Recently on this blog, I’ve written about the idea of replicability. Now onto something slightly different—generalizability. As behavioral neuroscientists, we hope that our findings generalize beyond the exact conditions of our experiments, and in many cases, beyond the species we choose to study. This is particularly true in labs that study models of psychiatric disease. Recent high profile co-publications on compulsive behavior and on depression, however, call this idea into question. Here I'll discuss these two pairs-of-papers, with an eye toward their implications for generalizability.
This last week, Science magazine published two papers on mouse models of compulsive behavior (1, 2). The first paper utilized Sapap3-KO mice (Sapap3 is a postsynaptic scaffolding protein), which show excessive grooming behavior. The authors provide evidence that optogenetically stimulating the pathway between the orbitofrontal cortex (OFC) and the striatum decreases activity in the striatum by way of feed-forward inhibition. They show that this stimulation reduces excessive grooming and argue that the drop in striatal activity drives this behavioral rescue.
The second paper utilizes wild-type mice. These authors, like the other authors, optogenetically stimulate the OFC-to-striatum pathway. They argue, however, that across multiple days, OFC-to-striatum stimulation leads to increased striatal activity, and they provide evidence that this increased activity induces pathological grooming behavior in previously normal mice.
Let’s assume both findings are true. It’s really the more interesting conjecture.
(full disclosure and potential conflict-of-interest: my advisor is an author on the second of these two papers).
Had the Sapap3-KO paper been published alone, it would have been easy to conclude that “increasing top-down inhibition” from OFC to striatum could have a therapeutic effect on compulsive behavior. The second paper adds an interesting twist. Across multiple days of stimulation, it appears that the striatum undergoes a bit of plasticity. After stimulation stops, striatal activity is even higher than it was before the experiment began, leading to a pathological grooming behavior in what-were-normal mice. This finding is, of course, beyond the scope of the first paper (which only stimulated animals on one day), but has potentially important implications about 1) the degree to which we can generalize findings from the first paper and 2) the use of such stimulation as treatment.
An additional concern, of course, lies in whether mouse-grooming provides a functional homology to human compulsion and not just an interesting analogy with some (perhaps limited) face validity. More on homology versus analogy in a few weeks…
Rodent Models of Depression
In December 2012, Nature co-published two papers on rodent models of depression (3,4). One group utilized a chronic mild stress model of depression. Their primary finding was that stimulating dopaminergic neurons in the VTA relieved the ‘depressive’ symptoms brought on by chronic mild stress. Inhibiting these same neurons, likewise, increased ‘depressive’ behavior in otherwise normal mice. Depression, in this case, was measured as sucrose preference and as struggling in a tail-suspension test.
The second paper utilized a social-defeat model of depression. The primary finding in the second paper was that stimulating dopaminergic neurons in the VTA exacerbated the ‘depressive’ symptoms brought on by social defeat. Likewise, inhibiting these same cells rescued the ‘depressive’ behavior. In this case, depressive behavior was measured by sucrose preference and by social interaction.
Again—let’s assume all findings hold.
(full disclosure and potential conflict-of-interest: my advisor is an author on both of these papers).
If we take the results of both of these papers at face value—that is, that VTA stimulation reverses the effects of chronic mild stress but exacerbates the effects of social defeat, then these papers can sit relatively comfortably on their adjacent pages. The two forms of stress are, in fact, quite different. In one, a mouse is exposed to 8-10 weeks of unpredictable stressors, some lasting for days, sometimes multiple within a day. For clarity, these stressors include: strobe light, wet bedding, food/water deprivation, cage tilt, cage shaking, and continuous loud white noise, among others. In the other, the subject mouse is placed into the home cage of a larger, more aggressive mouse who is allowed to beat our subject up for five-to-ten minutes a day over ten days. It is not difficult to believe that these two forms of stress work through different neural circuits to manifest changes in behavior.
If we are to believe, however, that both chronic mild stress and social defeat are “models of depression”, then these findings become difficult to reconcile. Such conflicting results clearly point to the need for better metrics to assess and validate animal models of psychiatric disease, and furthermore, clearly call for pressure from the scientific community (including publishers) to use models to assess only the outcomes they are validated to test. I’d love to hear your thoughts in the comments, and join me in a few weeks for more discussion on animal models.