On June 9, 2026, Life Biosciences issued a press release announcing that the first participant had been dosed. Based on publicly available trial registrations and company announcements through September 2026, this is the first Yamanaka factor partial reprogramming trial to have publicly entered human dosing.
The drug was injected into one eye.
Trial registration number: NCT07290244. This registration contains two kinds of fields, and they read like two different things.
The narrative fields are about aging. The detailed description states that ER-100 targets “age-related optic neuropathies,” that it is “designed to address cellular aging through epigenetic reprogramming,” and the registered keywords include Aging/Cellular Aging.
The evaluable fields are about something else. The listed indications are open-angle glaucoma and non-arteritic anterior ischemic optic neuropathy. All twenty-two primary endpoints are safety items. Aging appears in the description; it does not appear in any field that will be used to judge success or failure.
That is where this field actually stands today.
From “Revert to Embryo” to “Don’t Go Too Far”
In 2006, Takahashi and Yamanaka demonstrated in mouse fibroblasts that introducing four transcription factors — Oct4, Sox2, Klf4, and c-Myc — could reverse a fully differentiated somatic cell to a pluripotent state. These became known as the Yamanaka factors, and the work earned the Nobel Prize in Physiology or Medicine six years later.
The problem: this worked in a dish. Translating it into a living organism was another matter entirely. In 2013, Abad et al. reported in Nature that inducing full reprogramming in living mice did produce cells with pluripotent characteristics — along with teratomas. Once complete reprogramming was triggered, where the cells stopped was no longer under the control of whoever induced it.
So the field shifted toward an idea that sounds simple: don’t let it go all the way. Express the factors briefly, cyclically, move the epigenetic marks in a younger direction, and shut the switch before cells lose their identity. This is partial reprogramming.
The entire technical difficulty of that idea lives in five words: knowing when to stop.
The Animal Evidence Is Not One Continuous Line
Science writing often arranges animal experiments into a single unbroken chain: Ocampo proved you can extend lifespan, Lu proved you can restore vision, Browder proved longer treatment is more effective, Macip proved it works in older mice too. It reads as one technology steadily advancing.
There are actually at least three distinct lines here. Start with the two most commonly conflated.
The first is the OSKM transgenic mouse. The study published by Ocampo et al. in Cell in 2016 used transgenic mice carrying a four-factor cassette in the genome, controlled by doxycycline. Cyclic induction extended lifespan and improved multiple aging hallmarks in LAKI progeroid model mice. Worth noting: the same paper’s work in wild-type mice focused on recovery from metabolic disease and muscle injury. The paper does not report a lifespan experiment in wild-type mice. Browder et al.’s 2022 Nature Aging paper continues this same line, also using inducible OSKM transgenic mice.
The second is AAV-delivered OSK. Lu et al.’s 2020 Nature study removed c-Myc, used AAV2 delivered via intravitreal injection to transduce mouse retinal ganglion cells, and observed a youthened DNA methylation pattern, axon regeneration, and visual improvements in a glaucoma model and in aged mice. The study showed this effect depends on the demethylases TET1 and TET2. c-Myc was removed because it is a proto-oncogene and not required to initiate reprogramming — not because anyone had done a head-to-head comparison of OSKM versus OSK in the same model. ER-100, which has entered human trials, follows this line.
The two lines use different factor combinations, different delivery methods, and act on different cells. Taking the clock changes Browder observed in transgenic mice as evidence for AAV-OSK efficacy is connecting the wrong wires.
A note on conflicts of interest in Lu 2020. Corresponding author David Sinclair is an advisor, board member, named inventor, and equity holder in Iduna Therapeutics, a subsidiary of Life Biosciences; several co-authors also hold related equity, and two patents derived from this work had been filed. This is all disclosed in the paper itself — just read it that way. It does not invalidate the experimental data, but when you read subsequent company press releases, you should know this relationship exists.
Figure 1: Full reprogramming crosses the line of cellular identity, arriving at pluripotency and teratomas. Partial reprogramming shuts the switch before crossing. The three experimental lines below differ in factor combination and delivery method.
When the Clock Runs Back, the Whole Mouse Doesn’t Run Back Together
Browder 2022 deserves its own section, because the same paper that contains the most widely cited positive results in this field also contains a set of findings — written plainly in the original text — that are cited far less often.
The study put wild-type mice through long-term partial reprogramming: one group starting at 15 months of age for seven months, another starting at 12 months for ten months. In the group starting at 15 months, the LUC epigenetic clock declined significantly in kidney and skin. The next half-sentence in the original is cited much less often: other tested tissues did not follow. Liver, lung, muscle, and spleen showed no corresponding effect.
The same paper also includes a third group: mice starting at 25 months of age for only one month. In those tissues, the significant clock decline seen in the longer-treated groups did not appear. This group had small sample sizes, so “did not reach significance” is not the same as “no molecular changes at all” — but it is enough to establish something: the same treatment, with a different starting age and duration, produces different readings.
If partial reprogramming is resetting a body-wide clock, it does so very selectively by tissue, and by timing.
109% and 7% Are the Same Numbers
In 2024, Macip et al. (Rejuvenate Bio) published a result in Cellular Reprogramming: wild-type male mice aged 124 weeks, given systemic AAV9-delivered doxycycline-inducible OSK, showed a 109% increase in median remaining lifespan.
The numbers in the paper read as follows. The control group’s median total lifespan was approximately 133 weeks; the treatment group’s was 142.5 weeks. The 109% figure comes from a different calculation: at the time of injection, the control group had 8.86 weeks of remaining lifespan, and the treatment group had 18.5 weeks.
Same mice. Same data. With total lifespan as the denominator, the increase is roughly 7% (142.5 divided by 133 — my own calculation; the paper does not present it this way). With remaining lifespan at injection as the denominator, it is 109%.
Neither number is false. But which one a piece of writing chooses for its headline determines whether readers picture “living a bit longer” or “living twice as long.”
A few other conditions in this paper are worth keeping in view. The experiment used only male mice; the paper notes in its limitations that aged female mice were difficult to obtain. The control group received PBS-formulated buffer plus doxycycline, without an empty-vector AAV control, so the effect of the viral vector itself was not subtracted out. As for the “human cell epigenetic age reversal” data in the same paper: those experiments used HEK001, a keratinocyte line derived from a single 65-year-old male’s scalp, with n equal to 2 technical replicates.
I am not saying this paper should not be cited. I am saying that when it gets condensed into “OSK doubles mouse lifespan,” none of those conditions travel with it.
Figure 2: The difference in median total lifespan between control and treatment groups is roughly seven percentage points. The 109% figure comes from switching the denominator to remaining lifespan at the point of injection.
There Is Also a Positive Result That Shouldn’t Be Skipped
Presenting only the above would be its own form of cherry-picking.
In 2024, Sahu et al. published a study in Science Translational Medicine built on a different idea: rather than driving OSK expression uniformly across all cells in the body, make it preferentially active in cells that are already senescent or stressed. They placed OSK under the control of the Cdkn2a promoter, which is more active in senescent and stressed cells, so OSK expression was concentrated where cellular aging was already underway.
To be precise, this is a third distinct line. It is not cyclically induced OSKM in transgenic mice, nor AAV-OSK controlled by external dosing and a doxycycline switch — it hands control to the cell’s own endogenous promoter. So the results below cannot be read directly as a preview of what ER-100’s pathway will produce.
The findings: in HGPS progeroid mice, pro-inflammatory cytokine expression declined and lifespan extended. In naturally aging wild-type mice, aging phenotypes were delayed, lifespan extended, and observed tumor incidence was not altered. Intradermal injection improved wound healing in aged mice.
“Did not alter observed tumor incidence” should be read carefully. That sentence covers tumors only. Other toxicities, rare events, and longer-term risks are outside its scope.
Even so, this remains a relatively direct piece of lifespan evidence in wild-type mice. It moves the question from “can you make cells younger” to “can you make the cells that should be younger go first.”
Between Mice and Humans, the Bridge Was Built by the Company
Between mice and humans, there is one more stop: non-human primates. That stop matters especially for ER-100, because primate eyes are anatomically and functionally much closer to human eyes than mouse eyes are.
Life Biosciences presented this data at the 2024 American Academy of Ophthalmology annual meeting. The approach was to induce NAION-like injury in monkeys, administer a single intravitreal injection of ER-100, and give daily systemic doxycycline. Immunohistochemistry confirmed transcription factor expression in cells surrounding the fovea. Results showed that ER-100 attenuated deficits in pattern electroretinogram responses and axon density, with both prophylactic and rescue dosing timing tested.
The evidence level of this dataset deserves attention. It was presented by the company at a conference, not published as a peer-reviewed paper subject to independent verification. The difference between a conference abstract and a publication is not whether the content is reasonable — it is whether anyone else can examine it. The bridge from mice to humans here was built by the company; outside observers cannot yet walk across it and look for themselves.
Stopping Midway Is Itself a Risk
On the safety side, most writing stops at Abad 2013 and teratomas. That is not the most directly relevant risk.
The 2014 experiment by Ohnishi et al. in Cell addressed “partial” more precisely. They used doxycycline to control reprogramming factor expression in living mice, then withdrew the drug after brief expression. After withdrawal, tumors appeared in multiple tissues, composed of undifferentiated dysplastic cells and accompanied by global DNA methylation changes. The kidney tumors shared several characteristics with Wilms tumor.
The most significant step came next. The team derived iPSCs from those kidney tumor cells and reintroduced them into mice, where they could generate non-tumorigenic normal kidney cells. This result showed that the tumorigenic state was not irreversibly locked in, and supports the conclusion that epigenetic dysregulation alone is sufficient to drive tumor formation — though it does not prove that those cells lacked genetic mutations entirely.
The scope of this finding needs to be stated clearly. Ohnishi used doxycycline-inducible OSKM transgenic mice — the first line. What it establishes is the mechanistic risk that incomplete reprogramming can produce a tumorigenic state. It cannot be used to quantify the actual tumor risk of AAV-OSK or ER-100.
Even within that scope, what it says still holds: reversibility does not guarantee safety. Duration, dose, factor combination, and cellular identity each require separate verification. None of them can be waved through with “epigenetic changes are reversible anyway.”
More recent safety signals sit on the same line. A 2026 paper in Molecules and Cells reported that prolonged systemic OSKM induction caused early death via hepatocyte dedifferentiation and oxidative-stress-driven liver failure, not tumor formation. The same paper found that female mice survived significantly better than males, associated with a stronger antioxidant response. Again, this is OSKM transgenic mouse data — it does not transfer directly to ER-100. But it points at the same question: how long, and where to stop.
There is also a more granular version of the problem. Roux et al. in 2022 (Cell Systems) used single-cell genomics to trace partial reprogramming trajectories and found that as cells recovered a younger expression profile, the somatic identity program was transiently suppressed. A clock reading becoming younger, and a cell still knowing what it is supposed to do — these are two things that need to be verified separately.
Yang 2023: A Dispute Still Open
Any serious discussion of OSK has to pass through the 2023 Yang et al. Cell paper, and through the controversy it generated.
That paper proposed that epigenetic information loss is a cause of mammalian aging, using ICE mice: the I-PpoI endonuclease was used to induce DNA double-strand breaks. The authors argued that repair was faithful and sequences were unchanged, so the accelerated aging observed afterward resulted from epigenetic information loss rather than genetic damage. They also reported that OSK could reverse this process.
In February 2024, Timmons and Brenner published a formal Matters Arising in Cell. Their charges: I-PpoI had been shown to cause cell death and p53 induction in Sinclair’s own earlier papers, which were not cited; the paper did not provide dense time-series data from the first month after tamoxifen withdrawal — precisely the window for examining p53 induction, cell death, and cellular elimination; the graphical abstract showed functional rejuvenation by OSK without corresponding data in the main text; and the authors had not supplied requested raw data. Their alternative interpretation: cell death and tissue damage alone are sufficient to explain the observed aging phenotypes.
In the same issue, Yang et al. published a formal response.
A correction also appeared in the same issue. All three documents ran in the February 29, 2024 issue. The correction addressed methods and citations, without changing the reported results. It described how the team switched from intraperitoneal injection to dietary tamoxifen administration, specifically to limit I-PpoI expression and avoid genotoxic stress. It elaborated on the temporal and spatial control design for I-PpoI. It added three previously uncited relevant papers. It apologized for the omission of these methodological details.
Reading all three documents together shows what an ongoing scientific dispute looks like in practice. The charges are specific. The responses are specific. The correction fills in things that were genuinely missing. But filling in the methods does not automatically make the claim that “sequences were unchanged” a settled conclusion.
For readers, the significance of this is not who is right. It is that “epigenetic information loss is a cause of aging” remains a formally contested claim, not a field consensus. The partial reprogramming field has several competing mechanistic frameworks — cellular identity, chromatin state, transcriptional networks, damage response — but Sinclair and Lu’s research line interprets OSK’s mechanism substantially through this claim. ER-100 follows that line.
The Clock Runs Back. Then What?
The phrase “epigenetic clock” is used so casually that it has started to sound like a thermometer.
It is not. For one thing, it is not a single instrument. First-generation clocks (Horvath 2013, Hannum 2013) are predictive models trained with chronological age as the target variable. Second-generation clocks (PhenoAge 2018, GrimAge 2019) incorporate mortality and clinical health phenotypes. Third-generation clocks (DunedinPACE 2022) estimate the rate of aging. These three generations are asking different questions. Getting inconsistent readings from the same sample across different clocks is normal; no clock is broken.
Second, a clock reading is not evidence of function. Browder 2022 showed that clock responses vary by tissue: kidney and skin declined, other tested tissues did not. Claiming that function also recovered requires measuring function in those same tissues separately — the clock reading cannot do that work.
Third, the most practical point: as of September 2026, searching the FDA’s publicly available qualified biomarkers and surrogate endpoints resources, I find no epigenetic clock that has been formally qualified as a surrogate endpoint sufficient to support approval. Clocks typically appear in trials only as exploratory or secondary measures. To add a qualification: not having completed formal qualification does not mean a given marker will definitely be rejected in a specific product application — that is a separate pathway. But a regulator being willing to let you measure something is not the same as being willing to let you exchange it for approval.
What the First Human Trial Is Actually Testing
Back to NCT07290244.
This is a Phase 1 trial, non-randomized, open-label, with no parallel control group, enrolling up to eighteen participants: twelve in the glaucoma cohort, six in the NAION cohort. The glaucoma cohort proceeds through dose escalation, beginning with one sentinel participant per dose level, with the safety review committee evaluating before enrollment continues. The low dose is 2×10¹¹ vg per eye; the high dose is 6×10¹¹ vg. After the glaucoma cohort completes, one dose is selected for the NAION cohort. Administration is intravitreal injection of the AAV vector into one eye, followed by eight weeks of oral doxycycline to induce OSK expression.
The endpoint structure is what clarifies the picture.
The registration lists twenty-two primary endpoints, all safety: incidence of treatment-emergent adverse events, dose-limiting toxicity, twelve safety laboratory assessments, and eight ophthalmic assessments. Those eight include best-corrected visual acuity, Humphrey visual field, pattern electroretinogram, contrast sensitivity, optical coherence tomography measurements of ganglion cell layer and retinal nerve fiber layer thickness, intraocular pressure, and slit-lamp examination.
Seven of those eight functional or structural assessments also appear in the secondary endpoints. The difference is in the label appended to each. In the primary endpoints, the registered name ends in Safety. The same assessment listed in the short-term secondary endpoints ends in Efficacy. The slit-lamp examination carries only a safety designation in both locations.
The same test, placed in the primary endpoint, is there to confirm the eye was not harmed. Placed in the secondary endpoint, it is there to see whether things improved. This is the registration’s own labeling — the interpretation is not mine.
The secondary endpoints are not all efficacy, either. They also include long-term safety follow-up through year five, AAV2 neutralizing antibodies, cellular immune responses, viral shedding, and aqueous humor vector DNA distribution. These are standard equipment for a gene therapy trial. They are also not measuring aging.
Figure 3: All twenty-two primary endpoints are labeled as safety. Seven of the ophthalmic assessments only receive an Efficacy label when they appear in the secondary endpoints.
The trial started on March 2, 2026. The primary completion date is estimated as May 2027; the full study completion is estimated as March 2032. The registration was last updated May 19, 2026. That “June 9 first participant dosed” date deserves careful reading: it is the date the press release was issued, and the release states that the first participant “has completed dosing.” It establishes that dosing had occurred by that date; it does not establish that dosing happened on that specific day.
As for whether any other partial reprogramming therapy has entered human participants: based on publicly available trial registrations and company announcements through September 2026, I have not found a second one. YouthBio’s YB002 uses all four OSKM factors with an integration-deficient lentiviral vector targeting neurons, with Alzheimer’s disease as the intended indication, currently at the pre-IND stage with the company stating it expects to enter human trials within approximately three years. I am not writing “the only one” here, because trial registries across jurisdictions, investigator-initiated trials, and undisclosed dosing cannot all be captured in a single search.
Where the Judgment Lands
Putting the above together, here is what can be said.
Partial reprogramming is a field with real animal evidence. Sahu 2024’s approach of using an endogenous promoter to target senescent cells specifically moves the question from “can you make cells younger” to “can you make the right cells go first.”
At the same time, the evidence is inconsistent across factor combinations, delivery methods, and tissue responses. The human trial treats two specific optic nerve diseases, with safety as the primary endpoint, in at most eighteen participants, one eye each, with no control group. Primary completion is estimated for May 2027; when results will be published is a separate question.
“Humans have begun reversing their age” skips three levels at once: the indication is not aging, the endpoints are not efficacy, and the scale cannot support any population-level conclusion.
Skipping those three levels requires no dishonesty. It only requires that, in the retelling, “+109% remaining lifespan” stays while “+7% total lifespan” gets dropped; “clock declined in kidney and skin” stays while “other tissues did not” gets dropped; “first human trial” stays while “treating glaucoma” gets dropped.
Up to eighteen participants will produce the first human safety data on ER-100 in these two optic nerve diseases, at this dose range, via this delivery method. That is the question this trial is designed to answer. It is not yet the question everyone is waiting for.
💬 Comments
Loading...