Autism in the Age of COVID
Viral infection and the risk of neurodevelopmental disorders
Posted Aug 20, 2020
Few in the time of COVID-19 need to be reminded of the impact of viral illness. We recognise all too well the consequences: inflammation, fever, and (all too often) death. Not so obvious at present are the less visible consequences. I discussed in a previous blog post the potential impact of severe viral illness on the incidence of autism. The same inflammatory factors provoked by the virus that cause such a catastrophic impact on the lungs and other organs can also cross the placenta and affect fetal development.
We know this not from COVID-19, but from studies of other viral illnesses, particularly influenza. A credible study (1) coming from Denmark reported some years ago an almost 3-fold increase in the risk of a child subsequently being diagnosed with autism if the mother had a severe influenza infection during the early phase of pregnancy. Certainly, a 3-fold increase still represents a relatively small risk, from roughly 1% of live births to 3%. But COVID-19 is a more virulent virus than influenza, so the impact may be greater.
Several teams, including our own, are trying to understand more precisely how this inflammatory mechanism works. We have known for some years that pro-inflammatory cytokines are the primary culprits. These are the factors released into the bloodstream by the immune system that bring about the conspicuous features of a viral illness—fever, sickness behaviour, malaise. In a study published this week (2), we concentrated on one particular cytokine, called interferon-gamma, because it is a strong component of the body’s anti-viral response, and we thought it might help us answer a basic question: How does the brain remember that it has been exposed?
One of the conundrums surrounding the link between viral illness and autism is its longevity. We have all suffered viral illness and been bed-ridden for a few days, but the trauma passes. We get well, and as long as the sickness was not too profound—no scarring of the lungs, for example—we carry on as before. Mum might not even remember the illness. Why, therefore, might her baby be impacted to such an extent that two or three years later, a behavioural disturbance erupts?
We reasoned that this was where interferon-gamma might play a role. When immune cells like lymphocytes are stimulated with interferon-gamma, they turn on an antiviral response, the purpose of which is to remove the virally infected cells from the body. This involves activating certain genes and cellular pathways that direct the cells to attack the viral infection. Interestingly, lymphocytes that have been provoked in this way at one point respond more quickly and more decisively if they are exposed a second time. In other words, the lymphocytes remember that they have been provoked, and they are primed to go again.
The question arose, could the brain do something similar? But there was a problem with this hypothesis. Neurons – the brain’s primary signaling cells – are thought not to respond directly to interferon-gamma. The adult brain is affected by this cytokine, but the response is mediated by glial cells, the support cells of the brain. Here’s the problem: During early fetal brain development, glial cells are largely absent. They are produced later in development: neurons first, then glia. So when exposure to inflammatory cytokines seems most potent—during the first trimester—there are neurons, but few glia. This makes the hypothesis seem less tenable.
Except, there is another cell in the fetal brain to consider. During the first trimester, the brain is mostly composed of stem cells, the neural precursor cells that are generating neurons (and will ultimately also go on to produce the glia). Could they be responding to the cytokines?
What our study shows is that indeed they do. Using the cellular model of human brain development that I discussed in an earlier post, our team showed not only that these stem cells responded to interferon-gamma, but like lymphocytes, they remembered that they had been stimulated.
This had a very significant consequence. Not only did the stem cells themselves carry this memory, but the neurons that they subsequently generated also retained that memory. In other words, later during brain development, neurons remembered not that they themselves had seen a viral response, but rather that the stem cells from which they were derived had seen that response.
How did this cellular memory effect the behaviour of the neurons? They behaved more like neurons derived from individuals with autism than those from neurotypical individuals (which is what they actually were). We had previously shown (and I reported in that same earlier post) that brain stem cells derived from people with autism grew differently from cells taken from neurotypical people. They adopted a different size and shape and were less adept at making functional contacts (synapses) with other brain cells. This new study showed that exposure to the interferon-gamma pushed the neurotypical cells to adopt more autistic cell behaviour.
We feel that this result is important. It tells us that the brain itself remembers the consequences of the viral insult, rather than the memory residing in, say, the immune system alone. And it gives clear indication that the outcome is linked to the development of autistic features in the nerve cells themselves. It also helps explain how multiple different risk factors—genetic and environmental—all converge on the same set of outcomes, namely the core symptoms of autism. It thereby provides a potential explanatory mechanism of the “final common pathway” for autism: why multiple different genetic, molecular, and neural factors can give rise to it.
Identifying this effect is important, but for the information to be useful, we really needed to understand the mechanism. If we knew what the change was in these neurons that they are carrying forward into post-natal life, then the prospect exists (at least in principle) that the effect could be reversed.
Prime candidates for the cellular memory function are the structures within cells called PML nuclear bodies. When lymphocytes store the memory of an interferon-gamma stimulus, they do so by building these complex structures, which sit inside the nucleus of the cell, close to the genes that have been activated by the interferon-gamma signal, and keep those genes ‘open’. Thus, this response network is primed.
What our study showed was that the neurons were similarly primed. Just as in the lymphocytes, these PML nuclear bodies were increased in the primed neurons, and they tended to sit next to the interferon-gamma response genes. And what are the genes that these PML nuclear bodies are priming? We discovered that the key factors were a group of genes called MHC Class 1. These are well-known genes that mediate cell interactions across a range of immunological scenarios. Both the PML nuclear bodies and the MHC activation are required to produce the autism-like outcome in the cells.
Moreover, if we reversed this build-up of PML nuclear bodies, then the switch to autism-like behaviour never occurred in the cells. It seemed like they no longer remembered having been exposed to the inflammatory signal.
Whether such a reversal could be achieved therapeutically in patients is unknown and difficult to assess. Indeed, whether this pathology even exists in actual people with autism is currently unknown, although other studies have shown that PML and MCH class 1 genes are expressed at higher levels in the brains of these individuals (3). Even if it does exist, what is yet unknown is how significant these changes are in the etiology of autism, compared to other brain disturbances being discovered by other researchers, for example.
The Impact of COVID-19
Maternal immune activation caused by a viral infection certainly appears to be an important parameter in the emergence of neurodevelopmental disorders. A question now, of course, is how might this pan out in the Age of Covid? We don’t currently have the numbers that would be required to make an accurate assessment of the impact of COVID-19 on the incidence of autism, but my guess for what it’s worth, is that the effect will be negligible.
Infection rates with the COVID-19 virus vary across the globe, but to use England as an example, there are currently 2,400 new infections per day per 10,000 of population. This equates to roughly 72,000 new cases per month, or 0.13% of the population. Given their age, women of child-bearing age will be under-represented in that group, and we know that younger age groups are in any case less susceptible to the virus. One might also imagine that pregnant women will be more likely than most to self-isolate.
But taking the figure at face value: if the risk of infection is roughly 0.13% per month, and the three months of the first trimester represents the risk period for ASD, then the cummulative risk for any given pregnancy is 0.4%. There are roughly 700 thousand live births in the UK per annum, meaning 2800 women in the UK may be exposed to COVID during their first trimester. But note, the influenza data only link a severe infection requiring hospitalisation with an increased risk of autism. For COVID-19, rates of hospitalisation are falling, and official statistics do not seem to record the number of pregnant women who have been hospitalised. But if we pick an average figure of around 2%, then this reduces our ‘at risk’ population down to just 56 hospitalised pregnancies. If the autism risk for this cohort triples from 1% to 3%, then perhaps a single additional autism diagnosis in the UK would result in 2020 from the COVID-19 infection.
There are so many assumptions in this calculation. Infection rates vary considerably worldwide, and while rates in some countries are decreasing, others are still going up. All the comparisons with influenza might be seriously misleading. Certainly, more virulent viruses have a more dramatic effect. Rubella, for example, has been estimated to have had a much more substantial impact on the incidence of neurodevelopmental disorders in the past (4). More than just the hospitalised cohort might be at risk, and the impact might be considerably larger (or smaller) than three-fold. Moreover, the maternal immune activation I have described here is associated with risk not only of autism (though that is what our research project concentrated on) but a range of psychiatric disorders, including schizophrenia, anxiety, and bipolar disorder. So while this calculation implies an insignificant increase, the true outcome could be much more concerning.
So, how to advise pregnant women at this time? I would conclude only that avoiding any viral infection while pregnant would seem highly advisable. But then, there isn’t a woman alive who couldn’t work that out for herself.
Understanding the impact of COVID-19 has become the critical biomedical issue of our time. As my colleagues Grainne McAlonan, Declan Murphy, and David Edwards pointed out recently, in tackling this problem, we should not overlook an important risk group: those as yet unborn (5).
1. Atladóttir HO, Thorsen P, Østergaard L, Schendel DE, Lemcke S, Abdallah M, et al. Maternal infection requiring hospitalization during pregnancy and autism spectrum disorders. J Autism Dev Disord. 2010;40:1423–30
2. Warre-Cornish K, Perfect L, Nagy R, Duarte RRR, Reid MJ, Ravel P, et al. Interferon-gamma signaling in human iPSC–derived neurons recapitulates neurodevelopmental disorder phenotypes. Science Advances. 2020;6:eaay9506.
3. M. J. Gandal, P. Zhang, E. Hadjimichael, R. L. Walker, et al Transcriptome-wide isoform-level dysregulation in ASD, schizophrenia, and bipolar disorder. Science 362, eaat8127 (2018).
4. Hutton J. Does Rubella Cause Autism: A 2015 Reappraisal? Frontiers in Human Neuroscience. Frontiers; 2016;10:187–15.
5. McAlonan GM, Murphy DGM, Edwards AD. Multidisciplinary: research priorities for the COVID-19 pandemic. The Lancet Psychiatry. Elsevier Ltd; 2020;7:e35.