Scientists who study the brain have long tried to unlock the neural mysteries of schizophrenia and bipolar disorder. They can be difficult to diagnose and even harder to treat.

A group of scientists at Johns Hopkins University, Baltimore, may be a step closer to more accurately diagnosing those brain disorders. They developed lab-grown “tiny brains” with functional neurons from the cells of people with schizophrenia and bipolar disorder and used a novel machine learning “pipeline” to identify neural firing patterns linked to those conditions. The research was recently published in APL Bioengineering.

How Does One Create a ‘Tiny Brain’ From Scratch?

The scientists began their research by generating induced pluripotent stem cells — adult cells that are reprogrammed to an embryonic-like state which can then be differentiated into any cell type a scientist needs.

“So I can take a hair cell or skin cell or even blood cells and make pluripotent stem cells,” said coauthor Annie Kathuria, PhD, assistant professor of biomedical engineering at Johns Hopkins, who noted that scientists Shinya Yamanaka and Sir John B. Gurdon were awarded the Nobel Prize in 2012 for coming up with the technology.

Kathuria and her colleagues used blood cells from individuals with schizophrenia and bipolar disorders to create pluripotent stem cells.

“The beauty of it is, the blood cells carry the genetic signature of the person you take them from, so it’s very personalized to that person,” she said. “It is very, very useful for studying neuropsychiatric disorders because they’re so diverse in nature, and everybody gets affected a little differently.”

For example, there’s schizophrenia disorder, but there is also schizoaffective disorder. There is also treatment-resistant schizophrenia. There are several types of bipolar disorder. Doctors see similar variation with anxiety disorders, added Kathuria.

Once the stem cells were made, Kathuria, a tissue engineer, and colleagues used them to generate “brain-like structures in the dish.”

“We aggregate these stem cells. Basically, a lot of developmental biologists have figured out ways to add in small chemical molecules at different times of development and aggregate them literally into ball-like structures — teeny, tiny ball-like structures,” said Kathuria, explaining that the chemical culture they grow in is concocted mostly from a mix of sugar and certain nutrients.

At this stage, they are not yet visible to the human eye. But they grow and expand over time and eventually can be seen. It can take time — the tiny brains are in culture for 6 months before they reach 3-4 mm in diameter.

“Imagine a small pea or a small blueberry,” said Kathuria.

From Nothing to Neurons

By giving the cell balls different molecules at different times, scientists are able to disaggregate them into the type of structure they want to study; in this case, neurons, the primary functional cells of the brain.

“We start with neural development — we do neural tube formation. Once the neural tube is done [it takes about 7 days in culture], the next step depends on what area of the brain I am studying,” said Kathuria. “I will divide or push the cells to go in the direction I want. So if I’m trying to make the prefrontal cortex, I will add molecules that will trigger that pathway of the brain’s development. If I’m trying to get the brainstem or spinal cord in order to study movement disorders, I will use other molecules.”

For neuropsychiatric disorders, she said they always develop the prefrontal cortex, where cognitive ability — decision-making ability — happens in the brain.

As the cells develop into organoids, “Sometimes we have to play around with the dosing and the timing,” Kathuria said. “That’s where the engineering comes in — finding out what pathways to use, what dosing to use. And then, once the cells develop into the little brain-like structures and they are big enough, we encase them in an extracellular matrix.”

The extracellular matrix is the skeletal mass system of the body that gives humans the support structure to expand and grow.

“It’s like a sticky glue,” said Kathuria, and it can take at least 2 months or more after the stem cell stage to reach the brain organoid stage. From one stem cell, the researchers can make hundreds of these brain-like organoids.

Slicing, Dicing, and Using Machine Learning

Once the scientists developed the cells from people with schizophrenia and bipolar disorder into the brain-like organoids, they were ready for next steps. Kathuria and colleagues sliced the tiny brains and, using microscopy, stained the slices to look for different proteins.

“We looked at all of the cell types we differentiated from the stem cells. We also did a lot of RNA sequencing — single nuclear RNA sequencing. So what that means is, I will literally pull the organ-like structure apart and I will extract and sequence the nuclei of the neurons to understand what cell type markers they’re expressing,” said Kathuria.

Because the way brain cells communicate with each other, or fail to communicate, is believed to be one key to brain disorders, Kathuria and her team also collected lots of electrophysiological data — electrical current activity — from the organoids. Next, they ran their data through a novel machine learning pipeline (or algorithm) developed with “a lot of math and coding” by Biomedical Engineer Kai Cheng, PhD, and Sridevi V. Sarma, PhD, professor of biomedical engineering at Johns Hopkins, to identify neural firing patterns associated with the healthy and unhealthy tiny brains.

“It was blinded to us. We didn’t know which organoid was the control, which was schizophrenia, and which was bipolar,” Kathuria said.

She said the program basically created a map — a signature — of the electrical activity of each organoid, serving as biomarkers for schizophrenia and bipolar disorder. The researchers were able to tell with 83% accuracy which organoids came from patients with the mental health conditions.

What It Means

Constanza Morén, PhD, a researcher in the Basic and Translational Research Laboratory in Schizophrenia at the Hospital Clínic of Barcelona, and professor at the University of Barcelona, both in Barcelona, Spain, (who was not involved in the study), said, “I find this a fascinating and highly promising field of research. This study uses brain organoids and artificial intelligence to detect electrophysiological patterns specific to schizophrenia and bipolar disorder, representing an innovative and forward-looking approach toward [identifying] objective biomarkers in mental health.”

Morén said the study’s main strengths lie in the use of patient-derived induced pluripotent stem cell models, the combination of electrical stimulation with machine learning analysis, and the ability to distinguish between groups with high accuracy.

However, she said future research should include a larger sample size (the current study included 12 participants), “including medicated and unmedicated groups to disentangle the effects of the illness from those of treatment, and incorporating genetically characterized cohorts to refine the interpretation of the findings.”

Morén noted that genetics are not 100% responsible for mental health conditions, but that “the use of samples from diagnosed patients carrying mutations directly related to disease onset (such as 22q11.2 deletions, SHANK3 mutation, or others) could be valuable to include.”

She also pointed out that the procedures used for the research are costly in terms of time and resources, which “raises the question of how close we really are to translating these promising tools into clinical practice.”

Henry Brem, MD, Harvey Cushing professor and director of Neurosurgery at Johns Hopkins University, said, “I think it’s really breakthrough work because it takes very complex psychiatric problems and in a model in the laboratory, it allows [researchers] to study unique characteristics that are patient-specific and to develop treatments based on that.”

Brem, who is from the same academic institution as Kathuria but was not involved with the study, said, “The idea of having distinct electrophysiological signatures associated with psychiatric diseases is extraordinary. And being able to test different treatments in a laboratory in a safe manner and seeing if that electrophysiology changes, just opens up tremendous opportunities for both better understanding these diseases and for developing appropriate personalized medicine for patients.”

Brem also said the approach “is a big step forward” in the push to get away from using animals for experimentation and to instead use nonanimal models to study disease.

What Lies Ahead?

Kathuria said larger studies are the next goal.

“With this paper, we have established a machine learning pipeline and we can now make maps of different brain organoids derived from the cells of a control healthy person, a person with schizophrenia, and a person with bipolar disorder. We have only tested it on 12 samples, so far, so we need to increase that size to make it more robust,” she said.

She would also like to do a similar study adding EEG test results from the same patients who provide stem cells for the mini brains.

“I’d like to do the same electrophysiological analysis and mapping [using EEG data] that we performed on the organoids in this study to see how accurate we are compared to the actual human brain EEG,” she said.

This study was supported by National Institutes of Health grants. The authors have no conflicts to disclose.