A meeting of the minds drives discovery in the brain

In the lab of Hongjun Song, PhD, and Guo-li Ming, MD, PhD, something like a clinical trial is underway, except that no human patients are involved.

Instead, hundreds of “mini brain tumors” immersed in cocktails of culture media and chemotherapies are the study subjects. “We call these avatars for testing cancer treatment,” Song said.

These micro-sized organoids—three-dimensional cultures of brain cells—contain cells from surgical tumor samples of patients at Penn Medicine's Abramson Cancer Center with recurrent glioblastoma, a type of aggressive brain tumor that returns after treatment.

In a recent paper, Song, Ming, and other collaborators at Penn showed that the organoids they developed could serve as a real-time mimic of patients’ brain tumors. In other words, if the treatment works in the organoid, it works in the human patient as well. “The assumption here is that only certain treatments can work on certain patients,” Song said. “The key is to find a match.”

Unlike human patients, these organoids enable testing of hundreds, even thousands, of therapies on the patients’ brain tumors simultaneously—a flexibility that no clinical trials could ever achieve.

This latest work is the culmination of decades of research by Song and Ming, who are a married couple in addition to long-term professional collaborators, on brain organoid models for studying neurological diseases.

When asked to describe the journey that brought them to this point in their career, Song and Ming have two words: it's a series of twists and turns.

Song became chair of the department of Neuroscience at Penn Medicine in November 2025, and is the David J. Mahoney Professor of Neurological Sciences. Both he and Ming, who is a Perelman Professor of Neuroscience and associate director of the Institute for Regenerative Medicine, have been at Penn’s Perelman School of Medicine since 2017. The two share more than 250 peer-reviewed articles and the mentorship of numerous talented early-career scientists.

But long before Song and Ming were research partners, they were high school sweethearts in Wuhan, China.

When Song came to the U.S. from China in 1992 to pursue a PhD at Columbia University, Ming, whose parents aspired to have a doctor in the family, was four years into medical school—two years short of graduation. “My parents were even unsure that I should go to the U.S.,” she said.

Nevertheless, right after she graduated, Song and Ming got married and began laying down roots in the U.S., where they grew their family along with two scientific paths that steadily twined together.

And as they became partners in life, so are they partners in their research endeavours.

Working towards one goal

While Song pursued his PhD, Ming was considering whether to continue with a medical career or pursue one in research. A few years later, Song’s mentor, Mu Ming Poo, PhD, relocated to the University of California, San Diego (UCSD), where the two followed and Ming began working in the same lab, first as a research associate. A month after beginning as a PhD student in the lab, she gave birth to the couple’s first child, a son.

“But there’s a twist here,” Ming said. Halfway through her PhD, Poo moved again to Berkeley, but Ming decided to stay at UCSD to finish her PhD training while Song was continuing his postdoctoral training at the Salk Institute across the road.

In 2002, Song and Ming joined as assistant professors at the newly founded Institute for Cell Engineering at the Johns Hopkins University School of Medicine. A few months later, they welcomed their younger child, a daughter.

One of the first projects in Song’s lab focused on turning neural stem cells back into embryonic-like stem cells. The ability to create stem cells in this way, which was just emerging at the time, is a critical process in modeling how diseases develop and progress because it provides an unlimited source for patient-specific cells containing their genetic information. Meanwhile, Ming studied how neural connections develop using axons from frog spinal cord cells as a model system.

“There is another twist here,” Song said. Around that time, groundbreaking work of Japanese scientist and Nobel laureate Shinya Yamanaka, transformed the way scientists could generate these stem cells. While researchers were fusing somatic cells with embryonic stem cells to obtain a pluripotent or immature cell state, Yamanaka discovered a more efficient method using transcription factors. Inspired by this new method of inducing pluripotency, Song redirected his approach.

Not long after that, Ming followed suit. “As a trained physician, I realized that I was drawn to questions that have more clinical relevance in humans,” she said. She began to investigate a risk gene for schizophrenia and other major mental disorders called Disrupted-in-Schizophrenia 1 (DISC1) to understand how the disorder originates in early brain development. She went on to partner with a clinical colleague and ultimately derive induced pluripotent stem cells (iPSCs) from cell samples from multiple members of a family with a rare DISC1 mutation. That enabled the researchers to follow the course of brain development and link the genetic mutation in DISC1 to synaptic dysfunctions that characterize schizophrenia.

“Guo-li’s lab was one of the earliest labs in the whole world to use cells from patients with psychiatric diseases to generate iPSCs and model neurological diseases,” Song said.

Through that study, it became clear to both Song and Ming that effective disease modeling is the gateway to not just understand how diseases develop in the body, but also to discover new treatments that are likely to be effective in human patients.

“Many therapies or treatment strategies developed on model systems cannot translate to the clinic because they were based on animals,” Song said. However, accessing human brain cells is challenging. “That’s why the breakthrough of generating stem cells from patients is extremely useful and important,” he added. The entire genetic makeup of a patient is contained in their stem cells, allowing scientists to make direct observations rather than inferences from animals.

About a decade later, they showed that through genome editing, they could correct the mutations in the cells of patients with schizophrenia and restore the cells’ normal function. They further identified an effective drug treatment, and when mice with the same mutation were treated, their social and cognitive abnormalities were reversed.

“This was before CRISPR was a thing,” Song said. The ability to correct a mutated gene in living brain cells and show its effect was ahead of its time, he added. But they didn’t stop there.

After their success with cortical neurons, Song and Ming next began to explore if they could generate different types of neurons to study diseases using iPSCs. Even more, could they grow the relevant neurons in 3D to mimic the architecture of a brain?

At the time, other researchers had developed methods to grow brain organoids using cup-sized bioreactors. “These cups are huge, and the experiment is very, very costly to do,” Song said.

The solution eventually came from an unexpected source: three high school students doing summer research in the lab with Song and Ming. The idea was for the students to 3D-print tiny versions of the bioreactor in their classroom.

“It was supposed to be just a fun project,” Ming said. “We were really surprised that it actually worked.”

Afterwards, the team began putting together new-to-the-field protocols for generating brain-region-specific organoids, including mini midbrain, hypothalamus, and cortex, using these miniaturized bioreactors.

Next, they wanted to assess how these organoids could fare as models for neurological diseases.

As they were refining the protocols, one of their graduate students, Xuyu Qian, PhD, who is now an assistant professor of Pediatrics at Penn and Children’s Hospital of Philadelphia, proposed a test case that was making headlines that year: the Zika virus.

At the time of the 2015-2016 Zika outbreak, many conflicting theories had spread on why the virus caused microcephaly, a condition where infants born to an infected mother had smaller-than-normal head sizes. Working with Hengli Tang, PhD, a virologist at Florida State University who studies RNA viruses, they were able to grow Zika-infected brain organoids that grew at the same pace as real developing human brains—meaning they could monitor each step that a fetal brain would go through. Within days, they found a clue: The virus attacks the neural stem cells.

“We immediately know this is a very important hint,” Ming said. “Zika virus can infect the neural stem cells and cause cell death and reduced ability to proliferate, and that could be directly linked to microcephaly.” In merely 10 days, the researchers prepared and submitted a manuscript detailing the results. When the paper came out, it made headlines in major news outlets, including the New York Times.

A partnership advancing science through collaboration

Since coming to Penn not long after their Zika discovery, Song and Ming have focused on improving their brain organoids to model neural circuitry in the hope of recapitulating how different parts of the brain interact.

Their individual expertise made the partnership so effective, Song said. “I was trained purely as a basic scientist. But because of Guo-li’s clinical training, that allows us to think not only about the mechanisms of brain development, but also how they contribute to disease, and what treatment strategies we can come up with.”

Together, they “bring a rare combination of leadership and scientific vision that has pushed stem cell–based human models into a new era,” said Sergiu Pașca, MD, a Kenneth T. Norris Professor of Psychiatry and Behavioral Sciences at Stanford University, whose work often coincides with Ming and Song’s, and whose lab pioneers the development of “assembloids”—a neural model made out of a network of organoids.

At Penn, Song and Ming continued to collaborate across disciplines. They developed a sophisticated organoid model of the human neocortex containing six distinct layers, each with specific functions. With H. Isaac Chen, MD, an associate professor of Neurosurgery at Penn and the Veterans' Administration Medical Center, they also showed that these organoids can be transplanted and seamlessly integrated into the brain of an injured rat to restore visual functions.

“In the next frontier,” Ming said, “we want to study how brain organoids can be involved in brain repair in larger animals.”

Song and Ming also continue to push the boundaries of understanding and treating aggressive brain cancers based on work with organoid models. Their collaboration with Donald O’Rourke, MD, the John Templeton, Jr. M.D Endowed Professor in Neurosurgery and director of the Glioblastoma Translational Center of Excellence at the Abramson Cancer Center, began about 10 years ago, when the couple arrived at Penn.

“There were no models for glioblastoma that were great,” O’Rourke said. “We have used cell lines that are not representative of the tumor; nude mice that don’t have an immune system and only partially recapitulate the actual condition you’re trying to test.”

In addition, glioblastoma is notoriously heterogeneous, and genetic mutations that drive the cancer can be difficult to model. But the organoids offered a much-needed solution: a close biological resemblance to tumors for each patient.

In 2020, the team reported the first patient-derived glioblastoma organoids, complete with the tumor microenvironment. “When the tumor was compared to the organoid, there was 98 to 99 percent fidelity in genetic makeups,” O’Rourke said, “that's hugely advantageous, because now you can study the real problem in human tissue, not a mouse tissue.”

The cells around a tumor play a critical role in cancer drug resistance and relapse, Song explained. By surgically removing a small piece of the patient’s tumor and growing it in the same medium used for sustaining brain organoids, the researchers could preserve this tumor microenvironment, creating a more realistic model of the cancer.

“This is by far the best model that I’ve ever seen,” O’Rourke said. With the organoids, scientists can now gather correlative biological information on patients, which is useful for understanding genetic responses to treatments, he said.

Song attributed the success of the tumor organoid work to Penn’s collaborative environment. “This study needed cooperation from everybody: the pathologists, neurosurgeons, neuro-oncologists, and the students and postdocs in the lab,” he said. “This probably can only do well at Penn.”

“The use of organoids may really help us make advances that have just not been possible before,” said Peter Marks, MD, PhD, former director of the FDA's Center for Biologics Evaluation and Research. “The brain has always been a challenging area to investigate. This work helps bridge the basic biology over to the clinic, and I’m very excited for what they are going to be doing next.”