Researchers can use biomarkers as measures to test the efficacy of therapies in early-stage trials. From 22 to 24 July, SFARI held an informal meeting of minds at Stony Brook University in New York to discuss early-efficacy biomarkers for autism. The goal was to provoke discussion about potential biomarkers and to build and strengthen collaborations among attendees.
The U.S. National Institutes of Health defines biomarkers as “biological characteristics that can be objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes, or pharmacological responses to therapeutic intervention.”
The workshop participants unanimously agreed that biomarker development should be a priority for autism research. They also called for an improved “second wave” of biomarker research, which would set more realistic expectations for biomarkers than there have been in the past.
For one, they said, biomarkers should not replace clinical diagnosis. Biomarkers can, however, identify subgroups of individuals within the autism spectrum who could be treated with specific interventions.
Despite the extreme heterogeneity of autism, there is a highly conserved mechanism of social behavior that consistently goes awry in individuals with the disorder. Biomarkers can help scientists devise a therapeutic approach that can be applied across the spectrum.
It is also important to acknowledge that biomarker measures may not be fixed across an individual’s lifetime, especially in a developmental disorder such as autism, the attendees concluded. When interpreting functional significance and predicting clinical outcomes, “a single measure at a single time point can be very misleading,” said Charles Nelson, professor of pediatrics and neuroscience at Harvard Medical School. Even the direction of a measure compared with controls — such as whether brain activity is elevated or suppressed — can vary.
For example, electroencephalography (EEG) measurements show differences in brain activity at 18 months of age between the siblings of children with autism, who are at risk of developing the disorder, and typical controls, according to Nelson. At 24 months, however, measurements from the two groups become more similar. If the trend were to continue, measurements from the high-risk siblings could show a completely different pattern compared with controls when they reach 36 months, Nelson noted.
Biomarker development is further complicated by the fact that, because of clinical testing regulations, researchers generally need to validate biomarkers in adults. However, these markers are most useful for identifying young children who are candidates for early interventions.
Researchers must replicate biomarkers and use them to predict outcome in independent samples, the attendees pointed out. There are numerous reports of potential biomarkers published in the literature, yet few make it through the rigorous scientific process of verification, noted Walter Koroshetz, deputy director of the National Institute of Neurological Disorders and Stroke in Bethesda, Maryland.
The participants also advocated for a collaborative approach to biomarker research, because biomarkers are most powerful when they combine measures from multiple disciplines.
The participants considered a wide variety of biomarkers: genetic, biochemical, cellular, anatomical, physiological and clinical. They also addressed whether these should be used alone or combined into panels.
Biomarkers can be either proximal or distal, meaning they are either close to the clinical manifestation of the condition or more representative of the underlying mechanism. Several of the researchers showed examples of proximal, or clinical, biomarkers.
Because autism is associated with defects in attention to social cues, people with the disorder may miss out on opportunities for social learning. These deficiencies are associated with underlying neural processes, specifically electrophysiological responses to social stimuli, according to Geraldine Dawson, chief science officer at Autism Speaks. Individuals with autism also show atypical responses in the reward center of the brain, often preferring monetary to social rewards.
Ami Klin, director of Emory University’s Marcus Autism Center in Atlanta, uses eye-tracking technology to show that individuals with autism look at different aspects of a social scenario than typical individuals do. Like eye tracking, the best biomarkers should capture what happens in naturalistic social environments, because these are the situations in which individuals with autism have the most trouble, Klin said.
The Q-sensor is a wireless wristband developed by Matthew Goodwin and his colleagues at the Massachusetts Institute of Technology Media Lab. The sensor allows researchers to track, in real time, body motion and activity of the sympathetic nervous system, which regulates sweating, heart rate and pupil dilation. This type of technology may allow researchers to develop biomarkers of stress, arousal and anxiety in the everyday lives of individuals with autism. Goodwin also showed examples of video technology and computer analysis techniques used to capture behaviors in real-life settings.
These measures are particularly suited to assessing and quantifying repetitive behaviors and atypical gait in individuals with autism. Repetitive behaviors are proposed to be one of only two core domains of autism in the forthcoming edition of the Diagnostic and Statistical Manual of Mental Disorders, DSM-5, set to be published in 2013.
Children with autism also have atypical pupillary responses, such as delays in pupil constriction in response to light stimulation, noted Judith Miles, professor of pediatrics and pathology at the University of Missouri’s Thompson Center for Autism and Developmental Disorders. Miles has done significant work on dysmorphology, or atypical facial and physical features, another potential biomarker for autism. For example, using three-dimensional facial imaging, Miles has shown that children with autism have different patterns of facial features than controls do.
Brain imaging measures could represent biomarkers that lie in the middle of the distal-proximal axis, suggested Timothy Roberts, professor of radiology at the Children’s Hospital of Philadelphia. Because of this, imaging biomarkers are ideal for translation into the clinic and as endpoints for drug trials, he said.
For example, brain responses to changes in sound frequency, as measured by magnetoencephalography, or MEG, are delayed in children with autism compared with controls and children who have language delay.
Functional magnetic resonance imaging has also identified patterns of brain activity in different brain regions in children with autism. However, these measures are sometimes shared by unaffected siblings of children with autism, and may represent endophenotypes that indicate underlying genetic risk factors.
Unaffected siblings may also have compensatory biomarkers that protect them from developing the disorder. However, it is very difficult to separate out which biomarkers represent the underlying biology that causes autism and which arise as a consequence of the disorder, cautioned Kevin Pelphrey, associate professor of psychology at the Yale Child Study Center.
The researchers also discussed biomarkers more distal to the manifestation of autism, including genetic and proteomic measures.
Biomarkers based on genetic differences can help distinguish individuals with autism based on the underlying mechanism, or etiology, of their autism, said Daniel Geschwind, professor of neurology at the University of California, Los Angeles. However, gene expression differences between the blood cells of individuals with autism and their siblings represent a “weak signal,” Geschwind cautioned.
Geschwind has analyzed patterns of gene expression in the blood cells of 200 individuals with autism and 200 unaffected siblings and found distinct differences. But these distinct patterns cannot be reproduced in another 200 individuals with autism, Geschwind reported. People with autism do have reproducibly different gene expression patterns from unrelated controls, but this difference is not as meaningful as showing differences with unaffected siblings, because “sibs carry the same genetic risk, but don’t have a diagnosis,” Geschwind noted.
Geschwind has also seen differences in the patterns of gene expression in postmortem brains of individuals with autism compared with controls.
Researchers can use various mouse models of autism to link a particular genetic defect with mouse behaviors relevant to autism symptoms and with biomarkers such as gene expression profiles, heart rate or blood pressure, says Jacqueline Crawley chief of the laboratory of behavioral neuroscience at the National Institute of Mental Health in Bethesda, Maryland. Rigorously replicated behavioral profiles in mouse models are also well-suited for testing potential autism therapies.
Some mutant mice have significant defects in the levels of neurons that dampen signals in the brain and express the neurotransmitter gamma aminobutyric-acid or GABA, reported Hannah Monyer, professor of clinical neurobiology at the University of Heidelberg in Germany. Monyer has shown that GABA neurons can make long-range connections in the brain, implicating them in the connectivity defects seen in individuals with autism.
Attendees who are experts outside of autism research presented approaches to biomarker identification that are not yet common practice in the autism field. For example, there are nearly 6,000 peer-reviewed articles listed in the publications database PubMed that cross-reference cancer and proteomics, but only 9 that do so for autism and proteomics, noted Howard Schulman, executive advisor at Caprion Proteomics in Menlo Park, California.
Proteomics is a powerful way to identify changes in gene expression in the blood of individuals with a disorder, Schulman said. Researchers can screen thousands of proteins at once and resolve differences among samples using mass spectrometry. Because mass spectrometry can validate these proteomic markers in humans and animal models, researchers can screen for hundreds of different potential biomarkers at once, he said. Different patterns of protein expression may correlate with the level of symptoms and identify groups of individuals with specific traits, Schulman added.
For example, Schulman is collaborating with researchers at the University of California, Davis MIND Institute to study 200 children with autism, aged 4 to 7 years, and to match their proteomic signatures with brain imaging data.
Another proteomics approach is to use a peptoid library, which detects antibodies circulating in blood, reported Dwight German, professor of psychiatry at the University of Texas Southwestern Medical Center in Dallas. This approach could be more powerful than proteomic screens because it detects the body’s immune response to protein levels, which may be amplified compared with the protein levels themselves.
Ellen Li, professor of microbiology at Stony Brook University, discussed her research on the microbiome, the full range of symbiotic bacteria and other flora that live in the gut. Li follows microbiome changes across the lifetime of individuals with inflammatory gut disease. It is very important to correlate these data with changes in phenotype, Li noted.
The workshop concluded with a discussion of next steps, including plans to continue the discussion and to pursue collaborations online.
The attendees set priorities for developing autism biomarkers, such as using computational and statistical approaches to aggregate data and looking for dynamic trends in biomarkers over the course of development.
They also agreed that the field should identify the most promising biomarkers and ensure that data collection and research into these areas are standardized, allowing researchers to share data across studies.