Another digest of feeds brought a note: thirty years ago, twenty people with Parkinson's disease received transplants of dopaminergic neurons from embryonic tissue, while the control group had their skulls drilled and nothing injected. Seventeen out of twenty improved, PET showed increased dopamine activity, postmortem analysis confirmed synapses. The comments immediately rang out: "they preferred to forget a successful technology."
What hooked me in this story wasn't the cells. What hooked me was the design: people had their skulls drilled for the sake of experimental purity—in neurosurgery, no one had done that before. And what hooked me were the timelines. If neurons engrafted in the nineties, this is the longest continuous experiment on transplanting cells into the human brain, and it must have outcomes. They exist, and they diverge sharply from how this story is told.
To understand why this work is still discussed, you need to break it down into four independent storylines: how the design emerged, what it showed formally, what they later found in the brains of the deceased, and what grew out of it thirty years later.
Until 1993, U.S. federal policy prohibited the National Institutes of Health from funding the transplantation of human fetal neurons into people. President Clinton lifted that ban—and researchers from Colorado and Columbia immediately proposed not an open observation but a full controlled trial. That's how the first double-blind surgical trial in the history of neurosurgery came to be, and the fact that it emerged exactly there and then is a direct consequence of a political window.
Forty patients aged 34 to 75 with an average disease duration of 14 years were randomized to transplantation and to sham surgery. Transplantation meant cultured mesencephalic tissue from four embryos into both hemispheric putamens. Sham—as the English literature calls the imitation of an intervention—looked like this: they drilled holes in the skull and stopped there, they did not puncture the dura mater. The operations were performed under local anesthesia, the NIH grant was $4.8 million, among participants were 19 women and 21 men, exactly half over sixty.
The ethical war began before the results were published. Arguments against sham included informed consent, therapeutic misconception, bodily integrity, and risk-benefit balance: why maim people for the sake of a control group if you have within-subject measurements before and after. In defense—the argument that without sham you get false-positive results that are later more dangerous for the public. The discussion stretched on for years and still serves as a teaching case in bioethics.
A telling detail: study co-author Stanley Fahn explained to journalists that in Parkinson's disease, about thirty percent of patients feel better just from a placebo pill, and some of those who received the sham surgery sincerely reported improvement. That number was precisely the reason for doing all this.
The primary outcome was subjective global rating of change—a scale from minus three to plus three. After one year, the mean result was 0.0 ± 2.1 in the transplantation group and −0.4 ± 1.7 in the sham group. That is, by the main criterion they saw no difference; the sham surgery gave slight worsening, the real one gave zero.
The difference showed up in a subgroup. In patients sixty years old and younger, formal tests on the morning trial, before taking medications, gave a significant advantage to transplantation: p = 0.01 on the UPDRS scale and p = 0.006 on the Schwab and England scale. In those older than sixty, there was no advantage. In other words, the positive result of the trial rests entirely on half the sample, and age before surgery turned out to be more important than the procedure itself.
Now about the famous "17 out of 20." Growth of fibers from transplanted neurons was documented in 17 patients out of 20—by increased 18F-fluorodopa uptake on PET or at postmortem examination. This is a metric of tissue engraftment, not functional recovery. The popular retelling glues these things into one phrase "17 out of 20 had improved movements," and the story then lives on in distorted form. Curiously, the study's lead investigator himself, ten years later in a review article, described the result more boldly than the primary outcome of his own work: "grafts improved objective signs to better than the best effects of L-DOPA, while sham surgery produced no clinical changes."
Fifteen percent of those operated on, after the first year of improvement, developed recurrent dystonia and dyskinesias—even after reducing or discontinuing levodopa. More on this below, because this point is what stopped the field.
The most intriguing part of the story is postmortem. Patients died from things unrelated to the surgery, and each such case turned into a slice of a unique experiment that cannot be run again.
Already in 1995 the first autopsy appeared: a 59-year-old man, tissue from seven embryos, death after 18 months from massive pulmonary embolism. In his brain they found dense clusters of dopaminergic neurons and extensive striatal reinnervation in a patch-matrix pattern; they found no sprouting of the host's own processes at all. Twenty-four years later the same group described a case with nearly normal density of graft-derived putamen innervation—while the host's brain had degenerated and there were no signs of immune response. Another graft in a different study survived twenty-two years.
Then the real science begins, because in these same sections they found something else entirely. In neurons that were fourteen years old, they discovered Lewy body-type inclusions, stained for α-synuclein and ubiquitin, with reduced dopamine transporter expression. In the twenty-two-year graft, α-synuclein was present in 49% of pigmented neurons, and 1.2% contained true Lewy bodies; in grafts of 12–16 years there was less pathology—meaning it accumulated over time, like in the patient's own neurons.
Here I must mention a contradiction in the sources, because it's glaring. In the same issue of Nature Medicine, a few pages after the article about Lewy bodies, a paper came out with the directly opposite conclusion: five patients after 9–14 years, grafts survived, dopaminergic and serotonergic neurons without pathology. And in 2014 a third group reported that α-synuclein pathology sits in the patients' own substantia nigra and is not found in transplanted neurons. I cannot assemble these three texts into one picture, and judging by the fact that the field still cites all three, I'm not the only one. The difference may be in the staining method, in the timelines, in specific patients—but I won't assert that.
There's also a sadder metric. The same patient with the twenty-four-year graft lost clinical benefit after fourteen years—with preserved morphology. The authors formulate this as an independent conclusion: even a viable graft with extensive reinnervation loses effectiveness if degeneration spreads in the host's brain. And for four patients examined after 16–18 years, the picture is this: two walked without assistance and lived independently, but all four eventually got deep brain stimulation, and dyskinesias in the "off-medication, off-stimulation" state were documented in three.
Dyskinesias are involuntary movements familiar to everyone with Parkinson's disease who has taken levodopa for years. It turned out that transplanted neurons can reproduce the same effect even without drugs. In the Colorado trial, persistent dyskinesias developed in 5 out of 34 patients (15%) despite reducing and discontinuing medications; amantadine, clozapine, anticholinergics, and dopamine depletors helped to a limited extent, metyrosine eliminated movements but returned parkinsonian "off." Three eventually got pallidal stimulators.
In the New York trial with four donors per side, the rate was much higher: 56% of transplanted patients developed dyskinesias persisting after overnight withdrawal of dopaminergic medications. The authors honestly concluded that the method cannot be recommended as therapy.
The mechanism was figured out seven years after the field had already shut down. In two patients with substantial motor recovery after transplantation, they found excessive serotonergic innervation of the grafted striatum, and a 5-HT1A serotonin receptor agonist noticeably reduced movements. The culprits turned out to be serotonergic cells that got into the transplant along with the dopaminergic ones.
Now the connection that seems to me the strongest part of the whole story. Persistent dyskinesias developed in patients who before surgery were milder than the rest and after transplantation improved to almost complete absence of parkinsonism with levodopa withdrawal.
The best response to therapy and the worst side effect turned out to be the same event.
The nervous tissue worked like a pump without feedback: healthy neurons in the striatum regulate release according to need, while the transplanted ones dumped dopamine at their own discretion.
If you look at thirty years of these studies as therapy—you get a list of two failed randomized trials and the open TransEuro, where no overall clinical effect was seen at the primary endpoint, and outcomes depended on which instrument was used for implantation in each of the two centers.
But as an experiment it turned out to be priceless, and here's why. The transplanted neurons did not genetically belong to the patients. This was foreign tissue placed in a brain with Parkinson's disease—and over time Lewy bodies appeared in it. That means pathology spreads from cell to cell inside the human brain. Without these autopsies, such a conclusion would have to be assembled from animal models and circumstantial evidence; here it was obtained in humans, moreover in a controlled trial. From this finding grew the entire modern line of discussion about prion-like mechanisms of neurodegeneration—including the hypothesis that the disease starts in the gut or nasal cavity and crawls up nerve pathways.
And here's an honest rough edge. If α-synuclein spreads so well, why has transmission of Parkinson's disease itself from human to human never been confirmed? Recipients of cadaveric growth hormone are an ideal natural model: they were injected for decades with a preparation from the pituitary gland contaminated with the same biological material. For Alzheimer's disease this line worked: in 2024 they described iatrogenic cases, and in 2026—four recipients with high levels of Alzheimer neuropathological changes.
But for Parkinson's, examination of the same cohort in 2013 found not a single case. Three deaths from amyotrophic lateral sclerosis, zero parkinsonism. What to do with this, I don't know: either α-synuclein passes this pathway worse, or the observation horizon is shorter, or parkinsonism is simply harder to recognize in postmortem statistics. All three options explain the data, and there's nothing to choose between them.
The field returned twenty years later, changing the cell source. First in 2020 they described a personal case: autologous progenitors from induced pluripotent cells of the patient himself, manufactured according to GMP rules.
Then came series. In Kyoto they transplanted iPS-derived progenitors into seven patients aged 50–69 and observed for 24 months: no serious adverse events, 73 mild and moderate, no tumor growth on MRI, 18F-DOPA uptake in putamen increased on average by 44.7%, in four of six assessed the Hoehn and Yahr stage improved.
In Swedish STEM-PD—eight patients with moderate disease, two escalating doses, twelve months of immunosuppression, and one died from pulmonary infection; no serious events related to the cell product, no dyskinesias either. In the BlueRock trial—twelve patients, doses of 0.9 and 2.7 million cells, after eighteen months the motor part of UPDRS in the high dose improved on average by 23 points, no dyskinesias documented.
Two comparisons with the original suggest themselves. Embryonic grafts from the nineties survived beautifully without any immunosuppression—so much so that for years it served as proof of the brain's immune privilege. Modern products require a year of immune suppression, and the price is already visible: the death from pulmonary infection in the Swedish trial was attributed precisely to the immunosuppression regimen. And the logistics changed. Instead of four embryos per patient, which physically had to be obtained over weeks—a cryopreserved product from the warehouse: those very "off-the-shelf" cells that can be shipped between clinics in batches.
At the same time, immune privilege itself had to be reconsidered. In 2007, in patients with transplants of fetal tissue for Huntington's chorea, they found alloimmunization to donor antigens without signs of rejection in four of thirteen, and in a fifth—a documented rejection process, completely reversible under immunosuppression. The authors directly wrote that some results of past studies could have been distorted by unrecognized immune response.
Politics, by the way, had time to repeat itself. In 2019 restrictions on fetal tissue returned, in April 2021 they were lifted again. The same debate, the same sides, thirty years later—only now cells are taken from the patient's skin, and the political part of the question doesn't work for them at all.
The ending of this story is a draw, not a tragedy. Twenty people, each of whom had their skull drilled, gave medicine a twenty-four-year series of slices of the human brain—and showed along the way that the best effect of therapy can be its own main side effect. And fetal neurons in someone else's head, apparently, continue to work as long as the host can endure.