Baby teeth that arrive at the “Tooth Fairy” project’s research laboratory at the University of São Paulo (USP) are helping scientists better understand changes that can occur in the brain in some forms of autism. Autism is a group of disturbances of neurological origin that manifest in childhood and may impair cognitive, communication, and social interaction skills to varying degrees, as well as motor skills. Stemming from as yet mostly unknown causes, autism—or autism spectrum disorder, as experts prefer—includes such varied conditions as classic autism, which is marked by severe language and social interaction difficulties, and Asperger’s syndrome, in which intelligence is normal to above average and language acquisition occurs without problems, but repetitive gestures and lack of control with delicate movements are common. These disorders affect about 1% of children in England and the United States—there are no detailed studies of their frequency in Brazil—and are four times more common in boys than in girls.
Using baby teeth donated by children with and without autism, the groups led by Brazilian neuroscientists Patricia Beltrão Braga, from USP, and Alysson R. Muotri of the University of California, San Diego (UCSD), in the United States, have confirmed that an inflammation in brain cells called astrocytes may be associated with the development of a severe form of the disorder. More importantly: at least in the laboratory, the control of inflammation in the astrocytes reverses the changes it causes in neurons, the cells responsible for transmitting and storing information in the brain, and which are more immature in this form of autism.
Since it would be risky and unethical to extract brain cells from children, the researchers drew on an ingenious alternative to recreate them in the lab. They recovered cells from inside teeth that naturally fall out in childhood and used chemical compounds to regress them to a more versatile state, from which they could generate other cell types. They then added chemical compounds that stimulated them to transform into brain cells.
During her doctoral studies, supervised by Braga and Muotri, biologist Fabiele Russo obtained neurons and astrocytes from the baby teeth of three Brazilian boys. They had been treated by one of the country’s leading specialists in autism, child psychiatrist Marcos Mercadante, from the Federal University of São Paulo (UNIFESP), who died in 2011. All three boys exhibited a serious form of the disorder. They had difficulty learning tasks and information and relating to others, spoke only a few words, and performed repetitive movements.
Neurons generated from the tooth cells of these boys were less mature and less complex. They had fewer branches and made fewer connections with other neurons than those obtained from the teeth of children without autism. They were also less active (emitted fewer electrical impulses) than those of the boys in the control group. “Changes in the shape and functioning of neurons can impair the development and activity of brain circuits,” says Russo.
The influence of astrocytes
Results of the experiments, published in October 2017 in the journal Biological Psychiatry, indicate an inflammation appears to be behind the changes in the shape and functioning of these neurons. The evidence for inflammation comes through analysis of astrocytes. These star-shaped cells do much more than fill the space between neurons. It is now known that astrocytes control the formation of neuronal branches and regulate the concentration of compounds such as neurotransmitters, which are responsible for chemical communication between brain cells.
Russo observed that astrocytes generated from the cells of children with autism produced a larger amount of the inflammatory molecule interleukin 6 (IL-6). Earlier analyses of postmortem brain tissue had already linked high levels of IL-6 to autism, but didn’t confirm whether high concentrations of this molecule were the cause or consequence of the problem. The USP and UCSD researchers found evidence that the inflammation precedes autism, in a sequence of experiments in which neurons were cultured with astrocytes with and without signs of inflammation.
Healthy neurons that developed in the same container as inflamed astrocytes had fewer branches and connections and became less active. The opposite occurred when neurons from the teeth of children with autism were placed for a few days with astrocytes from boys without the disorder: in the company of healthy astrocytes, the neurons produced more branches and connections with other cells (see next page). The same was observed when an antibody that blocks the action of IL-6 was added to the culture of neurons from people with autism.
“These changes make us think that some forms of autism are caused by a problem in the astrocytes, and not in the neurons,” says Braga, coordinator of the Brazilian branch of the international “Tooth Fairy” project. It takes its name in reference to the figure from the popular imagination of various countries who rewards children for recently fallen baby teeth left under their pillows. If other studies confirm the influence of this neuroinflammation on certain forms of autism, it may be possible in the future to develop pharmacological treatment for the disorder.
Current therapies are based on psychotherapeutic treatment to stimulate interaction with other people and enable a more independent life, and include the use of medication to combat secondary problems such as difficulty sleeping or concentrating. “If someday it’s possible to reverse these cellular changes in the brain of a developing child, it may become feasible to mitigate the manifestations of autism,” the neuroscientist says.
It could be a long road. “Although it works well in vitro, treatment that inhibits IL-6 in humans should be investigated carefully, since there are reports that it may aggravate the inflammation of neurons,” warns Muotri, father of a boy with autism and cofounder, with Braga and collaborators, of Tismoo, a Brazilian biotechnology company with a branch in the United States. Tismoo performs genomic analyses in search of causes and possible therapies for autism and other genetic neurological disorders. In a recent clinical trial that tested a compound for blocking the action of IL-6 in people with a form of degeneration that affects the optic nerve and spinal cord, one participant developed a more serious condition, researchers reported in the journal Multiple Sclerosis in 2016.
Years earlier, Muotri had already confirmed that neurons generated from the reprogramming of skin cells from children with Rett syndrome—a disorder similar to autism that affects only girls—were also immature and created fewer connections. In the laboratory, this effect was reversed by applying the antibiotic gentamicin or the insulin-like growth factor 1 (IGF-1) to the cells. The problem, confirmed later, is that the former is relatively toxic and the latter doesn’t reach the brain in adequate quantities (see Pesquisa FAPESP, issue No. 184).
The diagnosis challenge
Although scientists are far from understanding the causes of autism (changes in 800 genes have already been associated with the disorder), cellular reprogramming and genetic analysis may help in diagnosis, which is not always easy and can take time. The identification of autism is done by a well-trained psychiatrist or neurologist who examines the child and evaluates his or her life history for signs of delays in development and the ability to interact socially. Symptoms may appear early, within the first few months of life, and early diagnosis is important as well, because the brain develops more rapidly in the first few years, the period when the child begins to acquire language and interact with adults and other children.
In general, however, the diagnosis is usually delayed. Some experts estimate that confirmed diagnoses of autism usually occur around age three. In Brazil, identification may take even longer. In an interview with Pesquisa FAPESP in 2011, UNIFESP’s Mercadante estimated that diagnosis only occurred at around the age of five or six, when impairments in sociability are more difficult to mitigate.
In order to identify the reasons for this delay, researchers from the Social Cognition Outpatient Clinic at UNIFESP, and Mackenzie Presbyterian University, conducted detailed interviews with 19 mothers of children diagnosed with autism in the city of São Paulo. Most of them (68%) noticed that there was something abnormal in the child’s development before the age of two—the child didn’t respond when called by name, didn’t make eye contact, or wouldn’t speak. After noticing the problem, almost all the women (79%) sought medical help within the following three months. However, the diagnosis was only confirmed, on average, three years after their initial suspicions (in some cases it took eight years). These are partial results from a larger study of 250 families of people with autism, presented in an article published in October 2017 in the Revista Brasileira de Psiquiatria (Brazilian journal of psychiatry).
Thus, one of the reasons for the delay in diagnosis would be doctors who are unprepared or misinformed. “Most of these mothers, 84%, first mentioned their concerns about their child’s atypical behavior to their pediatrician, but heard things like ‘children shouldn’t necessarily be compared to each other,’ ‘boys develop more slowly than girls,’ and ‘boys are more agitated,'” says Sabrina Ribeiro, a psychologist at UNIFESP’s Social Cognition Outpatient Clinic and the lead author of the study. “More than half of them, however, felt discouraged to express their concerns again,” says the researcher, who years earlier had conducted the largest survey to date of the frequency of autism among Brazilian children. Conducted in Atibaia, a small city in the state of São Paulo, under the guidance of Mercadante, the study revealed that 4 of the 1,470 children surveyed (0.3% of the total) had autism. “Identifying autism as soon as possible is essential for early intervention.”
In Porto Alegre, a team from the Pontifical Catholic University of Rio Grande do Sul (PUC-RS), led by neuroscientist Augusto Buchweitz and computer scientist Felipe Meneguzzi, has been using a computational tool called machine learning to create a sequence of autonomous procedures (algorithms), which attempts to distinguish individuals with autism from those without the disorder, through analysis of functioning brain images. The group developed an algorithm classified as deep learning, which learns to make this differentiation in an autonomous way, without the need of suggestions from the researchers. The strategy achieves a high level of accuracy under specific conditions but doesn’t seek to replace clinical diagnosis.
The PUC-RS team used their algorithm to evaluate about 3,500 brain images stored in the Autism Imaging Data Exchange database, from 1,035 individuals (approximately half with and half without autism) from different regions of the United States. As had been observed in previous studies with lower numbers of participants, there is a dyssynchrony between the functioning of anterior and posterior regions of the brain in autism—in individuals without autism, these areas are more synchronized. According to the results of the study, published online in the August issue of the journal NeuroImage: Clinical, the PUC-RS group’s algorithm correctly identified 70% of the cases with autism.
“This is the highest accuracy obtained yet in the analysis of a large and heterogeneous sample,” says Buchweitz, coordinator of functional magnetic resonance imaging research at the Brain Institute at PUC-RS. Previous studies have already achieved higher rates, close to 100%, but were conducted with only a few dozen participants and used algorithms that require the authors’ supervision in order to learn how to classify images. “Classification accuracy drops significantly in larger population samples and if data is from different sites,” the researchers wrote in the article.
The encouraging result, however, doesn’t mean that computational analysis would become available for real-life diagnoses anytime soon. “Replacing clinical diagnosis, which works, isn’t our objective,” says Buchweitz. One reason for this is that the predictive power of the algorithm is only high in cases where half of the participants have autism and half do not, as in the database used in the initial evaluation. This is far from what occurs naturally in the population, in which the frequency of the disorder is around 1%. “Under these conditions, the algorithm’s actual hit rate would drop to about 4%,” the neuroscientist says. “Our algorithm may aid in the identification of the most difficult cases, where the clinical examination is inconclusive, since it could indicate how much an individual’s brain function approximates that of a person with autism.”
Generation of induced pluripotent cells of patients with autistic disorder (No. 11/20683-0); Grant Mechanism PhD Grant; Principal Investigator Patricia Cristina Baleeiro Beltrão Braga (USP); Scholarship Beneficiary Fabiele Baldino Russo; Investment R$73,006.05.
HEINSFELD, A. et al. Identification of autism spectrum disorder using deep learning and the ABIDE dataset. NeuroImage: Clinical. Online. August 30, 2017.
RUSSO, F. et al. Modeling the interplay between neurons and astrocytes in autism using human-induced pluripotent stem cells. Biological Psychiatry. October 2017.
RIBEIRO, S. et al. Barriers to early identification of autism in Brazil. Revista Brasileira de Psiquiatria. October 2017.