) and GABAergic (inhibitory) neuronal activity; (ii) phosphorous (31P) [93] to study high energy phosphate metabolism and (iii) sodium (23Na) [94] to assess tissue sodium concentration as it changes significantly in stroke and tumors. The utility of MRS has been explored in various neurological and psychiatric disorders [95–97]. MRS may be a sensitive way to detect changes in brain chemistry and function before pathology occurs [96, 98] and to differentiate between disease sub-types of CNS disorders.

3. Adopting CNS Biomarkers

3.1 Practical Challenges

Objective biomarkers of CNS function have the potential to augment subjective assessments of CNS disease. Subjective assessments are often highly variable and so can lack power to guide novel approaches to improving current CNS therapies. The advent of new multimodal neuroimaging techniques (see Section 2 above) has undoubtedly opened a new era of CNS biomarker discovery (Figure 2). Current neuroimaging techniques are undoubtedly complex, and there are diverse practical challenges in using them to characterize patients, identify responders, monitor drug actions and define therapeutic outcomes. Imaging requires specialized equipment and trained individuals but most importantly its successful implementation depends upon standardization of measurement procedures and analytical processes to support routine use.

Biomarkers have been defined as “Biological substances or features that can be used to indicate normal biological processes, disease processes, or response stotherapy” (www.biobankcentral.org/resource/glossary.php). Successful biomarker development requires a rigorous process of scientific validation and clinical qualification. As for any biomarker [15], CNS imaging markers should (i) predict or differentiate an individual’s function within a generically defined group of volunteers, patients or class of disease; (ii) have a high degree of specificity; (iii) have a high degree of reproducibility with and across subjects i.e., a reliable/consistent response of the measured response in brain function/behavior to the drug; (iv) be able to be performed and measured easily; (v) be defined in the context of the underlying neurobiology of the specific disease in question; (vi) be validated; and (vii) have a low risk when performing the evaluation. The value to using healthy individuals for testing the clinical efficacy of drugs using specific biomarkers, is a subject of debate, since some brain systems undergo neuroplastic adaptive changes in progressive CNS diseases [99, 100] and these are essentially unchanged and intact in the absence of disease. The concomitant use of other susceptibility markers (e.g., genetic or epigenetic measures) in healthy volunteers may however help define populations that could be the basis of trial enrichment in the early phases of drug discovery.

Biomarkers undergo a “fit for purpose scientific validation and clinical qualification process (see Figure 5 [101]. Qualifying a biomarker for use in regulatory approval processes can be difficult even when a biomarker is scientifically validated and well defined. Indeed most biomarkers are used at risk, compared to clinical outcomes, for decision-making during drug discovery and development and few ever reach the level of surrogacy where they can substitute for a clinical outcome. Functional brain imaging alone is unlikely to improve success rates in the discovery of CNS drugs. However, when integrated with clinical subjective assessments it can provide an objective view that together with patient and physician reports may improve decision making in early clinical trials.

Figure 5. Biomarker Development.

Left Panel: The panel shows the classic drug development pipeline flow from initial discovery in the laboratory to the drug use in the clinic. Some of these processes are detailed in Figure 2.

Middle Panel: The panel shows CNS Biomarker Evolution from initial identification of the potential biomarker through processes that include exploration, demonstration, classification and it sues in the clinic before becoming a diagnostic in general medical and research use. Note that as progress toward a diagnostic use becomes more defined increasing levels of evidence for the biomarker are required (see Text and see [101])

Right Panel: The figure shows biomarker process (lifecycle) from initial definition to adoption. An initial observation suggesting a Potential CNS Biomarker may be observed in a small study that then needs to be evaluated in a larger Clinical Trial that contributes to Validation. Validation, demonstrating specificity and sensitivity of the biomarker assay is followed by Qualification (with demonstration of robust reproducibility) of the biomarker and then the required Regulatory Adoption. Once adopted, the process of Continued Evaluation of the biomarker defines its continued status (Continued Evaluation) that includes potential refinements as technologies and larger clinical datasets become available.

3.2 Scientific Challenges

Specific diagnoses of most CNS diseases are relatively straightforward in their fulminate condition. Diseases that affect the brain are usually progressive and sometimes dynamic producing alterations in brain phenotype and perhaps changing the targets for drug effects. Differences in brain state clearly change over time (e.g., early stage disease vs. fulminate disease state) and drugs themselves may induce changes in the brain through halting, reversing or altering disease pathophysiology [44], such that treatment history could impact the usefulness of functional biomarker measures to predict future efficacy and treatment response.

New developments in functional brain imaging will be able to define specific phenotypic markers at two ends of the clinical spectrum – Healthy Brain State vs. Diseased Brain State. However, early diagnosis and differential diagnosis during chronic CNS disease are difficult in many CNS disorders and it is likely that establishing neuroimaging biomarkers that reflect changes brain function will require parallel development of more sensitive clinical rating scales to anchor them in disease symptomatology. Further complicating factors are that brain activity is influenced by gender, genetic make-up, age, and new adaptive environmental experiences that impact memory, affect and learning [102–105]. For example in chronic neuropathic pain, the brain may be in transitional or in “altered modes” as the disease progresses. Such transitional states can result from a complex interaction between a number of factors including time of evolution of the disease, the effects of therapy, together with changes emotional, cognitive, and sensory state that can influence a patient’s subjective experience of pain [106–108]. Given that many of these factors underlie the subtleties of CNS disease subtypes and affect brain function in their own right, it is unlikely that functional neuroimaging biomarkers will alone be useful to describe CNS disease state. A combination of objective functional neuroimaging markers with other measures such as subjective clinical assessments, biochemical markers and genetics will be required to obtain the specificity required to detect and quantify the effects on drugs on CNS disease.

In the future the development and application of validated CNS imaging biomarkers may support personalized or stratified medicine approaches by identifying patients that will most likely respond to drug treatment or be a candidate for an adverse reaction to treatment [109]. In this way, clinical practice for some patients could be radically changed although the use of functional neuroimaging in a general healthcare setting would be severely limited by the high costs of implementation.

4. CNS Imaging Biomarkers – Why, What and How

4.1 The promise of CNS imaging biomarkers

We have discussed the potential of imaging biomarkers to improve CNS drug discovery but the implementation of an imaging strategy is expensive and will have to produce a return on the investment it requires. The value proposition is that objective CNS biomarker can speed decision making to advance, accelerate or stop drug discovery programs enabling ever scarcer research resources to be focused purposefully on those areas with the highest probability of success to deliver a useful therapeutic agent. We need then to consider what barriers remain to the pharmaceutical industry adopting and integrating CNS imaging biomarkers into their research operating plans in drug discovery today?

4.2 The Barriers to CNS biomarkers

The first barrier is undoubtedly the cost and time taken to validate and qualify the CNS biomarker measure. When biomarkers that are linked solely to proprietary molecular targets (e.g., a PET tracer) it is up to each company to judge the value to them of early indicators of target engagement and proof of biology compared to traditional outcome measures. Companies execute biomarker strategies using in house imaging capabilities or through links to academic and medical centers of excellence where studies are managed by dedicated in house expertise and/or specialized imaging clinical research organizations (CROs). These proprietary CNS biomarkers however rarely have applicability across different targets within the same disease area and the cost of developing them has to be justified and accounted for in the costs of specific drug discovery programs. In contrast, functional imaging biomarkers can generally be considered platform technologies that can characterize patient populations, disease state and response and so have potentially broad cross target utility with value to multiple companies pursuing diverse therapeutic approaches within a common disease area. In this case the second barrier is that there is little incentive to any one company to bear the considerable cost (in the case of the Alzheimer’s Disease Neuroimaging Initiative consortiums ADNI1 and ADNI2 a total of ~ $100 million) of the biomarker discovery programs and so a different shared solution is required.

4.3 The Solution

The recent shift to the formation of multimillion-dollar public-private consortia in Alzheimer’s disease and Parkinson’s disease to develop and standardize measurements of CNS biomarkers that can characterize patient populations and provide robust baselines for therapeutic trials have been an advance in the platform biomarker field. In the USA the foundation for the National Institutes of Health (fNIH) and in Europe through the European Medicines Agency (EMEA www.ema.europa.eu) or smaller groups such as the Imaging Consortium for Drug Development (ICD) [110] have championed these approaches. Indeed, the setting up of the Biomarker Consortium at the fNIH (http://www.fnih.org/) was a direct response to the FDA Critical Path initiative that called for a greater emphasis on biomarkers that could be deployed to speed the discovery of safe and effective targeted drugs for many common diseases including CNS disorders. The fNIH provides a neutral forum to bring together the National Institutes of Health (NIH), Federal Drug Agency (FDA), CMS, Pharmaceutical Organizations (PhRMA), Diagnostics companies, BIO, Patient Groups and Academia with the goal of definition and adoption of biomarkers. Once established, these platform biomarkers can be used to define responsive populations and perhaps to provide an early comparative assessment of the likely effectiveness of new mechanism drugs against established therapies. In the future, objective measures that define the most appropriate patients to treat or those who are most likely to benefit will become an important part of personalized medicine and the drug application, review and approval process.

4.4 The Key to Effective Consortia

The advent of precompetitive biomarker consortia is a welcome advance inmedical research but the execution of this strategy has several challenges that require strong yet inclusive management of all participants from the outset of a project proposal to its execution. These include: (i) aligning on common goals of what success looks like; (ii) preventing different institutions agendas from splitting and diluting the focus of the research effort to the point where impact on the field or interest to prospective members becomes minimized; (iii) changing of key personnel particularly in long term projects such as those following disease progression; and (iv) a difficult economic environment where projects without short-term proprietary benefit are prioritized against and financial considerations preclude commitment to long-term funding.

5. Future Progress

We see future progress in functional CNS imaging biomarkers and successful proof of concept for its value in drug discovery being driven by the following domains(see Figure 5: Examples of Applications of NMR Approaches):

5.1 Preclinical Imaging

The use of preclinical imaging to evaluate CNS drug candidates has become more common, particularly with the ability to image awake trained animals for phMRI (see above). The approach allows for novel observations of ‘CNS pharmacology in action’ and the generation of objective clinical hypotheses for CNS biomarker evaluations in drug discovery [60, 111–113].

5.2 Translational Clinical Medicine [114]

Neuroimaging the response of brain regions and circuits to stimuli and drugs can improve basic our understanding of behavior in preclinical and clinical neuroscience, highlighting potential markers of disease and drug response and directing future drug discovery efforts.

5.3 Reengineering Early Clinical Trials

The use of functional neuroimaging will increase as part of discovery and early discovery strategies for CNS therapies to provide objective readouts of CNS drug effects that can fuel go no go decision making for molecules and hypotheses (target engagement, CNS activity dose-response, regional activation and network connectivity, signature of potential side effects) The most advanced use of biomarkers in early discovery is in the field of oncology [115, 116]. In neuroscience, PET neuroimaging is now widely used to evaluate drug delivery to CNS targets (Hargreaves 2008), EEG to assess brain pharmacodynamic response and brain anatomical volumetric and molecular measures to stage neurodegenerative disease pathophysiology but other functional measures still require validation and qualification.

5.4 Demonstrating Value

Proving that CNS biomarkers can improve the cost effectiveness of CNS research is the critical issue for the adoption and of functional neuroimaging by the pharmaceutical industry. Future use will be driven i) by a positive return on investment through the demonstration of the real value of increasing early discovery spending in terms of better and faster decision making to decrease late stage attrition costs, ii) improving scientific insights around molecules and hypotheses perhaps identifying alternate indications for CNS drug therapies, iii) improving patient selection to decrease variability and increase signal to detect beneficial therapeutic effects and iv) using imaging biomarkers to support proof of biology and clinical proof of concept [117].

6. Expert Opinion

6.1 Current State

Objective biomarkers of CNS function have the potential to augment subjective assessments of patient symptomatology that are often highly variable and so sometimes inadequate for guiding many approaches to improving therapy. There is, therefore, a clear unmet medical need to develop better biomarkers for evaluating drug effects and disease state to advance therapy.

Functional imaging has provided new and exciting insights to CNS disease states [9]. Several approaches in imaging now have the potential to become CNS imaging biomarker tools. Integration of these functional, anatomical and neurochemical imaging techniques undoubtedly gives the potential of increased power to evaluate disease state and drug effects in brain diseases. The challenge of validating and qualifying imaging biomarkers for brain disease states (e.g., depression, chronic pain, schizophrenia) is not trivial and reducing them to practice in their simplest form to support cost effective and timely biomarker approaches to diagnosis, treatment and CNS drug discovery will probably require “big science” consortia based initiatives that are expensive to complete (see [118]). CNS biomarkers however promise to transform our current position of empirically based judgments to more rational scientific approaches based on objective markers that can augment our subjective analyses of brain diseases.

6.2 Future State – Opportunities to do Better

Some authors have justifiably added notes of caution to the excitement around the use of biomarkers [15]. “Biomarkers that are truly Found in Translation” [119] have the potential to become central to drug discovery and may impact clinical practice. CNS biomarkers enable the characterization of patient populations and quantification of the extent to which new drugs reach their intended targets, alter proposed pathophysiological mechanisms and achieve clinical outcome [120]. CNS biomarkers have the potential to inform decision making for pharmaceutical companies and regulatory agencies with respect to candidate drugs and their indications in order to bring safe and effective new medicines to the right patients more rapidly than they are today [120].

6.3 The Benefits of Neuroimaging Biomarkers

Several aspects of CNS drug discovery could be impacted positively by the discovery of functional neuroimaging biomarkers: (i) Improved lab to clinic translational approaches that are objective; both forward and backward translation could allow for the characterization and development of preclinical assays in the context of human brain disease conditions and pharmacological effects; (ii) More efficient early phase clinical trials (Phase 1) the data produced can define, target engagement CNS dosing and potential side effects and help evaluate candidate drug molecules for go-no go decisions; (iii) Potential, if objective and standardized methods can be defined, to a) use smaller numbers of similarly characterized patients to reduce the noise and increase detection signals in proof of concept clinical trials and b) use objective inclusion criteria to define patient cohorts for larger multicenter clinical trials; (iv) Ability to make frequent repeated measures to track progressive conditions over time allowing for longitudinal measures of the effects of drugs on symptom and disease modification because MR imaging is completely non-invasive and no radiation exposure is required; and (v) Support for faster CNS drug approvals and therapeutic indications through focusing on the right molecules, clinical hypotheses and patient populations

6.4 The Risks of CNS Biomarkers

Whilst the use of CNS biomarkers could improve clinical practice, therapy and the discovery of novel CNS drugs their adoption will depend on a number of issues including their sensitivity, specificity, costs, and, as support for drug registration, regulatory acceptance. The use of any biomarker involves increased risk compared to measuring true clinical outcomes and the degree of risk varies dependent on when they are used for decision-making. Biomarkers need to be available at the right time with the right fit for purpose level of scientific validation and clinical qualification to impact drug discovery and development. Biomarkers that come too late get ignored and so incur expense with no return. During early drug discovery decision-making based on biomarkers is a discovery risk that is assumed by companies to reduce costs and cycle time by getting an early focus on the molecules and clinical hypotheses having the highest probability of success. Consistent early implementation strategies that span CNS drug discovery portfolios spread the risk of decision making based on biomarkers rather than clinical outcomes and the overall gamble is that decisions made using biomarkers will on average be better than those made in their absence. In clinical practice the use of biomarkers to guide therapy carries different relative risks in terms of defining safety and effectiveness in targeted diseased populations.

6.5 Use of Research Resources

The discovery and development of imaging biomarkers especially for novel therapeutic mechanisms require significant resourcing. It can add time and expense to drug discovery particularly in the early stages but by eliminating potential failures early has the bold promise of focusing resources and research on higher quality molecules and proven hypotheses for clinical trials. There are multiple opportunities to integrate neuroimaging into preclinical, early and late clinical trials to improve decision making in drug discovery [6; see Figure 3]. Current measures of brain function using imaging are undoubtedly complex and require specialized and trained individuals and equipment. However, like many imaging processes, these obstacles can be overcome through purposeful harmonization of data acquisition techniques and the development of “turn-key” applications using standardized tools for data collection and analysis. The implementation of imaging as a strategy in novel CNS drug discovery will involve building (especially for preclinical work) or accessing sophisticated imaging capabilities and/or forming networks and consortia that link companies, academic centers of excellence and commercial imaging operations (e.g., clinical research organizations (CRO’s)) [110, 121–123].

6.6 CNS Imaging Biomarkers -Reality or Another Broken Promise

Only time will tell when, where and which CNS imaging biomarkers provide most value and a return on the investment they require but recent internal reviews seem to indicate that they can provide value [11]. To date there has not been a reported breakthrough in which NMR imaging has been transformative. Clearly a success would add impetus to the adoption of neuroimaging strategies as part of CNS drug development. Conceptually, the use of imaging biomarkers remains a very big experiment whose outcome will not be known for several years until the novel molecules selected using biomarker strategies are fully evaluated clinically and the outcome of biomarker guided therapeutic strategies is known. Let’s hope the outcome is positive!