Digital illustration of Prevotella bacteria.
Clarified consensus: Microbes that belong to the Prevotella genus are among those that make up the autism-linked microbiome signature.
KATERYNA KON / Science Source

Gut microbiome meta-analysis reveals consistent autism signal

But the field needs to move on from cross-sectional studies to gain insights into the causes and consequences of the association, experts say.

Despite a decade of inconsistent findings, the microbiomes of autistic and non-autistic children do differ, according to a recent meta-analysis of 10 studies and 15 large datasets.

“It’s not diagnostic just yet, and we still don’t have the causal relationship nailed down,” says co-lead investigator James Morton, former associate research scientist at the Simons Foundation, Spectrum’s parent organization. But, he adds, “this [meta-analysis] provided a hint that we can see these signals consistently across all these studies.”

The metabolites produced by both microbial and brain metabolic pathways diverged between autistic and non-autistic children in the meta-analysis. In addition, 591 microbes were more common in autistic children, and 169 microbes were more common in non-autistic children. The signatures for each group correlated with dietary habits, levels of immune molecules called cytokines, and gene-expression patterns in the brain, indicating that they reflect a global, whole-body relationship between autism and the microbiome.

“They are not able to give concrete pathways, specific bacteria interacting with specific receptors in the human, but they do find a generalized pattern of differences that persists between studies,” says Thomaz Bastiaanssen, postdoctoral scholar and lead bioinformatician in John Cryan’s lab at University College Cork in Ireland, who was not involved in the study. “In that sense, I think it’s more of a confirmation there is something, but what exactly it is is still very much in the open.”


ozens of studies have previously looked for a microbial signature for autism in stool samples from people and reported conflicting results. Most have shown that gut microbes differ between people with and without autism, but the extent of the difference and the exact microbes involved have varied from study to study. Further confounding those findings, a 2021 study suggested that any autism-related microbiome differences in autistic children can be chalked up to a less diverse diet.

Part of the issue is that most of the studies were cross-sectional, so they represent the microbiome at only one point in time. Plus, the studies compared averages between the entire autistic and non-autistic groups, rather than analyzing data on the individual participant level.

“Because most of the studies have been done that way, a lot of information was actually getting lost,” says co-lead investigator Gaspar Taroncher-Oldenburg, director of therapeutic alliances at New York University in New York City and consultant-in-residence at the Simons Foundation Autism Research Initiative (SFARI).

Taroncher-Oldenburg, Morton and their colleagues took a different approach. They developed a statistical algorithm that harmonized the data across the studies and then compared a total of 528 autistic people with 528 age- and sex-matched non-autistic controls. The results were published in June in Nature Neuroscience.

“It’s a statistical tour de force,” Bastiaanssen says.

Microbiome studies typically use siblings of autistic participants as controls because it can help reduce the effect of the environment, says study investigator Maude David, assistant professor of microbiology at Oregon State University in Corvallis. Siblings who live in the same household tend to have similar gut microbes because they drink the same water, eat food from the same stores, and interact with the same pets, she says.

But siblings are not a perfect control because the gut microbiome changes with age. In the meta-analysis, the autism-associated signal vanished when the team analyzed data from autistic people matched to sibling controls.

The results indicate that researchers should be careful to control for the relevant confounding variables based on what type of control is used, David says.


he microbes identified in the meta-analysis corroborate findings from a 2019 fecal matter transplant study. In the open-label phase 1 trial, 18 children with autism underwent two weeks of antibiotics and a bowel cleanse followed by a fecal matter transplant from a non-autistic donor. The treatment reduced gastrointestinal symptoms and eased autism traits as rated by the Childhood Autism Rating Scale. The benefits persisted for two years following the treatment.

Morton reanalyzed the raw data from the trial and found that some of the autism-associated microbes flagged by the meta-analysis were more abundant before treatment and decreased after the treatment.

“These microbes that were initially associated with ASD actually got modulated with a treatment,” says Rosa Krajmalnik-Brown, professor of engineering at Arizona State University in Tempe, who led the 2019 trial.

The meta-analysis marks a transition for the microbiome field, Bastiaanssen says. Now that the prior cross-sectional studies have been combined and the broad group differences parsed out, it’s time to design more specific studies. “In my opinion, the purpose of these large, omics datasets is to then formulate testable, mechanistic hypotheses.”

In order to glean any insight into the direction of causality — whether the microbiome contributes to autism or autism results in an altered microbiome — the next wave of research should come in the form of “longitudinal studies with some kind of intervention component,” Taroncher-Oldenburg says. “That’s the only way.”