Today’s Wednesday ORF comes to us from science journalist Virginia Hughes. She’s a wonderful writer with a particular focus on (among other things) neuroscience, and when the news broke recently that she’d been tapped to head the new science and health desk at BuzzfeedNews, the science writing community was justifiably excited. Her blog, Only Human, is published weekly at National Geographic‘s “science salon” Phenomena.
Many modern neuroscience researchers devote themselves to unearthing the mechanisms that underlie a whole host of cognitive disorders. At the recent annual Society for Neuroscience conference in Washington, D.C., quite literally hundreds, if not thousands, of poster presentations were submitted under the general heading “Disorders of the Nervous System” – which is, itself, broken up into 162 subtopics. Mood disorders? Sure, did you want to learn about antidepressants, or are biomarkers more your style? Schizophrenia? If you lik, you can learn how that presents in a specific subtype of cases. Alzheimer’s disease? Stroke? Addiction? If you can name it, there’s someone studying it. And if you look closely, you’ll see that, for almost every disorder studied, there’s at least one poster session about animal models.
But why should that be? Why should researchers, with so much technology and knowledge available to them, still need to optimize the models they use in their studies? The answer is simple:
Because studying the brain is hard.
In a recent piece titled “Category Fail,” Ms. Hughes succinctly addresses some of the challenges inherent in developing an animal model of psychiatric disease – in this case, autism (though the lessons are more widely applicable). The problem comes down, in part, to the more basic issue of diagnosis. According to the 5th Edition of the Diagnostic and Statistical Manual of Mental Disorders, or DSM – essentially, it’s the handbook of psychiatric diagnoses – autism spectrum disorders are diagnosed according to two major criteria: whether or not a person exhibits deficits in social behavior and communication, and whether or not that person also exhibits repetitive behaviors (PDF fact sheet here). Hughes observes that, while a variety of animal models have been developed that recapitulate these broad criteria, the human cases are typically much more varied, both in symptoms and in their degree, and often present with other, non-cognitive conditions (for example, gastrointestinal distress in some individuals). In other words, there’s a two-fold challenge here: the challenge of making an accurate diagnosis, given a wide range of symptoms (and whether or not it’s socially useful to do so), and the challenge of developing an informative, accurate animal model by which to study, and perhaps address in some tangible way, the diagnostic criteria. Both of these are phenomenally difficult tasks, and unfortunately, they’re often in opposition to each other – for perfectly legitimate scientific reasons. Read on for more detail.
Consider Rett syndrome – a rare neurodevelopmental disorder that exclusively affects girls and often dovetails with certain autism-like behaviors. Its cause has been linked primarily to a spontaneous mutation in one gene: MECP2, which is believed to modify the way genetic material organizes into chromosomes. When the cause of a disorder can be linked to a simple gene, the job is (relatively) straightforward: introduce the mutation into a mouse model and study its effects. The mice used in such experiments typically come from very inbred lineages, which reduces the chance that intrinsic genetic variability between individuals will muddy the waters and contribute to an uncertain result. While this uniformity makes it easier to determine what specific aspects of a disorder arise as a result of any one mutation, and to narrow down possible methods by which to address these pathologies, the limited genetic background necessarily can’t reflect the same degree of variability that one sees in humans.
With autism spectrum disorders, which have genetic risk factors galore but no single gene responsible, the problem of clarity versus authenticity in animal models is compounded. Researchers have bred mice that mimic the diagnostic criteria for autism – as closely as mice can, anyway – but it’s far more challenging to recapitulate the incredible variety in human autism spectrum disorders that make them so difficult to diagnose, study, and treat. Why would one person have a risk gene variant and be diagnosed, and another not? It’s most likely due to a combination of genetic complement and environmental factors (or, oversimplified, a combination of nature and nurture), but researchers can’t reliably perform basic science while also introducing such phenomenal variability. In a good scientific experiment, as many variables as possible must be controlled in order that the truth might reveal itself – and even then, only through careful experimental design.
This is one of the many reasons why, from an outside perspective, science can appear to move slowly – because it is slow, and with some exceptions, it is careful. We study a variable at a time whenever we can, and once we’ve answered the questions surrounding it in a satisfactory way, we move on to the next variable. Slowly but surely, we come toward answers – but in the meantime, we do the best we can.