3

Target Identification

Target identification is a critical step in the drug development pipeline. Speakers and participants discussed ways to improve this starting point and thereby accelerate therapeutic development through the use of stem cells, humanized animal models, increased knowledge of genetics, and imaging.

STEM CELLS

Stem cells, which are undifferentiated cells that can divide for indefinite periods of time and differentiate into specialized cells, may serve as a faster tool for the discovery of novel targets. Embryonic stem cells are cells that are derived from embryos that can develop into any cell and tissue throughout the body, whereas somatic (also referred to as adult) stem cells are cells that naturally exist in the body, responsible for maintaining and repairing tissue (NIH, 2002). Induced pluripotent stem cells (iPSCs) are somatic stem cells artificially reprogrammed with transcription factors to express pluripotent properties of embryonic stem cells—meaning they have the potential to differentiate into mature cells of all types. Because embryonic stem cells and iPSCs exhibit inherited traits and disorders, researchers can use them to better understand the underlying mechanisms of diseases and create disease models more representative of the actual disease than animal models (Rubin, 2008). Particularly for target identification, comparing patient-derived iPSCs to normal cells may offer researchers the ability to detect when physiological deviations occur at the cellular level (Grskovic et al., 2011; Yang et al., 2011). This, in turn, could improve identification of drug targets. Speakers provided examples of this for Alzheimer’s disease (AD) and amyotrophic lateral sclerosis (ALS), in addition to the use of stem cells to create humanized animal models.

Induced Pluripotent Stem Cells

iPSCs provide a valuable tool for studying diseases and understanding pathways needed to develop potential therapeutics, said Lawrence Goldstein, director of the University of California, San Diego (UCSD), Stem Cell Program and distinguished professor in the department of neurosciences at the UCSD School of Medicine. When differentiated into neurons and other nervous system cells, iPSCs can potentially overcome the translational limitations of animal models. iPSCs enable analysis of human-specific phenotypes that cannot be modeled in animals (Young and Goldstein, 2012). Lawrence Goldstein highlighted several benefits of using iPSCs; they

•   can be produced in unlimited quantities because of their capacity for self-renewal;

•   carry uniquely human biochemistry;

•   are euploid (contain the normal number of chromosomes) and are genetically stable;

•   capture the range of human genomic variation;

•   can be studied electrophysiologically;

•   are useful for basic science, disease modeling, and drug screening; and

•   can be manipulated, a beneficial attribute for mechanistic studies.

Although there are benefits to using iPSCs, a few participants noted that iPSCs may be simply another step in drug development rather than a means to truly accelerate the process to first-in-human trials. Particularly, several participants indicated that there could be significant benefit from defining the iPSC phenotypes that would be most suitable for study of human nervous system disorders. A participant noted that cellular readouts may be difficult to obtain for neuropsychiatric disorders, and, as a result, researchers might turn to physiological and imaging readouts to subset patient populations. A few participants noted that although there will be an influx of useful genetic data pouring into the field, researchers are, at times, unaware of specific disease phenotypes—making drawing conclusions from data challenging. As mentioned in Chapter 2, understanding phenotypes, in several participants’ opinions who spoke, is necessary to improve drug discovery.

Alzheimer’s Disease

Lawrence Goldstein and colleagues recently demonstrated that iPSCs could be successfully reprogrammed from human fibroblasts for the purpose of modeling AD (Israel et al., 2012). This was accomplished with fibroblasts from two patients with familial AD, two with sporadic AD, and two normal controls. The iPSC cultures differentiated into neural precursor cells and then neurons, which were harvested by cell-sorting technology. The cultures consisted of 90 percent neurons, many of which formed synapses and displayed normal neuronal electrophysiological activity. The neuron cultures from patients with familial AD and one with sporadic AD displayed higher levels of three proteins implicated in AD pathophysiology: β-amyloid, phosphorylated tau, and glycogen synthase kinase 3 (GSK3β). These findings model the biochemical phenotypes of AD. However, the cultures did not contain glial cells. In addi-

tion, the cultures contained more than one neurotransmitter type with a mix of γ-aminobutyric acid (GABA), glutamate, and some peptidergic neurons. Given differences in relative susceptibility of neuronal populations in AD, the ability to obtain cultures with one neurotransmitter phenotype containing glia is needed to control for variability, noted Lawrence Goldstein. He and his colleagues are working to overcome these limitations.

Lawrence Goldstein’s laboratory is also using iPSCs to study the role of sortilin 1 (SORL1), a trafficking factor, which is a susceptibility gene for late-onset AD (Lee et al., 2008). SORL1 protein levels control the rate at which amyloid precursor protein (APP) is processed. Risk variants of SORL1 are hypothesized to have decreased expression, which is thought to lead to an increase in β-amyloid production. Recently, Lawrence Goldstein’s laboratory found that, as neuronal stem cells were differentiating into neurons, cells with the risk variant of SORL1 did not respond to brain-derived neurotrophic factor (BDNF), which normally generates higher expression of SORL1. On the other hand, cells with a protective variant of SORL1 demonstrate strikingly increased expression in response to BDNF. If these findings are accurate, Lawrence Goldstein and colleagues suggest that increased SORL1 may lead to reduced β-amyloid. Consequently, drugs that increase SORL1 may be desirable for AD, and patients participating in clinical trials could benefit from stratification for risk vs. protective variant SORL1 genotypes.

Adrian Ivinson, director of the Harvard NeuroDiscovery Center at Harvard University, discussed a similar approach in which his laboratory tested its library of compounds against cells from 50–200 patient-derived, induced pluripotent stem neuronal cell lines. It is a departure from the standard convention of testing thousands of compounds against one assay designed to model the disease process. The risk of this widely practiced approach is that the assay may not be a good representation of the disease. In the case of AD, for example, one cell line could be from a patient with the inherited form of the disease, while another could be from a patient with the sporadic or late-onset form of disease, while yet another could be from a patient who has positron emission tomography (PET)-amyloid-positive imaging, but is asymptomatic. The goal is to build patient heterogeneity into the drug discovery process and hopefully find a pattern of hits that are active in some portion of patient-derived cell lines. Ivinson said this patient-stratified approach might be one method to accelerate drug discovery.

Amyotrophic Lateral Sclerosis

Stem cells could also speed go/no-go decisions by using them to interrogate potential drugs, said Kevin Eggan, associate professor at the Harvard Stem Cell Institute. Eggan observed that pluripotent stem cells allow hypotheses from animals to be tested in human neurons, in many forms of the disease, and in “human-specific” biology. ALS is a neuromuscular condition characterized by spreading loss of spinal motor neurons, which results in fatal paralysis. Early in the disease, motor neurons are selectively vulnerable to degeneration and death. Genetic progress has been made in recent years, with about 25 percent of ALS now explained by mutations in about a dozen different genes, including superoxide dismutase 1 (SOD1) and chromosome 9 open reading frame 72 (C9orf72), said Eggan. The genes encode proteins that are involved in many different aspects of neurobiology and cell biology. Researchers are trying to understand how mutations in these genes induce motor neuron degeneration and why motor neurons are selectively sensitive to the effects of mutation when other cells are not.

The answers to these questions have been slow in coming largely because of a lack of animal models. Most investigators rely on one particular mouse model, the SOD1 mouse, which offers a good phenocopy of ALS by virtue of displaying rapid motor neuron degeneration. But drugs in several clinical trials (e.g., minocycline) have failed despite showing success in the SOD1 mouse (Couzin, 2007). The lack of translation might be due to differences between mice and humans, or differences between SOD1 mutations and other forms of ALS.

Eggan said that a new and faster alternative to animal studies has emerged—the use of human stem cells. One approach is to reprogram skin fibroblasts into iPSCs, followed by directed differentiation into motor neurons. The latter is accomplished by treating iPSCs with an agonist of the sonic hedgehog signaling pathway and retinoic acid, and then by plating on laminin (Dimos et al., 2008). A second approach is by lineage conversion, starting with skin fibroblasts and, using transcription factors that are expressed in neurons, transdifferentiating the fibroblasts into motor neurons. Stem cells made by either method function normally, said Eggan.

Eggan’s current research examines motor neurons from differentiated iPSCs derived originally from ALS patients. Eggan and colleagues are trying to determine whether motor neuron degeneration is caused by de-

fects in cells interacting with the motor neuron (i.e., astrocytes and microglia), or whether the degeneration is caused by defects intrinsic to the motor neuron. The researchers found that patients had a defective prostanoid DP1 receptor. Eggan and colleagues then turned to animal models to determine the validity of this finding. After knocking out the DP1 gene, they found that ALS animals with one or two copies knocked out lived longer than those with the defective gene, said Eggan.

In closing, Eggan said there is a torrent of released genetic data that might support the use of iPSCs. Genetic information might help determine the location of variants that are causative for disease, although it will not provide much information on how the molecular biology of patients is changing. In his opinion, many of the genetic variants may be near neuronal genes in regulatory regions, which are not well conserved between humans and mice. Because of this, understanding how regulatory regions change the expression of nearby genes by using human iPSCs or human embryonic stem cell–derived neurons might lead to greater success compared to conventional animal models. In the end, Eggan advocated for a measured approach and suggested that rather than overinvesting in one hypothesis or target, a broad-based platform of investment that raises all areas could be beneficial to accelerate the development of therapeutics.

Humanized Animal Models

The use of humanized animal models may be another helpful technique in identifying molecular targets by examining the function of normal or genetically diseased human-tissue stem cells in mice. By genetically characterizing specific mouse strains with known genetic backgrounds, researchers might further understand the disease and identify associated traits (Schughart et al., 2012). Beginning in the 1980s, Irving Weissman, director of the Institute of Stem Cell Biology and Regenerative Medicine at Stanford University, and colleagues transplanted human fetal liver hematopoietic cells, thymus, and lymph node into a severe combined immunodeficiency (SCID) mouse (McCune et al., 1988). The fetal cells subsequently differentiated in the mouse host into functionally mature cells, leading to the use of humanized mouse models in disease research.

Weissman and colleagues isolated central nervous system stem cells from human neural fetal tissue using antibodies to early cell surface markers (Uchida et al., 2000). In vitro, the stem cells differentiate into

neurons and glia. When transplanted into the lateral ventricles of immunodeficient mice, the stem cells displayed engraftment, migration, and differentiation into neurons, astrocytes, and oligodendrocytes. The cells migrated throughout the brain to the cerebral cortex, corpus callosum, and cerebellum, as well as to sites of neurogenesis, where the cells continued to proliferate up to 7 months after the transplant. Weissman noted that the success of this experiment gave rise to clinical trials to replace tissues lost in two different conditions—hypomyelinating disease and spinal cord injury.

Weissman concluded with optimism that human nervous system stem cells hold promise for neurodegenerative diseases that exhibit widespread cell loss. He noted that stem cells themselves are a therapeutic entity and must be researched in vivo. Only stem cell transplants, in his view, have the potential to repopulate large swaths of tissue lost to disease-induced neuronal degeneration. Weissman hopes to work toward developing a true human neuronal mouse using iPSC-derived human neural stem cells to potentially validate his work and understand the disease pathogenesis to help develop therapeutics.

GENETICS

Using genetics as a tool to better understand variations in patient populations compared to healthy populations may facilitate the identification of novel therapeutic targets for nervous system disorders. Through DNA sequencing, researchers have the ability to locate molecular sites associated with specific diseases. More specifically, examining changes to gene expressions that contribute to or alter disease pathology may help pinpoint potential targets (Altar et al., 2009; Orth et al., 2004). Many speakers highlighted the usefulness of understanding genetics when selecting animal models and understanding overlap among diseases and molecular mechanisms of individual diseases.

Evolutionary Considerations

New analytic and genetic techniques exist that allow for the examination of species differences in the development and interpretation of disease models. Daniel Geschwind, Gordon and Virginia MacDonald Distinguished Professor in the school of medicine at the University of

California, Los Angeles, discussed how therapeutic development for nervous system disorders might be better served by considering evolutionary differences between species and humans when using animal models. Developing disease-altering medicines for nervous system disorders using animal (mouse) models has not been as successful as predicted, and although there are many reasons for this, evolutionary differences are seldom explored or highlighted (Bolker, 2012).

Mice and humans diverged at least 70 million years ago from when they shared a common ancestor. Genes in the brain show different patterns of expression in mice and humans, said Geschwind. Not only are there more neurons in humans, but there is also greater complexity of evolutionary regions, especially in the temporal and frontal lobes. The temporal lobe is involved in auditory and visual processing, language comprehension, memory, and emotion. The frontal lobe is responsible for high-level cognition, including multitasking, social cognition, and planning and manipulating of abstract representations (Konopka et al., 2012). In a study of FOXP2, a transcription factor implicated in human speech and language dysfunction, a change of two amino acid sequences leads to significant functional deficits, largely due to FOXP2’s regulatory role (Konopka et al., 2009). In general, when considering genes, it is important to understand connectivity, interactions, regulation, and network dynamics, said Geschwind.

In a recent study, Geschwind and colleagues used high-throughput DNA screening methods, “next-generation DNA sequencing,” and microarrays to compare the transcriptome of the telencephalon in the human, chimpanzee, and macaque (Konopka et al., 2012). They found a dramatic increase in transcriptional complexity specific to the human frontal lobe. Only one-quarter of frontal pole modules¹ were preserved in humans, whereas caudate nucleus modules were highly conserved across species. Non-conserved modules were likely to have hub genes (central genes within a module) under positive selection.² This applied to the CLOCK gene, a circadian rhythm gene that is implicated in bipolar disorder, and, in a separate study, this also applied to the presenilin gene implicated in AD (Miller, 2010). Geschwind and colleagues concluded in their study that genes within modules associated with higher connectivity—hub genes—are more likely to be disease genes.

______________

¹Sets of genes co-regulated to respond to different conditions (Segal et al., 2003, p. 166).

²Process in which beneficial/adaptive genetic variants increase throughout a population.

Case Study for Drug Discovery

David Goldstein, director of the Center for Human Genome Variation at Duke University, provided an overview of the first discovery project of the National Institute of Neurological Disorders and Stroke–funded consortium known as Epi4K to study the genetics of epileptic encephalopathies. The initiative is a “center without walls” for studying 4,000 cases of highly selected and well-characterized epilepsy (Epi4k Consortium, 2012). David Goldstein and colleagues’ project was to search for de novo mutations³ in two types of epileptic encephalopathies, infantile spasms (IS) and Lennox-Gastaut syndrome (LGS). IS is found in 1 in 3,000 live births, with onset between 4 and 12 months of life and characterized by chaotic interictal and electroencephalogram patterns of hypsarrhythmia. LGS has onset between 1 and 8 years and is characterized by mixed seizure types and intellectual disabilities. David Goldstein and colleagues hypothesized that a significant number of IS and LGS cases of unknown etiology are likely due to dominant de novo mutations. The approach was to conduct whole-exon sequencing on 300 family trios consisting of a child with IS or LGS and their unaffected parents, using samples collected by the Epilepsy Phenome/Genome Project.⁴

David Goldstein and colleagues found there was 1 de novo mutation in 92 children, 2 in 45 children, and 3 in 25 children. But not all of these mutations were in genes that caused or contributed to the disease. Consequently, David Goldstein and colleagues performed new statistical techniques to find causative mutations. The group found that genes intolerant to functional variation were more likely to carry mutations that cause or contribute to disease. Altogether, the analysis produced 25 epileptic encephalopathy genes, of which 11 were also implicated in other nervous system disorders.

David Goldstein noted that recent studies have shown the existence of genetic overlap across neuropsychiatric disorders (Serretti and Fabbri, 2013). The implication is that drugs found beneficial in one disorder might be beneficial in another. David Goldstein and colleagues also found that the de novo mutations playing an causative role could be grouped into functional categories. For example, six de novo mutations were found in GABA receptor subunits. Another group of mutations impinge on the epidermal growth factor (EGF)/extracellular signal-

______________

³De novo mutations are genetic mutations that neither parent possessed or transmitted.

⁴See http://www.epgp.org.

regulated kinases (ERK) signaling pathway. Even though mutations were likely to differ among patients, it seemed likely that patients could be stratified by functional groups. The functional characterization of the de novo mutations is an essential part of the interpretation of the data. It then may be possible to match patients having certain mutations with a specific type of antiepileptic drug that targets the affected functional pathway.

Case Study of Identifying New Targets

Daniel Weinberger, director and chief executive officer of the Lieber Institute for Brain Development at Johns Hopkins University, described a new approach, or roadmap, to target identification in schizophrenia, which involves translating genes of interest into molecular mechanisms of illness. The key feature of this roadmap is a reliance on RNA sequencing in the brain after a gene has been identified. Studying the transcriptome has the potential to shed light on non-coding variations that affect gene function, specifically associated with illness state and genetic risk. Molecular mechanisms of association can be studied in cellular and animal models once these variations are understood. Weinberger provided two examples illustrating this roapmap: a metabotropic glutamate receptor and a novel potassium channel.

The metabotropic glutamate receptor 3 (GRM3) “regulates synaptic glutamate via a presynativ mechanism and by regulating the expression of the glial glutamate transporter, which inactivates synaptic glutamate” (Tan et al., 2007). Gene variation in GRM3 has been found to be associated with schizophrenia in patient samples studied by the Psychiatric Genomics Consortium⁵ and by Weinberger’s laboratory (Egan et al., 2004). Several single-nucleotide polymorphisms (SNPs) found thus far are located in non-coding regions, prompting study of RNA transcripts. Weinberger and colleagues discovered an alternatively spliced form of GRM3, termed GRM34, which yields a truncated receptor protein (Sartorius et al., 2006). Weinberger and colleagues sought to quantify mRNA expression levels of GRM3 and GRM34 in regions of the brain most affected by schizophrenia, the dorsolateral prefrontal cortex and hippocampus. They found increased expression of GRM34 in the prefrontal cortex of schizophrenia postmortem brains

______________

⁵See https://pgc.unc.edu.

relative to controls (Sartorius et al., 2008). The fact that increased mRNA expression is associated with both disease state and genetic risk has propelled Weinberger and his colleagues to further study molecular mechanisms of association. This is a work in progress and all the RNA sequence data from thousands of brains will be available to other researchers through the Lieber Institute’s website.⁶

In a separate study, Weinberger and colleagues found a new schizophrenia susceptibility gene, an isoform of the gene KCNH2, that encodes a novel hERG potassium channel protein (KCNH2-3.1) (Huffaker et al., 2009). The 3.1 isoform, which turns out to be primate- and brainspecific, is encoded in close proximity to risk-associated SNPs. In a cell model system, the isoform lacks a domain necessary for slow channel deactivation and repolarization, leaving neurons to fire in rapid trains of activity (Huffaker et al., 2009). The isoform exhibits increased expression in schizophrenia and is associated with risk genotype. Weinberger and colleagues showed that schizophrenic patients with the genotype for the novel isoform KCNH2 were five times more likely to complete a clinical trial with the antipsychotic drug clozapine (Apud et al., 2012). Patients without the genotype were more likely to discontinue clozapine. The isoform, in short, modulates treatment response.

Finally, Weinberger’s laboratory has created a mouse model that overexpresses the novel hERG potassium channel, as previously discussed. The mouse demonstrates executive cognitive deficits, episodic memory defects, and electrophysiological channel characteristics seen in cellular models of schizophrenia. It also has a deficit in hippocampal long-term potentiation, but the deficit does not emerge until animals reach young adulthood—an important feature of a developmental model of schizophrenia. When the novel gene is repressed during adult life, the cognitive deficits in the animals are reversed. With an eye toward drug development, Weinberger and colleagues have identified several molecules that show selectivity for this brain-specific novel potassium channel.

A better understanding of the genetic underpinnings of nervous system disorders has the potential to facilitate drug development through improved target identification.

______________

⁶See http://www.libd.org.

IMAGING TECHNIQUES

Imaging technologies are another tool that may improve and accelerate therapeutic development. While functional imaging—used to characterize physiological changes in the body—can serve as an objective biomarker, molecular imaging—which attempts to illustrate biological processes at the cellular level—can be used to identify new targets (Rudin and Weissleder, 2003; Willmann et al., 2008). Through imaging of transgenic animals, researchers might be better equipped to determine molecular targets that correlate with nervous system disorders.

Translational Tools for Drug Development for Mood Disorders

Wayne Drevets, scientific vice president and disease area leader in mood disorders at Janssen Pharmaceutical Companies of Johnson & Johnson, discussed several translational tools and models relevant to therapeutic development in mood disorders. Drevets is currently studying mood disorders with bioimaging modalities due to the limited number of animal models that fully recapitulate the behavioral and biological abnormalities of mood disorders. His focus is on the medial prefrontal network, which includes the medial prefrontal cortex (mPFC) and anatomically related limbic, striatal, thalamic, and basal forebrain structures. The limbic structure of greatest interest is the subgenual anterior cingulate cortex (subACC). Targeting the subACC with deepbrain stimulation causes striking remission of symptoms in treatmentrefractory depression (Mayberg et al., 2005).

By imaging the mPFC, Drevets and colleagues (1997) found lower metabolic activity and decreased cortical volume in patients with bipolar and unipolar depression. Similarly, the subACC exhibited reduced metabolism and cortical volume in both types of depression (Drevets et al., 2008). Patients taking lithium or valproate, both of which have neurotrophic effects in experimental animals, had larger subACC volume. In a prospective study, lithium was found to increase gray-matter volume in the prefrontral cortex (PFC) and subACC (Moore et al., 2009).

The PFC and subACC regions are significant in the prognosis of mood disorders, Drevets said. Nineteen of 23 studies showed that higher pre-treatment ACC activity, measured with a variety of imaging techniques, predicted better response to antidepressant treatment

(Pizzagalli, 2011). People with more severe gray-matter reductions in the PFC and ACC tend to display a greater severity of major depressive disorder (Salvadore et al., 2011).

The subACC also serves as a potential region to study biomarkers of treatment response (Pizzagalli, 2011). A pattern shift in metabolic activity is associated with reversal of negative emotional processing bias in depressed patients. The pattern shift is evident upon administration of several antidepressant drugs, including selective serotonin reuptake inhibitors (SSRIs) and drugs with novel antidepressant action, said Drevets.

Ketamine, when given to animals, acts as a rapid antidepressant and reverses loss of dendritic spines in mPFC and loss of synaptic function (Duman and Aghajanian, 2012; Duman et al., 2012). Previous human and animal studies have found that depression is associated with dendritic atrophy, synapse loss, and reduction in glia and GABA-ergic interneurons in limbic-mPFC circuits, noted Drevets. Ketamine is thought to produce a beneficial effect by disinhibition of fast-spiking inhibitory interneurons. This increases glutamate release and leads to induction of synaptogenesis and dendritic spine outgrowth (Duman et al., 2012). According to Drevets, the success with ketamine and its mechanism of action has led to the identification of new targets for rapid antidepressant action with a new class of drugs that is less toxic than ketamine. Drevets and his colleagues are working on increasing studies conducted on people with depression rather than using old behavioral assays in animal models.

Identifying Molecular Targets in Traumatic Brain Injury

Ramon Diaz-Arrastia, professor of neurology at the Uniformed Services University of the Health Sciences, spoke about the use of imaging to identify new molecular targets in traumatic brain injury (TBI). TBI is responsible for 1.7 million emergency department visits each year. Eighty percent of cases are mild, 10 percent are moderate, and 10 percent are severe. Two percent of the U.S. population, or 5.3 million people, live with disabilities resulting from TBI, making it the most common cause of death or disability in people under age 45 (Thurman et al., 1999).

TBI is an umbrella category for many subtypes of traumatic injury, including epidural hematoma, subdural hematoma, parenchymal contusi-

on, diffuse axonal injury, and traumatic subarachnoid hemorrhage. Most patients tend to have mixed pathologies, said Diaz-Arrastia. In general, computed tomography (CT) scans underestimate the extent of injury, whereas magnetic resonance imaging (MRI) and diffusion tensor imaging (DTI) provide better resolution. A recent prospective study of whole-brain volume as a biomarker found that patients experienced substantial global atrophy, displaying a mean volume loss of 4.5 percent by an average of 8 months post-TBI (Warner et al., 2010). Atrophy continues to occur months after the injury, but the greatest atrophy occurs in the acute/early subacute time period.

Atrophy is not uniform after diffuse traumatic axonal injury. Subcortical structures that are vulnerable to TBI-induced atrophy include the hippocampus, amygdala, and thalamus. In the cortex, the most vulnerable regions are the precuneus, posterior cingulate, superior parietal cortex, and superior frontal cortex (Warner et al., 2010). These cortical regions overlap somewhat with regions of cortical atrophy in AD (Perrin et al., 2009). Atrophy detected with MRI can be used as a surrogate biomarker of neuroprotective efficacy. There is a strong relationship between atrophy and function: The greater the atrophy, the greater the extent of disability (Warner et al., 2010).

Several new classes of drugs are being tested against emerging targets in TBI, noted Diaz-Arrastia. One promising target is vascular injury. Consequently, researchers are interested in drugs that promote angiogenesis or growth of new blood vessels. Several U.S. Food and Drug Administration–approved drugs related to blood flow—erythropoietin, sildenafil, and statins—are being “repurposed” for TBI, said Diaz-Arrastia. Enriched endothelial progenitor cells are also under study to help revasculize the site of injury (Hristov et al., 2003). Another target is Nogo, a myelin-based protein that inhibits neurite outgrowth. Anti-Nogo monoclonal antibodies have witnessed some measure of success in animal models of TBI and spinal cord injury (Freund et al., 2007; Marklund et al., 2007). Bone marrow–derived mesenchymal stem cells are also under study, noted Diaz-Arrastia.

Diaz-Arrastia concluded with the observation that stem cells are more likely to work in TBI and spinal cord injury as opposed to AD and Parkinson’s diseases. The patients are generally younger, their brains are better equipped to regenerate, and an underlying neurodegerative disease process is not continuing to contribute to further pathology. In summary, imaging modalities show promise to identify new molecular targets in

the brain following TBI and have the potential to serve as biomarkers to help in the drug discovery process, said Diaz-Arrastia.

Top-Down Approaches

Beginning with imaging studies in humans and then validating findings in animal models, a top-down approach could improve and accelerate therapeutic development, said Scott Small, professor of neurology at Columbia University Medical Center. Small provided evidence for this approach using an example of neuroimaging to map human hippocampal pathology in vivo. The hippocampus is abnormal in several nervous system disorders, including AD. Small and colleagues sought to find out if different patterns of vulnerability were associated with different disorders that affect the hippocampus. If this could be demonstrated, Small noted it might be helpful for identifying biomarkers and underlying mechanisms of nervous system disorders. Small and colleagues conducted neuroimaging studies in humans to further understand hippocampal patterns for AD and schizophrenia, followed by validation studies in animals.

In a series of experiments, Small and colleagues sought to distinguish cognitive dysfunction in AD versus normal aging using functional MRI (fMRI). They found hypometabolism in the entorhinal cortex, but not in the dentate gyrus, whereas the opposite was true in normal aging. The results were first established in humans and later confirmed in monkeys and transgenic mice carrying an Alzheimer-related transgene (Moreno et al., 2007; Small et al., 2002, 2004). Hippocampal imaging thus has the potential to serve as a biomarker for drug discovery distinguishing AD from normal aging, noted Small.

Having established the dentate gyrus as a site affected by normal aging, Small and coauthors examined whether any interventions could restore function. The researchers took fMRI measurements of the dentate gyrus before and after several weeks of exercise in healthy mice and humans. Results were similar across both species: Exercise was correlated with increased cerebral blood volume (CBV) in the dentate gyrus, but not in other regions of the hippocampus (Pereira et al., 2007). In humans, increases in CBV are correlated with improved cognitive testing. Neurogenesis may underlie the improvement, given that the dentate gyrus is one of few regions of the brain known to display neurogenesis, according to the authors.

Small and colleagues sought to determine if abnormalities in hippocampal subregions were exhibited in patients with schizophrenia. Patients with prodromal psychosis, who are at risk for schizophrenia, exhibited hypermetabolism in the CA1 subregion of the hippocampus (Schobel et al., 2009). In a subsequent study, Small and colleagues used fMRI and structural MRI to map the hippocampus of prodromal schizophrenia (Schobel et al., 2013). After 3–4 years, the hypermetabolism seen in sub-region CA1 spread to the subiculum, another subregion of the hippocampus, resulting in atrophy. The atrophy, which was presumably caused by hypermetabolism, became apparent during the emergence of psychosis. In an effort to identify the underlying mechanisms, Small and coworkers turned to a mouse model of psychosis. Ketamine administration in this model was found to mimic the hippocampal pattern of hypermetabolism seen in schizophrenia. The investigators hypothesized that hypermetabolism was generated by increased activity of the neurotransmitter glutamate. A glutamate antagonist and direct measurements of extracellular glutamate revealed that glutamate was behind the abnormalities, hypermetabolism, and subsequent loss of hippocampal volume (Schobel et al., 2013). The investigators proposed that lowering glutamate might be an effective pharmacological strategy during early stages of schizophrenia to reduce hippocampal hypermetabolism, protect hippocampal volume, and halt disease progression.

Imaging is a tool that can be used in several facets of drug development for nervous system disorders. Particularly for target identification, researchers are able to visually map areas of the brain that are associated with specific disorders, which in turn may help to objectively determine potential targets.