A couple of years ago while at an IMFAR meeting I happened to attend a lecture on animal models in autism. The lecturer, a distinguished neuroscientist, made a point that similarities in brain parcellation and circuitry across species justified studies of autism in animal models.  As I sat and watched the PowerPoint presentation it became quite clear to me that most of the analogies presented as facts were wrong.

Is the brain of a mouse similar to that of a human?  Besides the obvious difference in size, the brain of mice lacks complexity in terms of gyrification.  Some people even refer to the same as being “lissencephalic”, or relatively flat, especially when compared to the rich convolutional pattern of humans.  The difference is of great importance when considering the blueprint of connectivity within our white matter. Below are examples of human (left) and mouse (right) brains, though not scaled to size.

The surface expansion of the human cortex is the product of supernumerary divisions of germinal cells that surround the cavities of the brain (ventricles). These germinal divisions in turn provide for an increased number of cell modules (minicolumns) within the cortex.  In order for these minicolumns to function they need to contact a large number of adjacent modules.  These connections are made through the white matter.  Just to say that during encephalization (brain growth as compared to body size across species) the increase in white matter has surpassed that of gray matter. Here is a cross-section of cortex displaying the vertical alignments of cells known as minicolumns:

The increase in white matter during encephalization is the product of short connections linking adjacent parts of the cortex through so-called arcuate fibers, i.e., projections which form an arch between gyri.  Otherwise larger connections, which are metabolically expensive to maintain, have been reduced in numbers.  When making comparisons across species, larger brains usually have an increased number of short connections at the expense of longer ones.  This disparity in connectivity has given rise to the phenomenon of cerebral dominance.  In larger brain species, the two hemispheres do not connect their homologous areas through long -range connections.  This allows them to behave independently and acquire different functions.  The left hemisphere may therefore be responsible for language functions while the right hemisphere may be of better use for spatial reasoning.

The smaller brains of mice have a different pattern of white matter connectivity as compared to humans.  However, anatomical differences do not end here.  Mice also have a reduced number of brain parcellations as compared to humans.  In effect, an area of the brain called the dorsolateral prefrontal cortex, repeatedly involved in the pathophysiology of autism, is completely absent in rodents!  This area is responsible, in part, for executive functions.

At the microscopic level, there are many differences between the brains of mice and humans. When examining glass slides with brain slices cut at 35 microns, the cortex of mice is extremely cellular.  The density in such slices may be approximately twice that seen in humans (approximating the density seen in our own visual cortex).  There is a great deal of cellular overlap when examining the brains of mice and there is great difficulty in identifying elementary features of modularity, like minicolumns.  This point is of importance as minicolumns have been involved in the pathophysiology of autism [1, 2].

It may be that an investigator may try to overcome the cellular morass observed in the brains of mice by using specialized stains.  Immunocytochemistry may help identify individual cell types.  Using a silver impregnation method Cajal was the first investigator to suggests that smaller neurons increased in proportion in larger brain species.  These smaller neurons (interneurons) are inhibitory cells that appear to be reduced or malfunctioning in the brains of some autistic individuals.  Abnormalities in these cells appear to account for the high prevalence of seizures among this patient population. Some of the interneurons implicated in the pathophysiology of autism (creating an excitatory-inhibitory imbalance), i.e., double bouquet cells, are completely absent in rodents!

The anatomical differences between the brains of mice and men are so fundamental that they should readily disqualify the use of this species for animal models.  However, there are other reasons that minimize the importance of animal models in autism.  One of them is that present day diagnostic criteria for autism rely exclusively in screening behavioral traits.  Such a diagnosis is too crude to be used in animal models.

I often say jokingly in lectures that I have developed my own animal model of autism.  Affected mice avoid social interactions, are grumpy, suffer from sparse communication, clumsy gait, and lack exploratory behaviors.  The mouse model has social, communication, and motor abnormalities that could suggest the diagnosis of autism.  In reality they all have a leg fracture.  The example, naive and overly simplistic, calls into question a diagnosis of a human condition in animals based solely on behavioral criteria.

The significance of an animal model for autism should rely on a Turing test.  I would suggest using a blinded examiner who would perform all sorts of behavioral tests in animals and then try to diagnose the condition it was meant to imitate.  I doubt whether any single of the presently used animal models would pass such a test.  Just to say that animal models copy a limited range of behaviors but seldom a human condition.

It is our obligation to examine how many of the research approaches we are funding actually provide for meaningful information.  How do they provide a difference to our patients now, rather than enrich the imagination of researchers?  Animal research in autism has thus far been based on a house of cards.  It provides for potentially meaningless results that are easily discredited as long as our diagnostic definition lacks in terms of construct validity.  It is of importance that we recognize the limitations of this approach. Thus far animal models have raised more questions as to their significance than provided answers of clinical significance to the conditions they imitate.

[Addendum 5/18/13 A recent article indicates that most studies on animal models are underpowered and statistically flawed. For the VPA animal model of autism about 91% of them have reported invalid findings. Stanley E Lazic and Laurent Essioux: Improving basic and translational science by accounting for litter-to-litter variation in animal models. BMC Neuroscience 2013, 14:37 or the following link http://www.biomedcentral.com/1471-2202/14/37/abstract ]

References

1.Casanova MF, Buxhoeveden D, Switala A, Roy E:  Minicolumnar pathology in autism.  Neurology, 58:428-432, 2002.

2. Casanova MF, van Kooten IAJ, Switala AE, van Engeland H, Heinsen H, Steinbusch HWM, Hof PR, Trippe J, Stone J, Schmitz C. Minicolumnar abnormalities in autism.  Acta Neuropathologica 112(3):287-303, 2006.