The modeler of cell development
It all begins with stem cells. Standing at a whiteboard, Dr. Markus Mittnenzweig sketches a horseshoe shape – a cross-section of a mouse embryo’s early cell structure known as the epiblast. This layer contains embryonic stem cells that can differentiate into any type of cell in the body – a muscle cell, a nerve cell, or a skin cell. But how do they know exactly what they will become, what identity they should assume? A neighboring cell layer sends them commands. These molecular messengers, known as morphogens, work to activate certain genes and thus develop into muscle cells, for example.
Mittnenzweig and his six-person team generate some data themselves in the laboratory. Three-dimensional cell cultures, known as gastruloids, replicate early embryonic development.
This finely tuned orchestra of molecular signaling gradually gives rise to the body’s various cell lineages and the overall blueprint of the embryo – followed by organ development and maturation. “We know the core components of cell differentiation and tissue formation, the key signaling molecules and genes,” says Mittnenzweig. “But we still lack comprehensive models that explain how these elements work together to steer a stem cell down a specific developmental path. To me, it's fascinating to see how a tiny cluster of cells gradually forms complex structures, and how individual cells manage to adopt exactly the states needed to support life.”
Mittnenzweig heads the Computational and developmental biology lab at the Max Delbrück Center. He develops quantitative computer models that analyze the mix of molecular and cellular dynamics needed for zebrafish embryos to develop. His models can estimate the likelihood that a stem cell will follow a particular developmental trajectory – toward cell type A or B, for instance – depending on the activity of specific molecular messengers and genes.
His goal is to advance our understanding of embryonic development: how cells communicate with each other, what regulates their gene activity, and how they differentiate and organize into tissues. But his models also have the potential to shed light on the molecular and cellular origins of disease.
“It’s interesting that the same classes of signaling molecules involved in embryonic development – such as BMP, Wnt, FGF, or SHH – are also used by adult stem cells during tissue differentiation,” he explains. “These molecules are relevant to many cancers and other diseases and could be interesting from a therapeutic perspective.”
Multi-omics, cell cultures, and machine learning
To build his models, Mittnenzweig relies on vast datasets. Genomic and transcriptomic data, for example, offer a snapshot of all active genes or RNA molecules in a cell. Spatial transcriptomics methods can map the spatial distribution of messenger RNAs in tissue, providing clues about how neighboring cells exchange signals. Imaging techniques like fluorescence microscopy add further layers of information.
“But all these measurements are just snapshots, because the cells are destroyed in the process,” Mittnenzweig says. “Our models, on the other hand, can calculate how gene expression looked in a cell hours earlier, or will look later. This allows us to reconstruct the dynamics of cell development.” He uses machine learning tools to integrate and interpret this complex omics data.
One result is a network flow model that uses gene expression data to map when different cell types emerge during embryogenesis – displayed as colorful streams that resemble the shape of a river delta. In future work, he plans to integrate live microscopy and spatial transcriptomics data into the model to reconstruct the dynamics of differentiating cells in three-dimensional tissue.
A look inside an incubator.
To better understand why certain cells change identity during embryonic development, Mittnenzweig also incorporates epigenomic data – information on how gene activity changes without altering the DNA sequence itself. One example is DNA methylation: when methyl groups are added to specific DNA bases, the associated genes are sometimes switched off. Enzymes can also remove these methyl marks. If this process is disrupted – if cells can no longer demethylate properly – it can severely impair development. “Then embryonic development doesn’t progress as it should, and organs don’t form properly,” he explains.
Many of the datasets that feed into his models are produced in collaboration with other research groups. But Mittnenzweig’s six-member team also generates data in their lab. Using specific cell culture techniques, they grow pluripotent stem cells into in three-dimensional structures known as gastruloids that simulate early embryonic development. With single-cell technologies, they then analyze which signals are particularly active during these early stages. “That makes it relatively easy to generate large amounts of valuable data,” Mittnenzweig says.
From mathematics to developmental biology
Mittnenzweig is a data-driven thinker – someone who loves the abstraction of mathematics and who is also fascinated with the molecular detail of biology. He brings both together through modeling. Even as a student at the Free University of Berlin, the Weierstrass Institute for Applied Analysis and Stochastics, and later at the École Normale Supérieure and the Université Pierre et Marie Curie in Paris, he blended mathematics with chemistry. His doctoral research focused on optimal transport problems and reaction-diffusion models, which describe how molecules react and diffuse. That was in 2018. “At the time, the first single-cell RNA studies in developmental biology were being published,” he says. “I found the technology incredibly exciting because it called for new mathematical models to extract meaning from the huge volumes of data.”
At the Weizmann Institute of Science in Israel, he turned his attention to developmental biology. He spent four and a half years in Rehovot, where his wife also held a postdoc position. Their second child was born there. “Watching a living being develop in real life is even more exciting than studying the process,” says Mittnenzweig, who enjoyed spending time with his wife and children at the beach, the campus pool, or at Shabbat dinners with friends. “In winter we’d take day-trips into the beautiful surroundings – up north, near Jerusalem, to the Dead Sea, or to the desert,” he recalls. Meanwhile, his work on optimal transport laid the foundation for the models he still uses today, particularly in studying how DNA methylation affects cell differentiation.
Markus Mittnenzweig with a floorball ball.
In 2023, Mittnenzweig – born in 1989 in Halle – returned to Berlin to establish his own research group at the Max Delbrück Center. “The world-class research here in stem cell differentiation, embryonic development, and epigenetics, plus the technological expertise in single-cell analysis, were compelling reasons to come,” he says. And, of course, he was also drawn back by the chance to rejoin his old floorball club, a sport similar to hockey, where he often plays center or defense.
Our goal is to understand the complex cocktail of signals that cells need for differentiation well enough so that we can eventually control it.
In recent years, biomedical data has been exploding, thanks to advances in omics technologies. “We now know of more and more cell states that we can incorporate into our models,” Mittnenzweig says. This knowledge has also expanded the number of potential therapeutic targets for disorders of embryonic development.
Researchers today can already measure the activity of 20,000 genes across millions of individual cells. The resulting datasets take up terabytes of storage. But for Mittnenzweig, that’s just the beginning. “It’s always great to have more data,” he says. “Our goal is to understand the complex cocktail of signals that cells need for differentiation well enough so that we can eventually control it.” From the very beginning, stem cells could then be steered in ways that ensure the healthiest outcomes.
Text: Mirco Lomoth
