Of autistic mice and men

First published in Cracking the Enigma, January 2011

Studies looking at potential environmental and genetic causes of autism are almost always correlational. They may identify risk factors, but they can only ever show that people exposed to a particular risk factor are more likely to have autism. They don’t show whether it actually causes autism. Eating ice cream is a risk factor for getting sunburnt, but (unless you get your ice cream and your sunscreen mixed up) there’s no sense in which ice cream causes sunburn. Even if we are confident that there is a genuine causal relationship between a risk factor and autism, it’s difficult to know how the risk factor operates.

Increasingly, animal models of autism are being used to help researchers answer some of these difficult questions. An elegant illustration of the kinds of insights afforded by such an approach (as well as some of its limitations) comes from a study just out in Biological Psychiatry, looking at the effects of prenatal valproic acid (VPA for short) in mice.

VPA is a drug widely prescribed to patients with epilepsy. Various studies have reported that women who take VPA during pregnancy are at considerably higher risk of carrying an autistic child. For instance, Bromley and colleagues recently followed up 64 children born to mothers taking VPA during pregnancy and found that four children (6% of the total) had been diagnosed with autism. This is considerably higher rate than the rate of 0.9% reported for children born to mothers who didn’t have epilepsy.

One possible interpretation is that VPA affects fetal brain development, leading to autism. However, it’s difficult to rule out the possibility that whatever causes the mothers to require VPA in the first place might also be a factor that causes autism [1]. Given that many people with autism have epilepsy, this isn’t entirely fanciful. What’s more, these epidemiological studies don’t tell us how VPA causes autism.

To address some of these questions, Mike Gandal and colleagues at University of Pennsylvania investigated the effects of VPA in mice. Pregnant female mice were divided into two groups and injected with either VPA or saline (salt-water) as a control. The researchers waited until the baby mice were born and then, over the first few weeks of their lives, conducted a battery of tests to determine the effects of VPA exposure.

One of these, the Rotarod test, sounds like a challenge on some weird Japanese gameshow. The mice had to balance on a cylinder spinning faster and faster until they fell off. It’s not clear from the methods section what they fell into. But in any case, it turns out that the VPA mice were absolutely fine on this test.

The VPA mice did, however, show evidence of socialization deficits (reduced sniffing other mice) and communication impairment (reduced vocalization during mating) as well as increased repetitive behaviours (self grooming). And, if you’re prepared to draw the relevant analogies then you could argue that the mice met all three criteria for autism.

What’s more, unlike in the human research, we can be confident in the direction of causation. We know that the experimenter administered the VPA and that this was done randomly, so there’s no way that characteristics of a particular mother mouse caused her to be administered VPA. Somehow, exposure to VPA caused the unusual behaviours in her offspring.

Obviously, the relevance of these particular findings to autism hangs on the validity of the comparison between human and mouse behaviour. Can we really say, for example, that a reduction in vocalization in mice during mating is equivalent to the communication impairments that are seen in autism? If an autistic child was suddenly metamorphosed into a mouse, would he suddenly start repetitively grooming himself? Would he be averse to sniffing other mice? These are rhetorical questions. But we have to be cautious about over-interpreting the mouse behaviour.

So far, these findings are broadly consistent with previous reports of ‘autistic’ behaviour in rats that were exposed to VPA prenatally. The study gets really interesting when the researchers start to look at brain responses.

At 12 weeks old, the mice had electrodes implanted in their brains near the auditory cortex. Then, after being given a week off testing to recover, each mouse’s electrical brain responses were recorded as pure tones (pips) were played into its cage.The researchers looked at a number of indexes of brain responses, but the most interesting of these was so-called gamma phase-locking factor. Hearing each sound resulted in a short burst of oscillatory electrical activity in the gamma frequency range (40 Hertz or thereabouts). Phase-locking refers to the precise timing of the oscillations relative to the start of the auditory tone. The VPA mice had a lower phase-locking factor, meaning that their gamma oscillations were not as tightly synchronized to the tone [2].

Having established that the VPA mice had abnormal brain responses, Gandal and colleagues then set out to determine whether similar responses were found in autistic kids. Obviously, they couldn’t go implanting electrodes in kids’ brains. So instead they used a technique known as magnetoencephalography – or MEG for short.

Whenever you have an electrical current, you also get a magnetic field. MEG records the magnetic fields produced by the brain using hundreds of super-sensitive sensors all  around the outside of a helmet. By combining the information from all of the sensors placed around the head, it’s possible to estimate what a sensor placed inside the child’s brain would have measured if it had been there.

Remarkably, the autistic kids’ brain responses were very similar to those of the VPA mice. In particular, they showed a reduced gamma phase-locking factor: Like the mice, their brain oscillations were less synchronized to the sound.

These findings are intriguing. They show that administering VPA to mice in utero results in mouse behaviour that could be construed as ‘autistic’. But more than that, VPA mice have brain responses that are atypical in the same way that autistic children’s brain responses are atypical. While we should be cautious about over-interpreting the mouse behaviour, the similarities between the brain responses are extremely compelling.

What’s really intriguing, however, is this. As far as I can ascertain, none of the autistic kids were exposed to VPA. In other words, VPA appears to have an effect on brain development that is similar in some respects to the effect of other (currently unknown) causal mechanisms that lead to autism. Further studies of the VPA mouse model may, therefore, have implications for autism in general, particularly if they can show how VPA causes reduced phase-locking.

Finally, it’s worth mentioning another MEG study conducted by Donald Rojas and colleagues at the University of Colorado.  They reported similar results (i.e., reduced gamma phase-locking) in people who don’t have autism but are related to someone who does. As always, we have to be cautious about assuming causation, but this suggests that genetic factors may well be at play. An obvious next step, therefore, would be to look at auditory brain responses in genetic mouse models of autism.

Rojas’s findings also demonstrate that reduced gamma-phase-locking doesn’t necessarily lead to autism. Despite showing this ‘biomarker’, the relatives did not have autism. It may be that the mechanisms underlying the reduced phase-locking are not causally implicated in autism. It may be that the brains of ‘unaffected’ relatives are somehow able to compensate for these differences. Or it may be the case that reduced phase-locking only ‘leads’ to autism in combination with other genetic or environmental factors.

Given this, a further important question is how genetic and environmental factors interact. Remember that, in Bromley’s study, discussed earlier, 4 out of 64 children exposed to VPA developed autism. Although this is a relatively high percentage compared with the general population, it’s still important to note that the vast majority of VPA-exposed kids (94%) did not meet criteria for autism. Further studies in humans may indicate genetic or environmental factors that mediate the relationship between VPA and autism, helping to determine whether a child ultimately develops autism or not. Mouse models may then come in to play again, allowing researchers to test their causal hypotheses.


[1] Bromley et al.’s study also included 47 children born to women with epilepsy who were not taking medication. None of these children were diagnosed with autism. However, it’s not clear why these mothers were not medicated. It’s possible, for example, that their epilepsy was less severe than those on VPA, which would represent a confound.

[2] Gandal et al also investigated the effect of MPEP, a drug affecting the glutamate neurotransmitter system, on the two groups of mice. MPEP reduced the group differences in phase-locking factor, although this failed to reach statistical significance.


Gandal MJ, Edgar JC, Ehrlichman RS, Mehta M, Roberts TP, & Siegel SJ (2010). Validating γ oscillations and delayed auditory responses as translational biomarkers of autism. Biological psychiatry, 68 (12), 1100-6 PMID: 21130222