Have you heard about the latest treatment to cure autism ... in mice?

Several people have asked me about the recently released study that, some writers promise, holds out hope for a "cure" for autism. Let's all just take a breath and slow down to walking speed here because even if these results--in mice--were to translate into something relevant to humans, getting to that point is a process of years. No sprinting here, the news release and news media stories (possibly paywalled) about it notwithstanding.

In fact, experience makes me skeptical that we're much beyond the starting gate. The study in question appeared in the journal Science and used mice to link what the authors call "nonsyndromic autism" to the autism-related syndrome Fragile X (which sometimes includes symptoms of autism). The mice they used are bred to lack a protein, neuroligin-3, responsible for building functional synapses, the connective junctions where our neurons "talk" to each other. Perhaps not surprisingly, when the researchers had the mice without neuroligin-3 start making it, the mice improved in the "autistic-like" behaviors they exhibited. More on these results in a mo', but first, a little something about the way headlines like, "Animal study offers prospect of autism treatment" make their way to our eyeballs even if they can't live up to their promises.

It can start with the paper itself. Like many researchers working on questions related to the human condition, the authors, led by senior investigator Peter Schieffele at the University of Basel, Switzerland, couldn't help but put a little line at the end of the abstract, saying that the results highlight "the possibility for reverting neuronal circuit alterations in autism after completion of development." That's essentially science speak for, "Hey, look--maybe we could take this information and use it to develop an autism treatment." There's a long line of animal studies that hold out that clinical promise in the abstract, regardless of the condition--autism, neurodegenerative diseases, cancer, autoimmune disease. Anyone who follows these things can tell you that the "fruition" part of these promises is a rare fruit, indeed.

I know the promises are frequent because I edit scientific papers for a living and often see--and delete or tone down--statements like that and because I write about science. Those of us who write about science receive news releases about studies on a daily basis. These releases offer variable quality, with some being quite good, some being really, really awful, but all with the same purpose: to capture a writer's interest and cast the institution the release represents in the warm glow of hot science. Some outlets pick up these releases and publish them wholesale, which is essentially the same thing as providing free advertisement for the institution and researchers involved. I've seen people all over social media link to these churnalism sites without noting--or knowing?--that what they're linking is the text-version of a billboard ad. These releases can be informative, but they obviously don't offer context or outside commentary. That's not their purpose.

A recent study that appeared in the Public Library of Science journal found that claims made in the abstract are the primary influence on the "spin"-related content of these releases. Indeed, these results are no surprise. After all, the scientists themselves make the claims. Peer reviewers and journal editors let those claims get by. Why not take these peer- and publisher-vetted promises of potential, of cure, and make those the centerpiece of your "sell" to journalists?

That's what seems to have happened here. Below is the opening paragraph from the news release that the authors' home university provided:
People with autism suffer from a pervasive developmental disorder of the brain that becomes evident in early childhood. Peter Scheiffele and Kaspar Vogt, Professors at the Biozentrum of the University of Basel, have identified a specific dysfunction in neuronal circuits that is caused by autism. In the journal Science, the scientists also report about their success in reversing these neuronal changes. These findings are an important step in drug development for the treatment for autism.
What the researchers really found was that taking mice that don't make neuroligin-3 and giving them neuroligin-3 mitigates the behavioral signs the mice showed because they lacked neuroligin-3. A different group of researchers developed these mice without neuroligin-3 as a "model" of autism because the protein has been implicated in some forms of autism that are heritable. Notice that I said "some forms of autism." I said that because what we call "autism" comes in many forms, probably by way of many pathways, with a huge range of manifestations in different areas. It also happens in part of be a matter of socio-cultural interpretation with an ever-changing diagnostic landscape and overlap with other diagnoses. In other words, the six letters in the word 'autism' represent volumes of variability in human expression.

Mice are great research animals if your goal is to home in on a specific gene and see what living with it, living without it, living with too much of its protein product, or environmentally jiggering with it might do to a mammal. These particular mice that lack neuroligin-3 show a suite of deficits, including reduced vocalization (mice make sounds our ears can't hear) and a "lack of social novelty preference." Turns out, that lack of social novelty preference might trace to the fact that these mice have an "olfactory deficiency." In other words, they don't smell too well. These features may add up to a picture of what happens to mice when they lack neuroligin-3, but they don't necessarily overlap with autistic humans, whose sense of smell can be famously refined (although apparently, a "subgroup" has an olfactory deficit) and some of whom (ahem, oldest son of mine) also are known for a whole lot of vocalizing. Mice use vocalizing as communication, just like people--including autistic--people do.

So while it's intriguing to look at these mice lacking neuroligin-3 because some autistic people have gene variants encoding the protein--which doesn't mean they inevitably lack it--a mouse missing neuroligin-3 is in no way--not shape, not form, not behaviorally--much like a person with autism or all persons with autism.

What the authors of this study really did that was worthy of publication in Science--and it is interesting--was show that the brain after birth is still open to targeted interventions that alter function. Their other interesting finding is that some of the neuronal construct atypicalities observed in mouse models of Fragile X, which the authors call "syndromic" autism models, are like those in mice lacking neuroligin-3. [NB: Their thinking is that mice lacking neuroligin-3 represent a "nonsyndromic" autism mouse model, but that mutant mice intended to model Fragile X are "syndromic." I don't quite understand that distinction, as lacking neuroligin-3 does appear to induce a syndrome. In addition, autism researchers with whom I've spoken consider it a distinct possibility that some of the suites of signs and symptoms we call autism are as-yet-unidentified syndromes.] 

This discovery of an overlap of synaptic problems in mice knocked out or mutant for two different genes is certainly of interest, supporting the idea that many genetic pathways might lead to the Rome we call 'autism'. It could alternatively have revealed another step in the pathway to the syndrome we call "Fragile X."

Tweaking the young mice to start expressing neuroligin-3 resulted in improvements in their behavioral signs, including some hints at motor coordination improvement. That might not seem too surprising, but the real take-home is what it shows about the malleability of the mouse brain. The fact that expressing this protein seems to have reversed some signs suggests that the brains of the mice--at least while they're still young--remain open to change and remodeling to some extent. The investigators say in their Discussion that "structural neurodevelopmental phenotypes can be rescued by intervention after the completion of development." I think that they're referencing  completion of embryonic/fetal development there, even as their results support the increasingly evident idea that development doesn't end at some specific timepoint in life. It's an ongoing process throughout life, including--in particular?--development of the brain.

The authors also note in their Discussion that their results "provide insight into the synaptic pathophysiology of a model of nonsyndromic autism." That's a reasonable conclusion, for mice, based on their definition of "nonsyndromic." But not necessarily for humans.

But the human context persists, from abstract to news release to the Wall Street Journal. In its lede, the Journal article on this paper, echoed the news release, saying:
Roche Holding AG, a Swiss drug maker, and the University of Basel's Biozentrum said Friday the study identified a way to reverse a dysfunction in the brain's wiring typically caused by the disorder, which stumps intellectual development and can cause aggressive and anti-social behavior, and becomes evident in early childhood.
From what I've seen, describing autism in news stories can be a challenge, as this lede illustrates. "Stumped intellectual development" is not a sign of autism. Intellectual disability, as measured using current testing methods, can accompany autism, but a majority of autistic people identified by the Centers for Disease Control and Prevention in its most recent prevalence roundup were not intellectually disabled. Autism can be associated in some (but by no means all) cases with "aggressive" behaviors--typically self-injurious behaviors--however, it's not identified with "anti-social behaviors" but rather with social communication struggles. Not the same thing. How to describe autism in a few words? "A (neurobiological) condition of varying intensity (and likely causes) that involves difficulty with social communication." Parentheticals optional.

This study does not show that the 'disorder causes the wiring', rather than the likelier 'wiring's causing the disorder', but you can clearly see that the writer picked up that phrasing from the lede in the news releaseThis echo and the extrapolated human context form the final link in the chain of influence from abstract to news release to writer to You. To its credit, the Wall Street Journal article contains a number of caveats about the limits of an animal study, with a paraphrase from study funder Roche that
It is too early to say how a possible treatment would work or what patient groups it would be given to or for how long, Roche said, stressing this is still basic research in an animal model.
Indeed. Drug development takes, according to California Biomedical Research Organization [PDF], an average of 12 years, although Roche appears anxious to speed that process up for autism therapeutics. Few rarely make it to clinical trials, and even when they do, they are very likely to fail. For example, more than 200 drug candidates for Alzheimer's have failed. The mouse model in the autism study, like many mouse models, involves mice missing a single protein. As I've noted, autism is a human condition with cultural and social aspects of interpretation and one that probably arises by many different pathways, with multiple genetic inputs and likely environmental interactions. The news release itself states that variants of more than 300 genes have been linked to autism--of which neuroligin-3 is only one.

The university's press office release concludes with a subhead, "Vision for the future: Medication for autism" and the following paragraph:
Autism currently cannot be cured. At present, only the symptoms of the disorder can be alleviated through behavioral therapy and other treatment. A new approach to its treatment, however, has been uncovered through the results of this study. In one of the European Union supported projects, EU-AIMS, the research groups from the Biozentrum are working in collaboration with Roche and other partners in industry on applying glutamate receptor antagonists for the treatment of autism and hope, that in the future, this disorder can be treated successfully in both children and adults.

The results of this study do not show a "new approach to treatment." They suggest a target pathway linked to an extremely complex neurobiological condition, one that provides a rationale for more analysis. 

I confess to being nitpicky and highly critical of just about everything I read about autism and in articles about scientific research in general. I have become particularly jaded about premature hints at human applicability. Also, with an acute awareness of the chain of (mis)information that travels from news release to churnalism site or writer to you, I worry--I lie awake at night, people--about how much our perception of what science can do gets skewed, to the detriment of everyone. Perhaps you, gentle reader, are less critical, less nitpicky. Do you think that extrapolations like this from mouse studies are defensible, or do they overpromise ... again?


  1. Hi Emily -

    Nice write up that I am largely in agreement with.

    I wonder though, when I was starting out on journey Internet-Autism and started trying to read articles, one of the things that I took home was the idea that the upper tier journals had a level of respectability that was warranted. Here, this paper is in Science, which, if I understand correctly, is a big deal. But is altering brain function / behavior after birth like this really that big a deal?

    Are these findings more transcendent than Patterson's studies finding bone marrow transplants had similar behavioral modifications in the treatment group? [Very similar to a paper in Nature with a similar methodology, but on a rett model, Wild-type microglia arrest pathology in a mouse model of Rett syndrome]

    Is this really more impressive than another knockout then treat model like Neuronal and glia abnormalities in Tsc1-deficient forebrain and partial rescue by rapamycin? Or an oxtocin based knockout mouser, Pharmacologic rescue of impaired cognitive flexibility, social deficits, increased aggression, and seizure susceptibility in oxytocin receptor null mice: a neurobehavioral model of autism [at least oxtocin has some human trials, i.e., Effects of intranasal oxytocin on social anxiety in males with fragile X syndrome!]

    Technically what they did in the Science paper was impressive; but I cannot help but wonder, is what we are witnessing here a function of getting published in Science making it more likely that a press conference happens?

    Your notes of caution about the costs, timeframes, and incredible failure rates in the drug discovery realm are very important with the repeated and expensive failures in the Alzheimer's realm are a good example. My take away from this effort is coloring a lot of my thought processes on autism lately. I've been reading the very excellent In The Pipeline blog for a few years now, and my take home message on our failures to develop drugs is that despite all of the effort, and it has been herculean, we still are largely unclear on a lot of the underlying biology of many disorders; especially in the CNS.

    Perhaps you, gentle reader, are less critical, less nitpicky.

    What I think is largely missing from the autism discussion realm is the application of this same level of critical analysis towards the incidence question.

    When the UK adult "study" came out, the Internet at Large was bursting with stories about how a 1% incidence of autism was found in adults. Nobody seemed concerned, at all, that the total number of people diagnosed was 19. Nobody seemed worried about the fact that the male to female ratio was an astronomically high 9:1! (They "found" two females). I've seen you write about underdiagnosis in females, but the England study was designed to overcome that by mass screening, and targeted evaluations, and they still couldn't find any females! Does this give you any insight into the methodologies of the study; or perhaps, of autism incidence?

    Then, when the Korea study came out, showing a 2.5% "incidence", nobody batted an eye, again! If the UK study was any good, and the Korea study was good, we have evidence that the incidence of Korean children with autism is two and a half times what English adults is, a sort of 'epidemic lite'. Where was the skepticism towards these studies?

    Whatever the problems are with the Neuroligin studies, at least those findings have a basis in the microscope; our studies on incidence are all from the telescope, and they are all over the place, but the over riding meme of a stable rate of autism seems to get a free pass on a critical analysis. Considering the implications of being wrong, I think we should apply at least the same skeptical eye towards incidence that we do other types of studies

    - pD

  2. Hi, pD. Thanks for posting. What I wrote about incidence goes back much further than the studies you cite, to Lorna Wing, in fact. As you note, what we see under a microscope is much more constricted and comparatively easily interpreted. The changing diagnostic landscape alone has made pinning incidence difficult, but the careful studies keep coming up with about the same values. The Korean study was of interest because its work was in a *general* rather than a *clinical* population, suggesting that were there other equally large-sized studies done, we'd round up even more people walking around *right now* who'd qualify for a spectrum diagnosis under current criteria with current scales. I read all of these papers and various evaluations of them very carefully and consulted with several autism experts before I wrote the piece for Discover. SFARI featured two very useful analyses, and Dorothy Bishop also has written useful posts about this; all linked in that piece, if I remember correctly.

  3. Hi Emily -

    I wasn't really referring to your Discover piece per se; in fact, I hadn't read it until today.

    The Korean study was of interest because its work was in a *general* rather...

    But when they tried this in England, also general population study, they failed to find the same values! The methodology wasn't that different than Korea, give everyone a screening tool, then evaluate the people who scored high with full, current diagnostics. In your Discover piece you said:

    One study in South Korea found a significantly higher rate of autism, but it used a different methodology and different study population.

    What's so different, methodology wise between Korea and England? Seriously. One used an ASSQ, the other a modifed ASQ as a pre-screener, and then standard diagnostics were used on people who scored high. I thought the whole idea of a general population study was to get away from the confounding problems of people not seeking diagnosis for one reason or another.

    Also from your Discover piece:

    In fact, the part of the Korean study that was most comparable to other studies found an autism rate of .8%—about the same as in other countries.

    Try finding a headline / press review from that time that included this nugget. The same lack of nuance and inattention to details that bothered you so much about the neuroglin was widespread when the Korea study came out.

    I read all of these papers and various evaluations of them very carefully and consulted with several autism experts before I wrote the piece for Discover.

    So you'd read the England study carefully, and you knew that they found a 9:1 male/female ratio? That didn't bug you when you read every other prevalence study that found a 4:1 ratio, max? You didn't think that this might inform your conclusion that they had "found as many adults as there were children walking around with autism, suggesting stable rates across generations". Is there another study, anywhere, that approaches a 9:1 male to female ratio?

    A different population study, Brief report: prevalence of pervasive developmental disorder in Brazil: a pilot study [also Fombonne] followed the same methodology as Korea and the England UK study, widescale screening and targeted evaluations, and came up with a .3% prevalence, a third of UK adults, and almost ten times less than Korean children. (!) And again, here, an ASQ was used as a screening device, and then full diagnostics used on high probability targets; except our numbers are crazy low. It would seem that in Brazil, there aren't that many children walking around right now who would qualify for a spectrum diagnosis.

    If three general population screening studies can give such wildly different results, why shouldn't this color our interpretation of all of our prevalence studies? These were tests designed to overcome the traditional problems of evaluating people who had gotten diagnosis through varied means, and they still come up with crazy different numbers. Are we allowed to pick the ones we like and the ones we don't like?

    If someone published a study on Neuroglin that found a three fold increase or decrease didn't affect manifestation in a mouse model, we would be seriously reconsidering whether or not it was a meaningful participant in autism. When we see the same thing in incidence studies, it can all get lost in the wash, someone says, 'Lorna Wing', and any true increase is chalked up as, 'likely quite small'; whatever that means. (1%? 10%? 25%?)

    I think that is a terrible, horrible way to answer a question as profoundly important as 'is this generation of children, different?' You and I probably have different ideas on that.

    - pD

    1. The Discover piece should contain links to very specific commentary and analyses about these studies that answer your questions, particularly the SFARI links.

      I say "Lorna Wing" because she estimated a prevalence a couple of decades ago that matches the currently identified prevalence, and she did so well before anyone was talking about "epidemics." That strikes me as relevant. If you think that an estimate from that long ago that reflects what we estimate today is a "terrible, horrible way" to answer a question about whether or not this generation of children--and which generation, exactly?--is different, then, yes, we have different ideas about how to answer that question. Are kids these days different from kids in the early 20th century? Probably, but kids in the early 20th century were likely quite different from kids in the Revolutionary period. I don't know how profound that really is.

      Someone ought to do the study you describe in neuroligin because knocking it out likely doesn't reflect the human condition.


Post a Comment