espectro autista, descubrimiento de medicamentos

The term “autism” was introduced in 1938 by Hans Asperger1 when he was investigating a type of autism spectrum disorder (ASD) now known as Asperger’s syndrome. Independently, in 1943 Leo Kanner used the term autism in a report in which he described 11 children with con-siderable behavioural similarities, including autistic alone-ness and insistence on sameness2. Today, the diagnostic category of ASD represents a neurodevelopmental con-dition that is characterized by the presence of persistent deficits in social interaction and communication, as well as restricted social interests and repetitive behaviours. These core symptoms are usually detectable and diag-nosed in children by 3 years of age. The symptoms are highly variable in severity and often present with other neuropsychiatric symptoms, including cognitive deficits, anxiety, hyperactivity, aggression and epilepsy, which together form a complex pattern of co-morbidities (BOX 1).

Unlike previously thought, ASD is not a rare disor-der. Indeed, the latest prevalence estimate for ASD was reported to be about 1% in the general population3 (BOX 1), which is similar to the prevalence of schizophrenia.

However, ASD has not been the object of comparable efforts to develop novel therapies, partly because the clinical definition of ASD and its epidemiological char-acterization was only recently defined. Additionally, there is a poor understanding of its molecular pathophysiology and the lack of preclinical and clinical drug development tools. As a result, the current pharmacotherapy of ASD still relies greatly on the off-label use of a wide range of different agents such as antidepressants, anxiolytics and antipsychotics. There are only two approved pharmaco-logical therapies for ASD — the atypical antipsychotics risperidone and aripiprazole — indicated for the treat-ment of irritability associated with ASD, but are ineffec-tive in addressing core symptoms in these patients4.

The aetiology of ASD is complex, and includes both genetic and environmental factors. In recent years, genetic studies have identified a large number of muta-tions that may collectively explain up to 15–20% of all ASD cases. Furthermore, the study of syndromic forms of ASD caused by single-gene mutations, such as fragile X syndrome (FXS) and tuberous sclerosis (TSC), has

1F. Hoffmann‑La Roche, pRED, Pharma Research and Early Development, DTA Neuroscience, Grenzacherstrasse 124, 4070 Basel, Switzerland.2Present address: Neurimmune Holding, Wagistrasse 13, 8952 Schlieren, Switzerland.Correspondence to A.G.: [email protected]:10.1038/nrd4102

Drug discovery for autism spectrum disorder: challenges and opportunitiesAnirvan Ghosh1, Aubin Michalon1,2, Lothar Lindemann1, Paulo Fontoura1 and Luca Santarelli1

Abstract | The rising rates of autism spectrum disorder (ASD) and the lack of effective medications to treat its core symptoms have led to an increased sense of urgency to identify therapies for this group of neurodevelopmental conditions. Developing drugs for ASD, however, has been challenging because of a limited understanding of its pathophysiology, difficulties in modelling the disease in vitro and in vivo, the heterogeneity of symptoms, and the dearth of prior experience in clinical development. In the past few years these challenges have been mitigated by considerable advances in our understanding of forms of ASD caused by single-gene alterations, such as fragile X syndrome and tuberous sclerosis. In these cases we have gained insights into the pathophysiological mechanisms underlying these conditions. In addition, they have aided in the development of animal models and compounds with the potential for disease modification in clinical development. Moreover, genetic studies are illuminating the molecular pathophysiology of ASD, and new tools such as induced pluripotent stem cells offer novel possibilities for drug screening and disease diagnostics. Finally, large-scale collaborations between academia and industry are starting to address some of the key barriers to developing drugs for ASD. Here, we propose a conceptual framework for drug discovery in ASD encompassing target identification, drug profiling and considerations for clinical trials in this novel area.

R E V I E W S

NATURE REVIEWS | DRUG DISCOVERY VOLUME 12 | OCTOBER 2013 | 777

© 2013 Macmillan Publishers Limited. All rights reserved

Face validityThe extent to which an animal model’s phenotype resembles human symptoms (e.g., an Alzheimer’s disease model that progressively develops amyloid plaques and memory impairment).

Construct validityThe extent to which an animal model has the same aetiology and underlying mechanism as the human disorder (e.g., an Alzheimer’s disease model generated by introducing a human presenilin mutation in the mouse)

provided important insights into their molecular under-pinnings. This knowledge is likely to be important in furthering our understanding of the biological basis of other types of ASD of multifactorial aetiology.

New techniques such as patient-derived, induced pluri-potent stem cells (iPSCs) provide an unprecedented oppor-tunity to test new therapeutic approaches and new drug candidates directly on human neural tissue that contains the molecular alterations associated with ASD. For several syndromic forms of ASD, the ability to create genetically modified animal models has provided a means to test in vivo novel pharmacological interventions on a broad range of therapeutically relevant phenotypes. These novel pharmacological interventions include the correction of abnormal synaptic morphology and physiology, as well as the reversal of behavioural alterations, in relevant animal models even when treatment is applied well after symptom onset5–9. These findings have triggered a paradigm shift in the way we think about pharmacological interventions for ASD. Indeed, they indicate that ASD may not be the result of an irreversible disruption of brain development and that pharmacotherapy may have the potential to offer benefit that extends beyond symptomatic improvements.

Here, we provide a brief update on the current under-standing of the pathophysiology of ASD, which includes the most recent insights derived from the rapidly evolv-ing genetics of ASD. Additionally, we propose a concep-tual framework for drug discovery and development in ASD. This includes not only concepts regarding target identification, preclinical drug profiling and clinical end

point development, but also our views on key diagnostic considerations to target novel drug candidates to appro-priate ASD subtypes that are characterized by common pathophysiological underpinnings.

Understanding the pathophysiology of ASDRational drug design is based on a mechanistic under-standing of the pathophysiology of the disease or dis-order and on the validation of the therapeutic targets in translational cellular and in vivo models with face validity and construct validity. Given the significant genetic contribution to the aetiology of ASD, there is a strong expectation that genetic studies will help shed light on its pathophysiological mechanisms.

Learning from genetic studies in ASD. Twin and family studies have shown that ASD is a highly heritable con-dition. Indeed, the concordance rate between monozy-gotic twins, which have been observed in various places at rates of between 60% and 91%, defines ASD as the most heritable psychiatric condition10. In addition, sev-eral monogenic syndromes present considerable symp-tomatic overlap with ASD, thereby providing examples in which autistic behaviours are primarily determined by genetic mutations. The overall frequency of such syndromic forms of ASD is low, the most frequent being FXS, which explains about 3% of ASD cases. Other syn-dromic forms of ASD include TSC (representing 2% of ASD cases), Rett syndrome (1%) and neurofibromatosis type 1 (1%). Similarly, cytogenetic abnormalities such as

Box 1 | Definition of autism spectrum disorder

Autism-related disorders are currently diagnosed according to the criteria of the fourth edition of the “Diagnostic and Statistical Manual of Mental Disorder” (DSM). This current classification system distinguishes autistic disorder, Rett syndrome, childhood disintegrative disorder, Asperger’s syndrome and pervasive developmental disorder not otherwise specified. This classification has been found to be inconsistent and artificial, and the fifth edition of the DSM proposes to reunify these different diagnostic labels under the unique name of autism spectrum disorder (ASD) (note the singular form). The two core symptoms that define a diagnostic for ASD are persistent deficits in social communication and social interaction, and restricted interests and repetitive behaviours. The presence of social and communication deficits should not be accounted for by a general delay in development, and the core symptoms should already be present in early childhood. The term “spectrum” has been incorporated in the name to indicate the heterogeneity of the clinical forms of the disorder, both in terms of severity of the core symptoms and diversity of co-morbid features. Intellectual disability, for example, is a possible co-morbid symptom that affects about one-third of the whole ASD population65. Similarly, patients with ASD may present pathological levels of anxiety, hyperactivity, impulsivity, agitation or aggressive behaviours, including self-mutilation66,67. At the functional level, patients with ASD may also exhibit sensory hypersensitivity that affects one or multiple sensory modalities68, as well as seizures69 or gastrointestinal disorders70.

The clinical development course of the disorder is, for most cases, progressive and begins early during infancy. Abnormal development is usually detected by the parents during the second year of life, and a formal clinical diagnosis is usually made by the age of 3 years. Prodromal signs of atypical development are in fact already present during the first year of life, but this goes mostly unnoticed and is revealed by retrospective analysis of family videotapes. Less frequently, children with ASD develop normally until they present a regression of their socio-cognitive capacities71.

The question of the prevalence of ASD is a highly debated topic, because epidemiological studies report a rapid prevalence increase in western countries. The latest report (March 2012) from the US Center for Disease Control and Prevention indicates a prevalence of 11.3 per 1,000 in 8‑year old children and a fivefold difference between boys (18.5 per 1,000) and girls (3.9 per 1,000)3. The inflation in prevalence is attributed largely to the increased awareness of ASD in the general population and by physicians, diagnostic substitution, broadening of the diagnostic criteria, and changes in the policies for special education. Nonetheless, these reasons may not explain the whole phenomenon and the possibility of a genuine increase in the incidence of ASD cannot be ruled out72.

R E V I E W S

778 | OCTOBER 2013 | VOLUME 12 www.nature.com/reviews/drugdisc


RAS–MAPK–ERK pathwayRAS–mitogen-activated protein kinase (MAPK)–extracellular signal-regulated kinase (ERK) pathway.

PI3K–AKT–mTOR pathwayPhosphoinositide 3-kinase (PI3K)–AKT–mammalian target of rapamycin (mTOR) pathway.

the maternal duplication at the 15q11-13 locus and the deletions or duplications at the 16p11 locus are observed at a frequency between 1% to 3% in patients with ASD11. Together, the identified genetic causes for ASD explain more than 15% of all cases12.

Tremendous technical progress during the past decade has provided fundamental insights into the genetics of ASD. Modern technologies such as high-density microarrays, which allow the detection of copy number variants (CNVs) across whole genomes with

unprecedented resolution, revealed the important con-tribution of very rare and highly penetrant CNVs to the aetiology of ASD13. Moreover, exome sequencing studies, which are capable of detecting single nucleotide variants (SNVs), revealed the elevated rate of de novo mutations in ASD cases and its positive correlation with paternal age14. These results support the idea that sporadic ASD is best explained by a large number of rare and highly penetrant mutations, and that common gene variants may not con-tribute a significant risk for developing ASD, or only to a very low level that is still below the detection limits of current technologies15 (FIG. 1). Increases in sequencing speed, and decreases in sequencing costs, will soon allow complete genome sequencing in patients. This should enable the identification of rare mutations in patients and pave the way to a molecular nosology of ASD and, ultimately, personalized medicine.

The fact that potentially hundreds of mutations lead to the cognitive and behavioural outcomes comprised in ASD (TABLE 1), warrants a pathophysiological approach that focuses on signalling networks instead of individual genes and proteins, whereby different mutations within the same network would result in similar phenotypic manifestations16. In keeping with this framework, genes that have been linked to ASD can be clustered into three broad categories: those involved in synapse structure and activity17–19, those involved in protein synthesis20 and those involved in gene expression regulation21 (FIG. 2).

Investigating syndromic forms of ASD. Mutations caus-ing syndromic forms of ASD were identified over two decades ago. These discoveries enabled the generation of genetically engineered mice carrying these mutations, with behavioural, cognitive and physiological alterations similar to the ones observed in patients. Such mouse models allowed the detailed investigation of the patho-physiological mechanisms of ASD and led to the emer-gence of multiple pathophysiological and therapeutic hypotheses. We consider here the examples of FXS and TSC, as these are the most advanced examples related to ASD in which a mechanistic understanding of the disease pathophysiology is leading the way to targeted therapies.

FXS results from the absence of the fragile X mental retardation protein (FMRP), a repressor of mRNA trans-lation that is particularly important for the regulation of activity-dependent protein synthesis in neurons. In both mice and patients the absence of FMRP leads to signifi-cant alterations in cognitive functions, dendritic spine morphology and intracellular signalling22. Importantly, reduced FMRP levels unleash excessive signalling of metabotropic glutamate receptors 5 (mGluR5), which, in turn, leads to a saturation of the RAS–MAPK–ERK path-way, the PI3K–AKT–mTOR pathway and the CREB pathway, endocytosis of AMPA receptors and a reduction in syn-aptic strength (efficiency of synaptic transmission)23. This impairs long-term potentiation (LTP) in both the cortex and amygdala, increases the LTP threshold and facilitates long-term depression (LTD) in the CA1 area of the hippocampus, and leads to learning and mem-ory impairments in Fmr1-knockout mice. Chronic

Figure 1 | Towards identification of the genetic basis of ASD. Progress in understanding the aetiology of autism spectrum disorder (ASD) is currently driven by genetic studies and serves as the basis for the identification of its pathophysiological mechanisms. Common gene variants, which are defined by a minor allele frequency (MAF) of >5% in the general population, are typically associated with small risk (odds ratio <1.5) to develop a complex disease such as ASD. Common variants are detected by means of association studies such as genome-wide association (GWA) studies, which compare allele frequency in populations of unrelated, affected or unaffected individuals. Very large-scale GWA studies (thousands of patient and control samples) are currently ongoing that are powered to detect less common gene variants (MAF ~1%) that are associated with a moderate risk for developing ASD76. In contrast to GWA studies, linkage analyses are able to detect rare and highly penetrant mutations that carry a large risk for a disease. Linkage analyses evaluate the co-segregation of alleles within families with affected and unaffected members. Progress in DNA sequencing technologies now allow complete exome and genome resequencing studies in patients with ASD. Such studies are identifying at an unprecedented pace ultra-rare and de novo mutations that are potentially associated with ASD77,78. Because the bare genetic sequence may not be sufficient per se to explain highly complex traits such as ASD — a phenomenon known as the “missing heritability”79 — elaborate models are being developed to integrate factors that contribute in a synergistic manner to the aetiology of the disorder. This includes the simultaneous occurrence of multiple genetic hits80, combining, for example, inherited and de novo mutations, the analysis of parental origin effects, as well as epigenetic81 and environmental factors. DISC1, disrupted in schizophrenia 1; FXS, fragile X syndrome; NLG, neuroligin; OXT, oxytocin; SHANK, SH3 and multiple ankyrin repeat domains; TSC, tuberous sclerosis; V1A, vasopressin 1A. Figure is modified, with permission, from REF. 82 © (2008) Macmillan Publishers Ltd. All rights reserved.

Nature Reviews | Drug Discovery

Effec

t siz

e

Smal

lM

oder

ate

Larg

e

0.01Very rare Rare Less common Common

0.1 1.0 10

Rare alleles, largeeffect size (e.g., FXS,TSC, 15q, NLG andSHANK)

Does not exist

Identificationin progress

Common variants,small effect size(e.g., V1A, DISC1and OXT)

Currently notidentifiable

Allele frequency (%)

Linkage studies Association studies

R E V I E W S



CREB pathwayCyclic AMP-response element-binding protein (CREB) pathway.

Long-term potentiation(LTP). A persistent strengthening of synaptic transmission in response to strong, correlated input.

Long-term depression(LTD). The converse of LTP; in LTD there is a long-lasting and activity-dependent decrease in synaptic efficacy.

reduction of mGluR5 activity corrects a broad range of phenotypes, including learning and memory deficits, sensitivity to audiogenic seizures, sensory hypersensi-tivity, hyperactivity, dendritic spine dysmorphogenesis and physiological alterations6,24.

Various approaches have been explored to regulate levels of matrix metalloproteinase 9 (MMP9) and gly-cogen synthase kinase 3β (GSK3β) activity for thera-peutic benefit in FXS25. Minocycline, an antibiotic of the tetracycline class, can inhibit MMP9 and its chronic administration promoted dendritic spine maturation and ameliorated cognitive deficits in Fmr1-knockout mice26. Similarly, lithium is commonly used in psychia-try for its mood-stabilizing properties that are possi-bly related to GSK3β inhibition. Chronic treatment of

Fmr1-knockout mice with lithium ameliorated several behavioural alterations, including social behaviours27,28. Finally, multiple lines of evidence indicate that there are alterations in the GABA (γ-aminobutyric acid)-mediated synaptic transmission in FXS and suggest that there is potential therapeutic benefit for a GABA-augmenting approach29. In Fmr1-knockout mice, arba-clofen, the r-enantiomer of the GABAB receptor agonist baclofen, corrects the elevated protein synthesis rate, AMPA receptor endocytosis and dendritic spine den-sity phenotypes30.

These pharmacological corrections of Fmr1-knockout mouse phenotypes in adult mice paved the way to test-ing these therapeutic hypotheses in patients. We address results from the corresponding clinical studies in the last

Table 1 | Examples of monogenic disorders associated with altered neurodevelopmental trajectories and autistic symptoms*

Protein name Protein symbol Condition

Synaptic proteins

Glutamate receptor, ionotropic, AMPA 3 GRIA3 (AMPA receptor subunit) X-linked mental retardation

Cadherin 9, cadherin 13 CDH9, CDH13 Landau–Kleffner syndrome

Contactin 4 CNTN4 Autism spectrum disorder

Homer homologue 1 HOMER1 Autism spectrum disorder

Neurexin 1 NRXN1 Pitt–Hopkins-like mental retardation

Neuroligin 3, neuroligin 4 NLGN3, NLGN4 X-linked mental retardation

Postsynaptic density protein 95 PSD95 (also known as DLG4) Autism spectrum disorder

Synapse-associated protein 97 SAP97 (also known as DLG1) Autism spectrum disorder

SAP90/PSD95-associated protein 2 SAPAP2 (also known as DLGAP2) Autism spectrum disorder

SH3 and multiple ankyrin repeat domains 2 SHANK2 Autism spectrum disorder

SH3 and multiple ankyrin repeat domains 3 SHANK3 Phelan–McDermid syndrome (22q13 deletion syndrome)

Cytoplasmic proteins

Fragile X mental retardation protein FMRP Fragile X syndrome

Neurofibromin 1 NF1 Neurofibromatosis (NF1 may also act at the synapse)

Phosphatase and tensin homologue PTEN Cowden syndrome

Tuberous sclerosis 1, tuberous sclerosis 2 TSC1, TSC2 Tuberous sclerosis

Harvey rat sarcoma viral oncogene homologue

HRAS Costello syndrome

Kirsten rat sarcoma viral oncogene homologue

KRAS Cardio-facio-cutaneous syndrome

Nuclear proteins

CREB-binding protein CBP (also known as CREBBP) Rubinstein–Taybi syndrome

Cyclic AMP-response element-binding protein

CREB Rubinstein–Taybi syndrome

Dual-specificity tyrosine-(Y)-phosphorylation regulated kinase 1A

DYRK1A Autism spectrum disorder and Down’s syndrome

Euchromatic histone-lysine N-methyltransferase 1

EHMT1 Kleefstra syndrome (9q subtelomeric deletion syndrome)

E1A binding protein p300 EP300 Rubinstein–Taybi syndrome

Histone deacetylase 4 HDAC4 Brachydactyly mental retardation syndrome (2q37 deletion syndrome)

Methyl CpG binding protein 2 MECP2 Rett syndrome

*Data from REF. 39.

R E V I E W S



Figure 2 | Convergence of ASD risk factors on specific intracellular mechanisms. Proteins with genetic variants associated with autism spectrum disorder (ASD) (excluding those in white ovals) are clustered in specific intracellular processes. For example, mutations conferring a significant risk to developing ASD affect proteins that are important for synaptic structure and activity, such as transmembrane proteins (neuroligin, cadherin and contactin) and postsynaptic scaffolding proteins (postsynaptic density protein 95 (PSD95), SH3 and multiple ankyrin repeat domains (SHANK), Homer, synapse-associated protein 97 (SAP97) and SAP90/PSD95-associated protein 2 (SAPAP2))83. These proteins form a functional network with ionotropic and metabotropic glutamate receptors (NMDAR, AMPAR and metabotropic glutamate (mGluR)), and mutations in the corresponding genes have been shown to affect glutamatergic transmission17. Similarly, mutations associated with syndromic forms of ASD affect proteins such as neurofibromatosis 1 (NF1), phosphatase and tensin homologue (PTEN), tuberous sclerosis 1 (TSC1) and TSC2, and fragile X mental retardation related protein (FMRP), which are parts of the extracellular signal-regulated kinase (ERK) and mammalian target of rapamycin (mTOR) pathways and are essential for the regulation of activity-dependent synthesis of synaptic proteins18. Mutations causing intellectual deficit and autistic behaviours also affect nuclear proteins involved in long-term regulation of gene expression, such as transcription factor modulators (cyclic AMP-response element-binding protein (CREB), CREB-binding protein (CBP), E1A binding protein p300 (EP300) and dual-specificity tyrosine-(Y)-phosphorylation regulated kinase 1A (DYRK1A)), DNA methylation (methyl CpG binding protein 2 (MECP2)) and histone methylation (euchromatic histone-lysine N-methyltransferase 1 (EHMT1)) or histone acetylation and deacetylation (histone deacetylase 4 (HDAC4))21. The convergence of mutations causing ASD suggests that patients could be clustered on the basis of similar pathophysiological mechanisms at the signalling level rather than similarities at the genetic or behavioural levels. According to this hypothesis, drugs developed for a specific genetic condition would also benefit patients affected by mutations in genes involved in the same signalling pathway. In colour, proteins with genetic variants associated with ASD, which are listed in TABLE 1; in white, proteins not directly associated with ASD; asterisks indicate proteins that are associated with a syndromic form of ASD for which specific pharmacological treatments are in development. CNTNAP2, contactin associated protein-like 2; MEK, MAPK/ERK kinase; PI3K, phosphoinositide 3-kinase.

DYRK1A

CREBCBP

EP300

RAF

RAS

MEK ERK

HDAC4*

NF1*

TSC1*

TSC2*

MECP2*

EHMT

PTEN*

PI3K AKT

AMPK

mTOR

mRNA translationGTP

FMRP*

Nucleus

Cytoplasm

Synapse ContactinCadherinAMPARNMDARmGluRNeuroligin3-4

SHANK3

SAPAP2

PSD95SAP97 CNTNAP2

Homer

SHANK2


R E V I E W S



Predictive validityAn animal model’s ability to help identify drugs with potential therapeutic value in humans (e.g., antidepressant drugs reliably stimulate escape behaviours in the forced swim test, so the assay predicts antidepressant efficacy even if it does not measure depression).

section of this Review. Overall, the encouraging observa-tions made in patients with FXS receiving treatment with an mGluR5 antagonist, minocycline, lithium or arba-clofen indicate that the Fmr1-knockout mouse model may also have predictive validity for human FXS.

TSC is an autosomal dominant disorder in which non-malignant tumours grow in multiple organs, includ-ing the brain. In addition, patients present with seizures, developmental delay and behavioural problems, includ-ing social avoidance and communication deficits. TSC is caused by inherited (~30% of cases) or de novo (~70% of cases) mutations in the TSC1 or TSC2 genes that result in disinhibited mTOR signalling31. In mice, the neuronal phenotype associated with the absence of TSC1 includes cell body hypertrophy, low spine density and increased surface expression of AMPA receptors32, as well as impaired formation of cortical layers, hypomyelination, epileptic seizures and limited survival33. Constitutive reduction of TSC2 activity by 75% leads to a less severe phenotype, which includes learning and memory deficits and facilitation of late-phase LTP5. In both mouse mod-els of TSC, inhibition of mTOR activity with rapamycin rescues these phenotypes5,9, which suggests that mTOR inhibitors could have disease-modifying effects in patients with TSC. These results from mouse models of TSC have translational value, as rapamycin treatment (sirolimus or its derivative, everolimus) is effective in diminishing tumour size in lungs, kidneys and brain in patients with TSC, and in improving specific cognitive functions34–36.

The cases of FXS and TSC illustrate the value of mouse models that reproduce the human genetic basis of ASD and have strong construct and face validity. An important question to address is whether the therapeutic effect of a treatment designed on the basis of a mecha-nistic understanding of a syndromic form of ASD will also be effective for idiopathic ASD cases. Attempts to address this question with mice have led to contrasting results. On the one hand, mGluR5 inhibition was shown to lower repetitive behaviours and to increase social responses in the BTBR mouse model for idiopathic ASD, thus suggesting that mGluR5 antagonists would also be beneficial for non-FXS autistic patients37. On the other hand, the Tsc2+/– mouse model for TSC exhibit molecu-lar and physiological phenotypes that are opposite to the ones observed in Fmr1-knockout mice. Moreover, in the Tsc2+/– mouse model, the phenotype was corrected by increasing, instead of decreasing, mGluR5 activity, thereby suggesting that patients with different forms of ASD may require different treatments38.

Another important lesson learned from the TSC and FXS examples is that identifying a genetic cause for syn-dromic ASD opens up a wider therapeutic space beyond the mutated gene or protein, which may be therapeutically relevant for sporadic forms of ASD. Correction, or com-pensation, for this genetic defect can come from targeting either upstream or downstream genes or proteins involved in the same pathway. This is exemplified by the use of an mGluR5 antagonist, a GABAB receptor agonist, minocy-cline, lithium or rapamycin in FXS and TSC, which do not act directly on FMRP, TSC1 or TSC2, but indirectly compensate for the absence of these proteins (FIG. 3).

Selecting targets and developing drugs for ASDAn important revelation from the identification of ASD-associated genes is that rare mutations in one of many genes can lead to typical ASD39. An assessment of vari-ous mutations associated with ASD indicates that many of these genes have a role in regulating synaptic func-tion39 (TABLE 1). For example, ASD-associated mutations have been described in neurexin 1(NRXN1), neuroligin 3 (NLGN3), NLGN4, SH3 and multiple ankyrin repeat domains 3 (SHANK3), glutamate receptor, ionotropic, N-methyl d-aspartate 2A (GRIN2A) and ubiquitin pro-tein ligase E3A (UBE3A), which are all involved in regu-lating excitatory synapse function. However, mutations in any single gene account for a very small fraction of ASD cases.

These observations suggest that synaptic dysfunction is likely to be a core feature of ASD, but rule out the possibility of devising a therapeutic strategy that aims at correcting the deficit of individual genes. Moreover, the heterogeneity of the clinical presentation of ASD and the poor correlation between symptoms and pathophysiol-ogy also excludes the possibility of a therapeutic strategy that aims at ameliorating symptoms.

We therefore postulate that the search for drugs in ASD should focus on the definition of clusters of conver-gent pathophysiology and aim to restore normal synaptic and neural circuit function rather than to correct gene dysfunctions or target behavioural phenotypes (FIG. 4). Key considerations in selecting targets for drug discov-ery in ASD are whether such targets are tractable for drug development, and these considerations are discussed in the next section.

Selecting druggable targets for ASD. Selection of a drugga-ble target consists of identifying molecular processes that can be enhanced or inhibited in order to restore homeo-stasis. It is often the case that drugs do not directly target the mutated gene product but compensate for this altera-tion by acting on proximal or distal processes. Moreover, the choice of the druggable target is constrained by the pharmacological toolset currently available. There are a few good examples of syndromic forms of ASD in which understanding the genetic basis of the condition has led to the identification of tractable targets, of which the so-called “mGluR5 theory of FXS” is the most well known23. But there are also examples in which even with the understanding of the molecular basis of disease, the presumably natural targets are not easily druggable, such as in Angelman syndrome caused by mutations in UBE3A40. Thus, tractability is a highly important consid-eration for the development of viable therapies.

An analysis of the new molecular entities (NMEs) approved by the US Food and Drug Administration (FDA) during the past 5 years provides a good indication regarding the nature of the druggable targets for which drugs are successfully developed. During the period 2008–2012, a total of 131 pharmacologically active NMEs (that is, excluding contrast agents) received FDA approval for marketing (data taken from the Drugs@FDA website; see Further information). Among them, 75% are small molecules (98 NMEs) and 25% are biological compounds

R E V I E W S



(33 NMEs); notably, so far, no biological compound has been approved for the treatment of psychiatric or devel-opmental disorders of the central nervous system (CNS). Considering all indications together, the top two modes of actions that collectively comprise over three-quarters of all NMEs are the modulation of transmembrane pro-tein activity (47.3%) and the modulation of enzyme activity (30.5%) (FIG. 5). Alternative modes of actions targeting, for example, protein–protein interactions or gene expression regulation, have currently an anecdotal importance in terms of the number of NMEs accessible to patients. This is even more true for CNS diseases, for

which all NMEs approved by the FDA during the past 5 years exclusively target transmembrane proteins and enzymes. Two divergent conclusions can be drawn from this observation. On the one hand, current pharmaco-logical treatments target a limited number of biological processes, and the possibilities for new drug targets and innovative modes of action are almost unlimited. On the other hand, the almost complete absence of compounds based on novel modes of action is a good indicator of the extreme scientific and technical challenges involved in developing innovative therapeutic modalities that are suitable for patient use.

Figure 3 | Highly interconnected pathways offer opportunities for drug development. The postsynaptic glutamatergic synapse is a dense network of highly interconnected proteins (for definitions of the protein symbols please refer to the Universal Protein Resource (UniProt) database). The translation repressor fragile X mental retardation protein (FMRP), the absence of which causes fragile X syndrome (FXS), regulates the expression level of many of these proteins (orange symbols). Inhibitors of metabotropic glutamate receptor 5 (mGluR5; also known as GRM5) are currently in clinical development for FXS, and proteins directly or indirectly associated with mGluR5 (blue symbols) are highly responsive to mGluR5 pharmacological modulation. The overlap between the FMRP-regulated and the mGluR5-sensitive networks (red symbols) illustrates how mGluR5 inhibition has the capacity to compensate for the absence of FMRP in FXS. More generally, the high density of biological networks and their high level of interconnectivity offer many possibilities to compensate for the effect of a specific genetic mutation by acting pharmacologically on a distal target. The complexity of biological networks, which contrasts with the usual oversimplified representation of signalling pathways (see FIG. 2 for example), is an underexplored source of novel targets for drug development. Method: the postsynaptic glutamatergic network was built by retrieving protein interaction partners to the GRMs and DLG4 from the STRING (Search Tool for the Retrieval of Interacting Genes/Proteins) database84. The FMRP-regulated proteins are defined on the basis of the FMRP-binding RNA transcripts identified by Darnell et al. 85 using HIT-CLIP in vivo. The graphical representation was produced in Cytoscape86.


NLGN2

NLGN3

GRIN1

RELN GRIA1 GRIN2A GRIN2B GRIN2C

GRIN2D

GRIK1GRIK5

GRM1

CREBBP EP300

SP1 CREB1

CAMK2D CAMK2G

CAMK2A

SHANK1

APBA1ADRBK1

PTK2

PTK2B

PRKAR2A

PRKACG

HRAS

RASGRF1

NDOR1

PTPN11

CIT

NOS1

AKAP5 AKAP9 ADCY8 SYNGAP1HTT PRKCG DAPK1PRKCB

SHANK2

AP2B1

HOMER1

HOMER2

DLG1

DLG2

DLG3

DLGAP1DLGAP2

DLGAP3GNAO1

GNASGNAZ

ARHGEF7ARHGAP32

FYN MAP2 MAP1A MAP1B

CAMK2B

CALM1

DLG4

GRM5

GRM2GRM3GRM4

GRM6

GRM7

GRM8 ITPR1

ATP2B2

ATP2B4

LRP1

LRP8

KCNA2

KCND2

ASIC2

LPHN3CACNG2

Synapse

Nucleus

Membrane

Cytoplasm

R E V I E W S



Creating drugs for CNS disorders requires highly stringent chemistry, as beyond the typical drug-like char-acteristics compounds need to be able to achieve stable CNS exposure, which requires crossing the blood–brain barrier, and avoiding active removal from this compart-ment. Oral dosing is, in general, the preferred administra-tion route, which means compounds should be soluble and stable in the gastrointestinal tract and resistant to first-pass liver metabolism, and they should be sufficiently stable in vivo to support a reasonable dosing regimen. In the case of ASD, the prospect of life-long treatment, pos-sibly starting at juvenile age, also puts high requirements on treatment safety, absence of side effects and long-term toxicity. In particular, the possibility for transgenera-tional toxicity should be evaluated in appropriate species. Optimum treatment efficacy in ASD will presumably require starting treatment at young ages, which would probably exclude potential drug targets such as GSK3β and mTOR that have pleiotropic function and are essen-tial for body growth. Finally, a large proportion of the ASD patient population, especially in the United States of America, is already receiving pharmacological treat-ments for symptom management. This is an additional constraint on new treatments, which should be compat-ible with ongoing treatments and present as little drug–drug interaction as possible. While drug metabolism is well described in adults and children, drug metabolism is more variable and less well characterized before the age of 5 years. As such, the translation from in vitro testing to in vivo pharmacodynamic and pharmacokinetic pre-dictions is much less established in young children than in adults.

Patient-derived iPSCs for drug screening and testing. A central requirement for successful drug screening strategies is the availability of suitable in vitro assays. Most commonly, these are used to identify compounds

that bind to or activate a target of interest and are a cor-nerstone of the drug development process. Whereas this approach can be used to screen compounds for molecu-lar targets that are known to or thought to underlie the pathophysiology of ASD, the lack of validated targets poses a particular challenge. In the case of disorders such as ASD, in which the molecular pathophysiology is poorly understood, there is a need to develop platforms that can be used to identify ASD-associated phenotypes and to use for screening campaigns.

The recent advances in generating patient-derived iPSCs provide a unique opportunity to adapt this emerg-ing technology for drug discovery in ASD. The proce-dures for differentiating iPSCs into neurons of defined identities are fairly well established. For example, iPSCs can be differentiated into neurons that have character-istic features of motor neurons or cortical neurons41,42. A particularly exciting advance in the field has been the demonstration that iPSCs derived from patients with neurodevelopmental disorders exhibit phenotypes dis-tinct from those of cells from healthy volunteers. For example, Muotri and colleagues reported that iPSC-derived neurons from patients with Rett syndrome have a defect in synapse density and network activ-ity43. Similarly, it has been reported that iPSC-derived neurons from patients with Timothy syndrome display electrophysiological defects44. These studies suggest that patient-derived iPSCs provide an avenue for identifying phenotypes that can be used for drug discovery.

In the case of ASD, the genetic heterogeneity poses a potential challenge to identifying reliable cellular phe-notypes. One way to increase the likelihood of identify-ing robust phenotypes would be to start with patients that carry mutations in known risk-factor genes for ASD as they are likely to share cellular and molecular alterations. The most useful phenotypic assessments would be those that are relevant to the disorder and

Figure 4 | Points of intervention for ASD. Numerous rare genetic mutations have been linked to autism spectrum disorder (ASD). Many of these converge on molecular pathways related to the regulation of synapses and neuronal circuits, which are probably responsible for the behavioural symptoms associated with ASD. Targeting synaptic and neural circuit dysfunction in ASD may provide a tractable approach for development of new therapies.


Synapses

Target for pharmacological intervention

Circuits BehaviourGenes

Social impairment

Restrictedinterests

and repetitivebehaviours

R E V I E W S



maintained in vitro. In particular, analysis of cellular synaptic development as assessed by immunofluores-cence, morphometry, functional imaging and elec-trophysiology should reveal ASD-associated cellular phenotypes. In addition, analysis of gene expression could provide important insights into molecular path-ways that are altered in ASD. Comparison of these phenotypes across patient-derived iPSCs should allow the identification of shared phenotypes that could be the target of phenotypic reversal by pharmacological intervention. Compounds that reverse ASD-associated cellular phenotypes could be tested in animal models of ASD, and should lead to the identification of drug candidates for clinical development (FIG. 6).

Testing druggable targets in animal models for early proof-of-concept studies. An essential stage in the drug development process is the evaluation of new com-pounds for their therapeutic potential using specific animal models mimicking disease symptoms and/or pathophysiology. Animal models are usually evaluated according to their construct, face and predictive validity.

An extensive repertoire of mutant mouse lines bear-ing single ASD-causing mutations has been created by genetic engineering. These mice are able to reproduce phenotypes associated with syndromic forms of ASD, as well as carry very rare mutations associated with ASD. The major advantage of these genetic models is their excellent construct validity. However, the main draw-back is the limited size of the patient population being modelled. Besides single-gene models, the BTBR mouse strain has been identified as an inbred strain showing a range of ASD-like behaviours45, and is therefore, on the basis of face validity, an interesting model of idiopathic ASD. Finally, intrauterine exposure to toxic chemicals (such as valproic acid) or to immune challenges are used

to induce cognitive and behavioural deficits in rodents. This approach is used to model environmental risk factors for ASD46. Although the understanding of the pathophysiological mechanisms in such models is lim-ited, they offer at least the possibility of testing whether therapeutic approaches developed for a specific muta-tion could also be beneficial to a larger fraction of the patient population.

The evaluation of pharmacological effects of prospec-tive therapeutics relies on behavioural tests related to the core and co-morbid symptoms of ASD47. Social behav-iours are characterized by means of direct and indirect social interaction tests, which differ essentially in the pos-sibility for the animals to engage or not in direct contact with each other. The three-chamber test, for example, evaluates the presence of a natural preference for social contacts over exploration of a novel object and has been instrumental in revealing a lack of social preference in various animal models of ASD48. Rodents emit ultrasonic vocalizations (USVs) under specific circumstances; for example, male mice emit USVs in response to the scent of an oestrous female, and mouse pups emit USVs upon physical separation from the mother. A deficit in USVs, interpreted as a phenotype resembling communication deficits in patients with ASD, has been observed in specific mouse models for ASD, such as in BTBR mice49, in TSC2 mice50, and in Fmr1-knockout mice51,52. Furthermore, phenotypes such as excessive grooming, low level of alter-nations in a Y-shaped maze, and absence of preference for novelty, which have been observed in numerous mouse strains, are interpreted in light of the repetitive behaviours and preference for sameness in patients with ASD. Thus, a broad repertoire of behavioural tests offers the possibility to evaluate pharmacological effects on all core symptoms of ASD, using models that reproduce the heterogeneity in the aetiology of the disorder.

With regard to predictive validity, discussion is restricted by the limited number of molecules approved for the treatment of specific symptoms in ASD, and the absence of disease-modifying treatments on the market. It will be interesting to see whether the congruent behav-ioural effects observed in different animal models upon treatment are confirmed in clinical trials.

Challenges for clinical development in ASDNew molecules emerging from this stringent preclini-cal development process then need to be taken through an equally difficult clinical development process, which offers challenges beyond what are typically found for other neuropsychiatric disorders. One initial paradox is that while optimal efficacy for pharmacological treat-ment of neurodevelopmental disorders probably comes from intervention in childhood (offering the prospect of disease modification or correction of a developmental trajectory), traditional drug development and regulatory pathways require demonstration of safety and potential prospect of direct benefit in adult populations before paediatric studies can be conducted. While it is under-standable that stringent safety requirements should be met before younger populations are exposed to experi-mental medicines, there is cause to believe that efficacy

Figure 5 | Modes of action for the FDA-approved NMEs. New molecular entities (NMEs) approved by the US Food and Drug Administration (FDA) were retrieved from the FDA website (see Further information) for the 5 year period 2008–2012. Information related to the mode of action of the drugs were collected from literature searches.


0 20 40 60 80 100

Drug modes of action (%)

Allindications

Centralnervous

systemdiseases

47.3

88.9 11.1

30.5 7.6

Receptor, channeland transporteractivity modulation

Enzymeactivity

modulationSoluble

factor capture

Protein–proteininteraction modulation

DNA and geneexpression modulation

Synapticreleasemodulation

Others

R E V I E W S



Aberrant Behaviour Checklist(ABC). A symptom checklist developed to assess problem behaviours of children and adults with severe intellectual disability, mapping behaviours into 5 subscales: irritability and agitation, lethargy and social withdrawal, stereotypical behaviour, hyperactivity, and inappropriate speech.

Social Responsiveness Scale(SRS). A scale for quantitative measure of autism spectrum symptoms as they occur in natural social settings, specifically for social impairments such as social awareness, social information processing, capacity for reciprocal social communication, social anxiety/avoidance, and autistic preoccupations and traits. The SRS generates scores for five subscales: receptive, cognitive, expressive, and motivational aspects of social behaviour, as well as autistic preoccupations.

in adult patients might not be achievable in some cases, or might not be fully predictive of the potential thera-peutic efficacy in younger age ranges.

There is very limited experience in clinical trials for ASD, including for such key design elements as patient selection criteria or the definition of appropriate end points, which can significantly reduce the probability of detecting clinical benefit even for highly active and pre-clinically validated compounds. Additionally, although the only approved drugs for treating ASD are antipsy-chotics that target very specific behavioural problems, patients with ASD are usually medicated with several other classes of CNS-active compounds for co-mor-bidities (such as mood and anxiety disorders, epilepsy and behavioural problems), and often receive additional non-pharmacological interventions (such as physical therapy, psychotherapy and special educational sup-port). Given that it would be difficult to standardize all these therapeutic variables in a clinical trial, detecting the treatment effects of a new compound is therefore much more complex.

Looking at the known clinical heterogeneity in ASD, it is inadequate that crucial success factors such as selec-tion of patients or their stratification in trials are cur-rently based only on surface phenotypic characteristics, such as IQ, level of functioning (high versus low), pres-ence of specific behavioural problems (for example, irri-tability) or other co-morbidities. The current situation does not allow a rational matching of drugs or targets to patients, which emphasizes the urgent need to develop biomarkers that could be used to either categorize diag-nosis or predict response to new therapies. Likely candi-date biomarkers could come from molecular signatures detectable in peripheral cells or fluids (such as high-con-tent genetic or epigenetic, genomic or proteomic analysis

of iPSC-derived neurons, or simply blood samples from patients), or the use of electrophysiology or brain imag-ing to interrogate key brain circuitry.

These biomarkers might then be used to both clus-ter patients in pathophysiologically relevant diagnostic groups, such as dysfunctions in synaptic, protein or gene regulation (as discussed above; FIG. 2), or imbalances in excitatory–inhibitory neurotransmitter levels, and also to identify groups of ‘responders’ to specific drug targets or modes of action. This would allow a more rational selection of patients for trials and increase probabilities of success. So far, this approach has been most success-fully accomplished in Alzheimer’s disease, in which the identification of amyloid or tau pathology as key target pathways, together with the development of imaging (for example, ligands for positron emission tomography imaging) and molecular (for example, cerebrospinal fluid) biomarkers has allowed a redefinition of diagnos-tic categories and earlier intervention.

Likewise, clinical end points that have been used in past trials, such as the Aberrant Behaviour Checklist (ABC) were not initially developed for use in clinical development. Consequently, they suffer from diverse psychometric problems and there is a lack of under-standing of their reliability in the context of clinical tri-als. Furthermore, end points targeting core symptoms of ASD, such as the Social Responsiveness Scale53 have only recently been developed and validated, and have yet to be used in larger scale clinical trials.

Overcoming these challenges is an endeavour that requires the pooling of resources and knowledge from several stakeholders, including the pharmaceutical indus-try, health authorities, patient associations and academic and clinical investigators. Indeed, there are now several precompetitive initiatives, such as the European Autism Interventions — A Multicentre Study for Developing New Medications (EU-AIMS) project from the Innovative Medicines Initiative (BOX 2), which aims to achieve this goal. Additionally, we, at Roche, have modified our early development strategy to consistently include explora-tory, non-interventional and observational studies to explore biomarkers and validate end points. These are run in close connection to the early development stages of new medicines; for example, looking at end points for an age-matched population during the presumed dura-tion of treatment in Phase II or Phase III, to help design these studies by taking into account their variability and psychometric performance (FIG. 7).

Another approach that should improve translatability and help overcome the above-mentioned challenges in ASD drug discovery is focusing clinical development of new therapies to syndromic forms of ASD with known pathophysiology, such as FXS or TSC. In fact, several Phase I and Phase II clinical studies in FXS have been published and at least two Phase III programmes are cur-rently underway (sponsored by Novartis54 and Seaside Therapeutics55). Besides the fact that there is now evi-dence that FXS may be amenable to pharmacological modulation, exploring the results of these studies offers important lessons for future drug development in FXS and potentially ASD.

Figure 6 | Use of iPSCs for ASD drug discovery. Patient-derived fibroblasts can be reprogrammed to generate induced pluripotent stem cells (iPSCs) that can be differentiated into neurons to assess cellular phenotypes associated with autism spectrum disorder (ASD). These phenotypes can provide a basis for screening for compounds that reverse ASD-associated phenotypes, which could be tested in animal models and eventually be developed for clinical studies.


Drugtherapy

Patient Patient-derivedskin fibroblasts

Drug screening Characterization ofdisease phenotype

iPSC linesNeuronaldifferentiation

R E V I E W S



Vineland Adaptive Behaviour Scale(VABS). A semi-structured interview instrument that measures adaptive behaviour in children and adults, covering major personal and social skills required for everyday living. It is commonly used to support the diagnosis of intellectual and developmental disabilities including autism. Five major domains (including specific subdomains) are assessed, including communication, daily living skills, socialization, motor skills and maladaptive behaviour.

Repeatable Battery for the Assessment of Neuropsychological Status(RBANS). A brief neuro-psychological battery that assesses cognitive decline or improvement, used as a neuropsychological screening battery for younger patients (down to 12 years). It tests five cognitive domains: immediate memory, visuospatial/constructional, language, attention and delayed memory.

Visual Analogue ScaleAn instrument to assess subjective characteristics or attitudes that cannot be directly measured, in which respondents specify their level of agreement to a statement by indicating a position along a continuous line between two end points.

Clinical Global Impression (CGI) scaleA commonly used summary measure of a patients’ global functioning. The CGI Severity is a seven-point categorical scale (from normal to extremely ill) that rates the severity of the disease at the time of assessment in comparison with past experience of patients with the same diagnosis. The CGI Improvement is a seven-point categorical scale (from very much improved to very much worse) that measures change in patient status from baseline in response to an intervention.

Several small-scale, open label clinical trials have been conducted in recent years in FXS. However, we should not underestimate the problems associated with short duration and especially open-label studies, which are prone to large placebo effects. An initial study of lithium in 15 patients with FXS showed improvements in several behavioural scales (ABC-C, Vineland Adaptive Behaviour Scale (VABS) and Repeatable Battery for the Assessment of Neuropsychological Status (RBANS))56, but these results are yet to be replicated in a controlled manner.

Similarly, minocycline was tested initially in an open-label, short duration study in 20 adult and adolescent patients, and showed significant improvement in the ABC-C-Irritability subscale, as well as in a Visual Analogue Scale for behaviour and in the Clinical Global Impression (CGI) scale57. Recently, results of a double-blind, placebo-controlled trial in 55 infant and adolescent patients (aged 3.5–16 years) treated for 3 months have been published, with less positive results. In this trial, only the CGI was significant, and the authors themselves highlighted potential design issues related to unblinding and investi-gator bias, which may have negatively affected the study58. These findings illustrate the need to take into account methodological problems, such as placebo effects related to investigator and patient or caregiver bias, in the design of future clinical trials in ASD.

Alternative benefits of small pilot studies have been the development of clinical translational tools with con-struct validity for the core symptoms of ASD and FXS.

Imaging tools such as functional magnetic resonance imaging59, or behavioural surrogates such as eye-track-ing and pupillometry60 have been used in small-scale settings to detect differences in the ‘social circuitry’ of patients versus controls, but also to detect drug effects, notably for oxytocin61,62. It remains to be seen how reli-able and scalable these tools are across different research sites, which are a pre-requisite for their widespread use in drug development.

More interestingly, positive initial results from larger scale controlled trials in FXS using an mGluR5 antago-nist (AFQ056) and GABAB receptor agonists (STX209 and arbaclofen) have generated much enthusiasm, but have also illustrated additional challenges in trial design that need be taken into account.

In a proof-of-concept study of 30 adult patients with FXS receiving AFQ056 (double-blind, placebo-con-trolled, up-titration two-period crossover), the primary efficacy analysis on the ABC-C scale was negative, but a subpopulation analysis in patients with FMR1 ‘full meth-ylation’ (seven patients, defined by concordant results in PCR and bisulphite sequencing) revealed significant results, mainly in the ABC-Irritability subscale54. Should these findings be confirmed, it would appear that FMR1 methylation status might be a predictive biomarker for response to therapy (at least for mGluR5 antagonism), highlighting the fact that even in more homogenous syndromic ASD there are additional levels of biological complexity to consider in trial design for confirmatory

Box 2 | Industry–academic collaborations in drug discovery for ASD

Industry–academic collaborations are an integral part of today’s drug discovery and development landscape. The size and complexity of such collaborations range from simple bilateral exchanges of tools and expertise, to large international consortia with multiple stakeholders and multimillion-dollar budgets operating over many years.

In its simplest form, industry–academic collaborations are based on the exchange of tools such as pharmacological compounds, antibodies, cell lines or animal models. These exchanges exist in both directions, which may also include consultancy and fee-for-service work, and their main purpose is to provide the other party with tools or services that are not commercially available. In addition to that, training programmes can nurture collaborations between academic partners and scientists from industry on specific projects.

On a larger scale, research foundations have an important role in structuring collaborations between industry and academic institutions. Research foundations typically aim for advancing academic research and drug discovery in a particular indication area, and facilitate the exchange of scientific expertise and tools between partners. For instance, the Simons Foundation has played a crucial role in supporting much of the research that led to the identification of rare genetic mutations in autism spectrum disorder (ASD). In addition, the Simons Foundation Autism Research Initiative and the Fondation FondaMentale are actively promoting the development of databases, sample repositories, disease models, often in the form of public and private partnerships, that will support clinical trials. Autism Speaks is another foundation that has supported much work to increase our understanding of ASD, and is an active participant in the European Autism Interventions — A Multicentre Study for Developing New Medications (EU-AIMS) from the Innovative Medicines Initiative (IMI), described below.

Recently, very large public–private partnerships involving numerous stakeholders from the pharmaceutical industry, academic institutions and government agencies have been set up to conduct research on a precompetitive basis.

The IMI is Europe’s largest public–private initiative aiming to accelerate the development of new medicines for specific indications, such as diabetes, oncology, pain and ASD. Related topics critical for drug development are also addressed, such as the improved handling of patient data, the development of biomarkers or improved chemical manufacturing techniques73,74.

The EU-AIMS project is focused on ASD75, and is an international collaboration led by scientists from Roche and Kings College London, UK. It consists of 14 European academic centres, the patient organization Autism Speaks, representatives of patients and caregivers (Autism Europe), three medium-size enterprises and six members of the European Federation of Pharmaceutical Industries and Associations. Key objectives of EU‑AIMS are to develop and validate translational approaches for the advancement of novel therapies, setting new standards in research and clinical development to foster drug discovery, and to identify and develop expert clinical sites to conduct clinical trials in ASD.

R E V I E W S



http://sfari.org/

http://sfari.org/

http://www.fondation-fondamental.org/index.php?lang=EN

http://www.autismspeaks.org/

http://www.eu-aims.eu/

http://www.autismeurope.org/

trials. In this case, different levels of silencing of the FMR1 gene are associated with synaptic dysfunction and clinical response.

Arbaclofen, the r-enantiomer of baclofen (STX209) has been studied in children and adults with both FXS and ASD, and results of a Phase II study involving 63 patients with FXS have recently been published55. Similar to the mGluR5 trial54, the primary efficacy analysis using the ABC-Irritability subscale was negative, although trends were found in several other behavioural scales (CGI and VAS)55. However, in a post-hoc analysis using the ABC Social Avoidance subscale (an FXS-specific refactoring of the ABC that was previously validated in 630 patients)63 arbaclofen showed significant effects. Based on these results, two Phase III studies (one in adults and one in children and adolescents) were started. It can be legitimately argued that in these settings, refac-toring of existing end points may be required to detect treatment effects in ASD, and that additional strategies such as defining response by a combination of improve-ment in specific behavioural end points and more global measures of functioning might be required to improve detection. Unfortunately, Seaside Therapeutics publicly announced a decision (see Further information) to ter-minate the open-label extension phase of the adult study, based on lack of efficacy on this primary end point, but no further details have been revealed and results from the paediatric study have also not yet been publicly disclosed.

Figure 7 | Building clinical capabilities in neurodevelopmental disorders. During the early development stages (Phase I–II) of new projects, beyond the core clinical safety and efficacy studies, an array or additional exploratory studies (for example, observational, imaging and biomarker) are conducted to generate a comprehensive data package to allow a decision to progress to confirmatory development (Phase III). EiH, entry into human; PET, positron emission tomography.


Single ascending dose• EiH healthy volunteers• Safety and tolerability• Pharmacokinetics

PET study• Healthy volunteers and patients• Receptor occupancy

Multiple ascending dose• Healthy volunteers and patients• Safety and tolerability• Pharmacokinetics• Pharmacodynamics

Proof-of-concept• Patients• Safety and tolerability• Clinical efficacy (end point)

Observational studies• Patients (target population)• Clinical scale validation• Biomarkers

Functional imaging study• Healthy volunteers and patients• Proof-of-mechanism• Dose selection

Biomarker studies• Patients • Diagnostic development

Biomarker studies• Patients • Diagnostic development

Phase I Phase II Phase III

Finally, results of a 12-week placebo-controlled Phase IIb study of arbaclofen in 150 patients with ASD were recently presented64. Once again, although the primary end point (ABC-Lethargy Social Withdrawal) was not met, a significant improvement was seen on a global scale (CGI-S), and a post-hoc analysis of the VABS detected benefits in socialization for patients with higher IQ (>70) or in which the scale had been administered by the same clinician and caregiver.

Reproducibility in psychiatric drug development continues to be a major challenge, and only when full results have been analysed will it be possible to discern whether there are lessons to be learned from these stud-ies that help progress the field. Taken together, however, they do re-emphasize that success in drug development for ASD requires meticulous care in clinical develop-ment, including pathophysiology-related biomarkers for patient stratification (of which FRM1 methylation could be the first example) and end point validation that might include refactoring of existing scales, development of new scales or combination of multiple tools to detect true responders to treatment.

Conclusions and next stepsRemarkable progress has been made in creating the foundations needed to embark on drug development for ASD and other neurodevelopmental disorders. The elucidation of key pathophysiological pathways, as well as the explosion of new genetic data for syndromic and

R E V I E W S



non-syndromic ASD, has helped create a landscape for target exploration that, until recently, was non-exist-ent. Interestingly, this landscape is not too unfamiliar and has already yielded several potentially druggable targets, especially in synaptic proteins and key neuro-transmitter systems. Animal models with good face and construct validity exist for several of the more common syndromic ASD. Indeed, phenotypic models that rep-licate the deficits in social behaviour, communication deficits and repetitive and restrictive behaviours are now available to test NMEs. Much effort is being made to develop and characterize animal models with good construct validity based on ASD-associated mutations, which should be of great value for preclinical drug development.

We see an urgent need to lay the groundwork for clini-cal development in ASD. Importantly, there is much to be done in terms of identifying and validating biomarkers that can be used for the stratification of patients based on pathophysiology. The possibility of novel approaches, such as characterization of patent-derived iPSCs hold promise, but are far from proof of concept. There are initial signs of hope, however, that with the application of translational tools, such as imaging, careful adapta-tion of existing clinical trial methodology, and attention to patient and end point selection, including early trials in paediatric patients, it will be possible to dramatically change the therapeutic landscape for neurodevelopmen-tal disorders, creating new medicines that will improve the lives of patients and their families.

1. Kanner, L. Autistic disturbances of affective contact. Nerv. Child. 2, 217–250 (1943).

2. Asperger, H. The psychically abnormal child, Wien Klin. Wochenschr. 51, 1314–1317 (1938).

3. Centers for Disease Control and Prevention. Prevalence of autism spectrum disorders — Autism and Developmental Disabilities Monitoring Network, 14 sites, United States, 2008. MMWR Surveill. Summ. 61, 1–19 (2012).

4. Politte, L. C. & McDougle, C. J. Atypical antipsychotics in the treatment of children and adolescents with pervasive developmental disorders. Psychopharmacology (Berl). 2013 (dx.http://dx.doi.org/10.1007/s00213-013-3068-y).

5. Ehninger, D. et al. Reversal of learning deficits in a Tsc2+/– mouse model of tuberous sclerosis. Nature Med. 14, 843–848 (2008).

6. Michalon, A. et al. Chronic pharmacological mGlu5 inhibition corrects fragile X in adult mice. Neuron 74, 49–56 (2012).

7. Costa, R. M. et al. Mechanism for the learning deficits in a mouse model of neurofibromatosis type 1. Nature 415, 526–530 (2002).

8. Tropea, D. et al. Partial reversal of Rett syndrome-like symptoms in MeCP2 mutant mice. Proc. Natl Acad. Sci. USA 106, 2029–2034 (2009).

9. Meikle, L. et al. Response of a neuronal model of tuberous sclerosis to mammalian target of rapamycin (mTOR) inhibitors: effects on mTORC1 and Akt signaling lead to improved survival and function. J. Neurosci. 28, 5422–5432 (2008).

10. Ronald, A. & Hoekstra, R. A. Autism spectrum disorders and autistic traits: a decade of new twin studies. Am. J. Med. Genet. B Neuropsychiatr. Genet. 156B, 255–274 (2011).

11. Kumar, R. A. & Christian, S. L. Genetics of autism spectrum disorders. Curr. Neurol. Neurosci. Rep. 9, 188–197 (2009).

12. Betancur, C. Etiological heterogeneity in autism spectrum disorders: more than 100 genetic and genomic disorders and still counting. Brain Res. 1380, 42–77 (2011).

13. Marshall, C. R. & Scherer, S. W. Detection and characterization of copy number variation in autism spectrum disorder. Methods Mol. Biol. 838, 115–135 (2012).

14. Kong, A. et al. Rate of de novo mutations and the importance of father’s age to disease risk. Nature 488, 471–475 (2012).

15. Gibson, G. Rare and common variants: twenty arguments. Nature Rev. Genet. 13, 135–145 (2012).

16. Barabasi, A. L., Gulbahce, N. & Loscalzo, J. Network medicine: a network-based approach to human disease. Nature Rev. Genet. 12, 56–68 (2011).

17. Peca, J. & Feng, G. Cellular and synaptic network defects in autism. Curr. Opin. Neurobiol. 22, 866–872 (2012).

18. Etherton, M. R., Tabuchi, K., Sharma, M., Ko, J. & Sudhof, T. C. An autism-associated point mutation in the neuroligin cytoplasmic tail selectively impairs AMPA receptor-mediated synaptic transmission in hippocampus. EMBO J. 30, 2908–2919 (2011).

19. Etherton, M. et al. Autism-linked neuroligin-3 R451C mutation differentially alters hippocampal and cortical synaptic function. Proc. Natl Acad. Sci. USA 108, 13764–13769 (2011).

20. Kelleher, R. J., 3rd & Bear, M. F. The autistic neuron: troubled translation? Cell 135, 401–406 (2008).

21. van Bokhoven, H. Genetic and epigenetic networks in intellectual disabilities. Annu. Rev. Genet. 45, 81–104 (2011).

22. Schneider, A., Hagerman, R. J. & Hessl, D. Fragile X syndrome — from genes to cognition. Dev. Disabil Res. Rev. 15, 333–342 (2009).

23. Krueger, D. D. & Bear, M. F. Toward fulfilling the promise of molecular medicine in fragile X syndrome. Annu. Rev. Med. 62, 411–429 (2011).

24. Dölen, G. et al. Correction of fragile X syndrome in mice. Neuron 56, 955–962 (2007).

25. Hur, E. M. & Zhou, F. Q. GSK3 signalling in neural development. Nature Rev. Neurosci. 11, 539–551 (2010).

26. Bilousova, T. V. et al. Minocycline promotes dendritic spine maturation and improves behavioural performance in the fragile X mouse model. J. Med. Genet. 46, 94–102 (2009).

27. Yuskaitis, C. J. et al. Lithium ameliorates altered glycogen synthase kinase-3 and behavior in a mouse model of fragile X syndrome. Biochem. Pharmacol. 79, 632–646 (2010).

28. Mines, M. A., Yuskaitis, C. J., King, M. K., Beurel, E. & Jope, R. S. GSK3 influences social preference and anxiety-related behaviors during social interaction in a mouse model of fragile X syndrome and autism. PLoS ONE 5, e9706 (2010).

29. Paluszkiewicz, S. M., Martin, B. S. & Huntsman, M. M. Fragile X syndrome: the GABAergic system and circuit dysfunction. Dev. Neurosci. 33, 349–364 (2011).

30. Henderson, C. et al. Reversal of disease-related pathologies in the fragile X mouse model by selective activation of GABAB receptors with arbaclofen. Sci. Transl. Med. 4, 152ra128 (2012).

31. Curatolo, P., Bombardieri, R. & Jozwiak, S. Tuberous sclerosis. Lancet 372, 657–668 (2008).

32. Tavazoie, S. F., Alvarez, V. A., Ridenour, D. A., Kwiatkowski, D. J. & Sabatini, B. L. Regulation of neuronal morphology and function by the tumor suppressors Tsc1 and Tsc2. Nature Neurosci. 8, 1727–1734 (2005).

33. Meikle, L. et al. A mouse model of tuberous sclerosis: neuronal loss of Tsc1 causes dysplastic and ectopic neurons, reduced myelination, seizure activity, and limited survival. J. Neurosci. 27, 5546–5558 (2007).

34. Davies, D. M. et al. Sirolimus therapy for angiomyolipoma in tuberous sclerosis and sporadic lymphangioleiomyomatosis: a phase 2 trial. Clin. Cancer Res. 17, 4071–4081 (2011).

35. Bissler, J. J. et al. Sirolimus for angiomyolipoma in tuberous sclerosis complex or lymphangioleiomyomatosis. N. Engl. J. Med. 358, 140–151 (2008).

36. Dabora, S. L. et al. Multicenter phase 2 trial of sirolimus for tuberous sclerosis: kidney angiomyolipomas and other tumors regress and VEGF- D levels decrease. PLoS ONE 6, e23379 (2011).

37. Silverman, J. L. et al. Negative allosteric modulation of the mGluR5 receptor reduces repetitive behaviors and rescues social deficits in mouse models of autism. Sci. Transl. Med. 4, 131ra51 (2012).

38. Auerbach, B. D., Osterweil, E. K. & Bear, M. F. Mutations causing syndromic autism define an axis of synaptic pathophysiology. Nature 480, 63–68 (2011).

39. Spooren, W., Lindemann, L., Ghosh, A. & Santarelli, L. Synapse dysfunction in autism: a molecular medicine approach to drug discovery in neurodevelopmental disorders. Trends Pharm. Sci 33, 669–684 (2012).

40. Mabb, A. M., Judson, M. C., Zylka, M. J. & Philpot, B. D. Angelman syndrome: insights into genomic imprinting and neurodevelopmental phenotypes. Trends Neurosci. 34, 293–303 (2012).

41. Kim, J. E. et al. Investigating synapse formation and function using human pluripotent stem cell-derived neurons. Proc. Natl Acad. Sci. USA 108, 3005–3010 (2011).

42. Dimos, J. T. et al. Induced pluripotent stem cells generated from patients with ALS can be differentiated into motor neurons. Science 321, 1218–1221 (2008).

43. Marchetto, M. C. et al. A model for neural development and treatment of Rett syndrome using human induced pluripotent stem cells. Cell 143, 527–539 (2010).

44. Pasca, S. P. et al. Using iPSC-derived neurons to uncover cellular phenotypes associated with Timothy syndrome. Nature Med. 17, 1657–1662 (2011).

45. McFarlane, H. G. et al. Autism-like behavioral phenotypes in BTBR T+tf/J mice. Genes Brain Behav. 7, 152–163 (2008).

46. Roullet, F. I., Lai, J. K. & Foster, J. A. In utero exposure to valproic acid and autism — a current review of clinical and animal studies. Neurotoxicol. Teratol. 36, 47–56 (2013).

47. Silverman, J. L., Yang, M., Lord, C. & Crawley, J. N. Behavioural phenotyping assays for mouse models of autism. Nature Rev. Neurosci. 11, 490–502 (2010).

48. Crawley, J. N. Translational animal models of autism and neurodevelopmental disorders. Dialogues Clin. Neurosci. 14, 293–305 (2012).

49. Scattoni, M. L., Ricceri, L. & Crawley, J. N. Unusual repertoire of vocalizations in adult BTBR T+tf/J mice during three types of social encounters. Genes Brain Behav. 10, 44–56 (2011).

50. Young, D. M., Schenk, A. K., Yang, S. B., Jan, Y. N. & Jan, L. Y. Altered ultrasonic vocalizations in a tuberous sclerosis mouse model of autism. Proc. Natl Acad. Sci. USA 107, 11074–11079 (2010).

51. Roy, S., Watkins, N. & Heck, D. Comprehensive analysis of ultrasonic vocalizations in a mouse model of fragile X syndrome reveals limited, call type specific deficits. PLoS ONE 7, e44816 (2012).

R E V I E W S



52. Rotschafer, S. E., Trujillo, M. S., Dansie, L. E., Ethell, I. M. & Razak, K. A. Minocycline treatment reverses ultrasonic vocalization production deficit in a mouse model of fragile X syndrome. Brain Res. 1439, 7–14 (2012).

53. Constantino, J. N. et al. The factor structure of autistic traits. J. Child Psychol. Psychiatry 45, 719–726 (2004).

54. Jacquemont, S. et al. Epigenetic modification of the FMR1 gene in fragile X syndrome is associated with differential response to the mGluR5 antagonist AFQ056. Sci. Transl. Med. 3, 64ra1 (2011).

55. Berry-Kravis, E. M. et al. Effects of STX209 (arbaclofen) on neurobehavioral function in children and adults with fragile X syndrome: a randomized, controlled, Phase 2 trial. Sci. Transl. Med. 4, 152ra127 (2012).

56. Berry-Kravis, E. et al. Open-label treatment trial of lithium to target the underlying defect in fragile X syndrome. J. Dev. Behav. Pediatr. 29, 293–302 (2008).

57. Paribello, C. et al. Open-label add-on treatment trial of minocycline in fragile X syndrome. BMC Neurol. 10, 91 (2010).

58. Leigh, M. J. et al. A randomized double-blind, placebo-controlled trial of minocycline in children and adolescents with fragile X syndrome. J. Dev. Behav. Pediatr. 34, 147–155 (2013).

59. Pelphrey, K. A., Shultz, S., Hudac, C. M. & Vander Wyk, B. C. Research review: Constraining heterogeneity: the social brain and its development in autism spectrum disorder. J. Child Psychol. Psychiatry 52, 631–644 (2011).

60. Farzin, F., Scaggs, F., Hervey, C., Berry-Kravis, E. & Hessl, D. Reliability of eye tracking and pupillometry measures in individuals with fragile X syndrome. J. Autism Dev. Disord. 41, 1515–1522 (2011).

61. Zink, C. F. & Meyer-Lindenberg, A. Human neuroimaging of oxytocin and vasopressin in social cognition. Horm. Behav. 61, 400–409 (2012).

62. Andari, E. et al. Promoting social behavior with oxytocin in high-functioning autism spectrum disorders. Proc. Natl Acad. Sci. USA 107, 4389–4394 (2010).

63. Sansone, S. M. et al. Psychometric study of the aberrant behavior checklist in fragile X syndrome and implications for targeted treatment. J. Autism Dev. Disord. 42, 1377–1392 (2012).

64. Veenstra-VanderWeele, J. et al. Randomized, controlled, Phase 2 trial of STX209 for social function in ASD. International Society for Autism Research [online], https://imfar.confex.com/imfar/2013/webprogram/Paper15177.html (2013).

65. Chakrabarti, S. & Fombonne, E. Pervasive developmental disorders in preschool children: confirmation of high prevalence. Am. J. Psychiatry 162, 1133–1141 (2005).

66. Simonoff, E. et al. Psychiatric disorders in children with autism spectrum disorders: prevalence, comorbidity, and associated factors in a population-derived sample. J. Am. Acad. Child Adolesc. Psychiatry 47, 921–929 (2008).

67. Mattila, M. L. et al. Comorbid psychiatric disorders associated with Asperger syndrome/high-functioning autism: a community- and clinic-based study. J. Autism Dev. Disord. 40, 1080–1093 (2010).

68. Boyd, B. A. et al. Sensory features and repetitive behaviors in children with autism and developmental delays. Autism Res. 3, 78–87 (2010).

69. Chez, M. G. et al. Frequency of epileptiform EEG abnormalities in a sequential screening of autistic patients with no known clinical epilepsy from 1996 to 2005. Epilepsy Behav. 8, 267–271 (2006).

70. Buie, T. et al. Evaluation, diagnosis, and treatment of gastrointestinal disorders in individuals with ASDs: a consensus report. Pediatrics 125 (Suppl. 1), 1–18 (2010).

71. Stefanatos, G. A. Regression in autistic spectrum disorders. Neuropsychol. Rev. 18, 305–319 (2008).

72. Fombonne, E. Epidemiology of pervasive developmental disorders. Pediatr. Res. 65, 591–598 (2009).

73. Goldman, M. The innovative medicines initiative: a European response to the innovation challenge. Clin. Pharmacol. Ther. 91, 418–425 (2012).

74. Laverty, H., Gunn, M. & Goldman, M. Improving R&D productivity of pharmaceutical companies through public–private partnership: experiences from the Innovative Medicines Initiative. Expert Rev. Pharmacoecon. Outcomes Res. 12, 545–548 (2012).

75. Murphy, D. & Spooren, W. EU-AIMS: a boost to autism research. Nature Rev. Drug Discov. 11, 815–816 (2012).

76. Klei, L. et al. Common genetic variants, acting additively, are a major source of risk for autism. Mol. Autism 3, 9 (2012).

77. Neale, B. M. et al. Patterns and rates of exonic de novo mutations in autism spectrum disorders. Nature 485, 242–245 (2012).

78. Michaelson, J. J. et al. Whole-genome sequencing in autism identifies hot spots for de novo germline mutation. Cell 151, 1431–1442 (2012).

79. Manolio, T. A. et al. Finding the missing heritability of complex diseases. Nature 461, 747–753 (2009).

80. Girirajan, S. et al. Phenotypic heterogeneity of genomic disorders and rare copy-number variants. N. Engl. J. Med. 367, 1321–1331 (2012).

81. Ollikainen, M. & Craig, J. M. Epigenetic discordance at imprinting control regions in twins. Epigenomics 3, 295–306 (2011).

82. McCarthy, M. I. et al. Genome-wide association studies for complex traits: consensus, uncertainty and challenges. Nature Rev. Genet. 9, 356–369 (2008).

83. Betancur, C., Sakurai, T. & Buxbaum, J. D. The emerging role of synaptic cell-adhesion pathways in the pathogenesis of autism spectrum disorders. Trends Neurosci. 32, 402–412 (2009).

84. Franceschini, A. et al. STRING v9.1: protein–protein interaction networks, with increased coverage and integration. Nucleic Acids Res. 41, D808–D815 (2013).

85. Darnell, J. C. et al. FMRP stalls ribosomal translocation on mRNAs linked to synaptic function and autism. Cell 146, 247–261 (2011).

86. Saito, R. et al. A travel guide to Cytoscape plugins. Nature Methods 9, 1069–1076 (2012).

AcknowledgementsThe authors would like to thank M. Ebelin and J. Gottowick for their excellent support with the generation of the synaptic network for figure 3 and with information retrieval from drug databases for figure 5.

Competing interests statementThe authors declare competing financial interests: see Web version for details.

FURTHER INFORMATIONAutism Europe: http://www.autismeurope.orgAutism Speaks: http://www.autismspeaks.orgDrugs@FDA: http://www.accessdata.fda.gov/scripts/cder/drugsatfda/index.cfmEU-AIMS: http://www.eu-aims.euFondation FondaMentale: http://www.fondation-fondamental.orgSimons Foundation Autism Research Initiative: http://sfari.orgA letter from Seaside to the Fragile X community regarding 209FX303: http://www.seasidetherapeutics.com/sites/default/files/STX209%20FXS%20website%20letter%2020May13_RC.PDFUniProt: http://www.uniprot.org

ALL LINKS ARE ACTIVE IN THE ONLINE PDF

R E V I E W S



http://www.autismeurope.org

http://www.autismspeaks.org

http://www.accessdata.fda.gov/scripts/cder/drugsatfda/index.cfm

http://www.accessdata.fda.gov/scripts/cder/drugsatfda/index.cfm

http://www.eu-aims.eu

http://www.fondation-fondamental.org

http://sfari.org

http://www.uniprot.org

espectro autista, descubrimiento de medicamentos

Documents

aetiology of asd

asd cases

clinical definition

current pharmacotherapy

diagnostic category

core symptoms

neuropsychiatric symptoms

heterogeneity of symptoms