Over his short career, 44-year-old Sergiu Pașca, MD, has changed how scientists study the developing human nervous system.

It started in 2017, with Pașca and his Stanford University colleagues building more complex three-dimensional neural development models using human pluripotent stem cells (hPSCs). These organoids, the oft-misnomered “mini brains,” allowed researchers to observe human neurodevelopment in the lab and assemble neural regions into functional circuits.

In 2020, Pașca’s group reported fusing organoids of different lineages to create self-organizing 3D tissue models. One such assembloid was created by merging human cortical neuron organoids, spinal cord motor neuron organoids, and striated muscle organoids that translated cortical activity into muscle contraction. The textbook corticomuscular diagram of a white space with a brain, spinal cord slice, and muscle wired together came to life.

But such systems were limited because a brain doesn’t develop in a dish. Organoids lack sensory, vascular, immune, long-range circuitry, and behavioral outputs. Later in 2022, Pașca’s group transplanted human cortical organoids into newborn rats’ brains. The tissue grew and integrated with the host nervous system: human neurons responded to sensory input, extended projections through the rat brain, and could influence behavior when activated experimentally. The model has enabled researchers to study patient-derived human neurons in a living neural circuit and link molecular and cellular abnormalities to circuit and behavioral phenotypes.

That entire time, Pașca was working on a related project that was far more complex, both experimentally and ethically. In a new Nature study, Pașca and his colleagues describe xenocortication, an approach in which human cortical organoids are transplanted into mice genetically engineered so that most of the neocortex and hippocampus never form. By three months, human tissue took advantage of the space, generated diverse populations of human cortical cells, formed long-range connections with the mouse nervous system, and developed organized electrical activity.

By providing a large volume of developing human cortical tissue inside a living nervous system, the model potentially allows researchers to investigate human cells and circuits that are difficult to study in vitro. For proof-of-concept, Pașca’s group demonstrated several applications, including modeling neuropsychiatric disease and hypoxic injury. “This is not an all-in-one, universal system that is here to replace previous models, but rather to complement them,” Pașca told Inside Precision Medicine.

These hypoxic injury experiments highlighted, for Pașca, the degree of engraftment and functional integration between the human tissue and the host mouse. “When we induce hypoxia in the xenocortical mice, the microglia from the mouse react to hypoxia,” said Pașca. “But they’re not reacting to a hypoxic injury of the mouse cortex because there is essentially no mouse cortex there. They’re reacting to the injury of human cells.”

In six years, Pașca’s group has moved from assembling pieces of human neural circuitry in a dish to integrating them into a living rodent brain to redesigning the host brain itself to give human neural tissue room to grow.

Making room for human tissue

The work centers on immunodeficient “apallial” mice, which lack much of the dorsal and medial pallium, which gives rise to the neocortex and hippocampus. In cells expressing the pallial marker Emx1 on an immunocompromised SCID background, the team deleted Esco2, a sister chromatid cohesion gene. Creating mice without most of the neocortex and hippocampus gave researchers plenty of space to transplant human cortical organoids shortly after birth. 

The engraftment of three different hPSC lines was effective, with 86.2% of the 29 transplanted animals showing successful engraftment. Graft volume increased 4.7-fold in two to three months after transplantation. Human-derived tissue made up 91.9% of cortical tissue volume after three months. “It’s still not a fully formed human cortex, but it contains a large diversity of cortical cell types, including astrocytes,” said Pașca.

Size wasn’t everything. The extra space seemed to affect how much human tissue could grow and how it connected to the nervous system. Human neurons projected organized axonal tracts toward the superior colliculus from subcortical structures. In contrast, mouse neurons from the paleocortex, thalamus, and pallidum projected into the human graft. Human-derived projections were even found in the cervical spinal cord, which were not seen when cortical organoids were transplanted into mice with intact cortex.

Modeling disease and injury in human tissue

The researchers then examined graft cooperation. Wide-field calcium imaging showed synchronized events across human tissue, while electrophysiological recordings showed coordinated bursts throughout the graft. In mice, human graft activity correlated with orofacial movements.

That the model can go from human cellular phenotypes to circuits and behavior may be its most important feature. “Just because the neuron is hyperexcitable in a dish doesn’t mean they will result in a seizure or in the EEG changes that are characteristic of that condition,” Pașca said.

The distinction is crucial for studying neuropsychiatric disease. A cellular phenotype in vitro can reveal a key mechanism, but networks of interacting cells alter cognition and behavior to cause psychiatric disorders. “For many neuropsychiatric disorders, circuit and behavioral readouts are very important because psychiatric disorders are behaviorally defined,” Pașca said.

Xenocortication allows patient-derived human cells to be studied molecularly and physiologically while part of an animal’s circuitry. “We think these are a somewhat narrow but important set of applications,” Pașca said, “that capture a unique feature of the model.” The researchers also modeled severe hypoxia. After oxygen deprivation, human grafts showed strong HIF1α immunoreactivity, while adjacent mouse paleocortex did not show a similar signal. Injuries also altered motor behavior.

Pașca envisions xenocortication being used to investigate genetic and environmental perturbations, test therapeutics across relatively large volumes of human neural tissue, and potentially inform preclinical studies of cell therapies for conditions such as microcephaly or severe ischemic injury early in development.

Solving space, but not time

Despite expanding territory, xenocortication cannot eliminate another major drawback of cross-species transplantation. “While we create more space, we still don’t solve the problem of time,” Pașca said. 

Neural development follows a human schedule. Grafted human glutamatergic neurons were transcriptionally comparable to the developing human cortex around mid-gestation 24 weeks after differentiation. “The cells are still progressing at their own pace,” Pașca said. The mouse nervous system, meanwhile, matures far more quickly. “The critical periods of the mouse are closing one by one over the ensuing weeks after transplantation,” he said. 

That mismatch may limit how completely the two nervous systems can integrate. The human grafts lacked several hallmarks of mature cortical organization. There was no canonical cortical lamination, for example, although related neuronal populations showed evidence of local self-organization. “There’s no lamination in the human graft in the xenocortical animals,” Pașca said. “That’s partly because the cells don’t know where up and down really is. They’re not anchored in that way.”

There were, however, “attempts at cytoarchitecture,” he said, and the graft contained a broad diversity of human cortical cell types. The result is not a fully formed human cortex but rather developing human cortical tissue occupying an unusually large volume inside a living mammalian nervous system.

Yet anatomical connectivity alone does not establish what those human neurons are functionally contributing. Pathway-specific experiments will be needed to determine which graft-host connections are functional and whether particular human neuronal populations are necessary or sufficient for specific behavioral effects.

The experiment inside the experiment

Yet xenocortication also produced another finding, which goes back to the apallial mice and has relatively little to do with the human graft itself. Scientists created mice without most of the neocortex and hippocampus to accommodate human tissue. The mechanism is different from destroying a developed cortex, says Pașca. “This is not a depletion. It’s not an ablation,” he said. “This is sort of like a prevention of the formation. It’s a blockade.”

Those mice functioned far better than their anatomy might suggest. Apallial mice could move around their environment, see, hear, and smell. Gait, limb coordination, and spontaneous behavior were abnormal, but gross locomotion was preserved. Cognitive deficits also emerged.

The finding does not mean the cortex is dispensable. Pasca suggests that their nervous systems developed early around its absence, allowing other neural structures to compensate. “The nervous system therefore has to deal with the lack of cortex very early on,” Pașca said. “I think that probably creates some interesting opportunities for that plastic nervous system to compensate to some extent.” He suspects preserved subcortical structures may be assuming functions that would ordinarily involve cortical circuitry.

Xenocortication was built to ask what human cortical tissue can do when it is finally given room to grow. The mice required to make that experiment possible may now offer another model altogether: a way to ask how a developing nervous system reorganizes when the cortex that normally dominates it was barely there in the first place.

How far should xenocortication go?

As human neural grafts become larger and more integrated, the work also raises questions about how far such models should go. For example, one theoretical way to overcome the developmental mismatch in timing between neuronal and rodent cells is to use an animal with a longer developmental timeline or one that is evolutionary closer to humans. But Pașca is wary of that step. “If this were to be done in an animal where there is more space, or an animal that is evolutionarily closer to us, then the probability of integration is larger,” he said. “I don’t think it’s justified at this point.”

Pașca said the project underwent extensive ethical scrutiny, including consultation with an external committee, and the paper calls for proactive ethical engagement as researchers contemplate more mature or complex human neural grafts.

The immediate goal is to see what researchers can do with the platform they already have, and Pașca expects stem cell and animal model labs to adopt it without much technical difficulty. “You don’t need to request those because you can pretty much make them,” he said of the mice. “The Emx1-Cre mice already exist. You just cross them with the right line and then cross them to make them immunocompromised.” The human side of the system is increasingly accessible as well. Pașca said his laboratory has taught nearly 500 laboratories around the world to generate organoids.

Still, xenocortication is not a plug-and-play technique. “This is not like CRISPR that you can easily implement in lab,” Pașca said. “This is much more elaborate. It involves animals, live animals, human stem cells, and long-term experiments.”

But broad adoption was one reason for creating a reproducible system rather than a one-off experiment. “It’s not going to be a trivial experiment to do, and not many labs are going to do it,” Pașca said. “But hopefully others will do it, too. We built the system so others can also use it. We’re not going to be able to do everything.”

So, what’s next for Pașca? It may seem like he’s jumping from model to model, but he insists that it’s all been done to be a critical model for the field and his lab, not to push ethical or Frankensteinian boundaries. “My approach has always been quite utilitarian,” he said. “We have a specific goal in terms of disease, and then we model that in a specific way. We built a system for about seven years so that we can actually apply it, not to build the next one.”