Repair Has a Cell of Origin

Cancer biology spent a decade learning that a tumor's behavior is set largely by which healthy cell it started from and how far along its developmental path that cell had traveled before it stalled — the so-called cell of origin. In medulloblastoma, for instance, the developmental stage a cell had reached before maturation arrest directly governs how heterogeneous and aggressive the tumor becomes Sonic Hedgehog medulloblastoma subtypes driven by developmental stage at the time of maturation arrest, a finding corroborated by Developmental basis of SHH medulloblastoma heterogeneity, and in ependymoma, malignant cells were shown to hijack specific fetal developmental programs by computationally "docking" tumor profiles onto a healthy developmental atlas A spatiotemporal single-cell atlas of the developing human spinal cord. Regenerative medicine almost never asks the mirror-image question: after an injury, the repair response is also carried out by descendants of particular cells, and the identity of those founding cells should determine whether tissue is rebuilt or replaced by scar.

Nowhere does this matter more than the heart, the textbook case of limited regeneration in adult mammals: after injury it heals by scarring rather than muscle regrowth, and it does so progressively worse with age. We propose that this age-dependent failure is, at least in part, an origin shift. Aging does not simply shrink the pool of repair-competent cells; it changes which cells in the lineage hierarchy launch the repair, biasing the launch point toward cells that have already entered a mesenchymal-transition program — EMT, the gene-expression switch by which adherent epithelium-like cells lose their contacts and become motile, matrix-secreting cells, a program that in the heart also runs in epicardial and endothelial variants and is the cellular engine of scarring. If this is right, then transient EMT inhibition early after injury should redirect repair back toward a regenerative origin — but only in hearts that still contain a functional regenerative subpopulation, which the aging niche may already have silently lost.

Core Claim

The hypothesis has three falsifiable parts. First, the founding cells of the dominant post-injury repair clones shift position along the developmental trajectory as a function of age: in young animals they dock near regenerative progenitors; in aged animals they dock near mesenchymal-transitioned, senescence-prone states. Second, transient inhibition of the mesenchymal-transition program during the early post-injury window reduces this shift in young and middle-aged hearts. Third, in aged hearts the same inhibition fails or backfires, because the alternative origin pool is depleted — and prior rejuvenation can restore redirectability.

To make the first part precise, suppose the atlas-docking step assigns every recovered repair clone $c$ a position $s(c)$ on the developmental trajectory of a reference atlas (a pseudotime-like coordinate where low means early progenitor and high means mature or transitioned), and let $w(c)$ be that clone's share of the repair tissue. The origin shift at age $a$ is then

$$\Delta(a) \;=\; \sum_c w(c)\, s(c) \;\Big|_{\text{age } a} \;-\; \sum_c w(c)\, s(c) \Big|_{\text{young}} ,$$

and the claim is that $\Delta(a)$ grows monotonically with age, that clone-level aging scores track it, and that EMT inhibition within a defined early window drives $\Delta$ back down where, and only where, the regenerative subpopulation survives.

The proposal deliberately treats the four source directions asymmetrically. Cell-of-origin identification is the methodological core, transplanted from cancer to repair. Age-related stem cell dysfunction is the driving variable and the constraint — it determines whether redirection is even possible. Limited cardiac regeneration is the application regime. EMT inhibition is the intervention lever.

Research Program

The first component builds the coordinate system: a single-cell, multimodal reference atlas of the cardiac stromal, epicardial, and endothelial lineage hierarchy, sampled across the age span over which regenerative capacity regresses, from the early postnatal window through old age. This follows the docking methodology established in the cancer-origin literature — A spatiotemporal single-cell atlas of the developing human spinal cord built exactly this kind of healthy developmental coordinate system to locate tumor origins, and Sonic Hedgehog medulloblastoma subtypes driven by developmental stage at the time of maturation arrest showed that position along such a trajectory is the variable that governs outcome — but extends it with the age dimension those studies lack, and with chromatin measurements alongside transcriptomes.

The chromatin addition is not optional decoration. Lineage tracing on transcriptional landscapes links state to fate during differentiation used expressed DNA barcodes to show that a neural network predicting a progenitor's fate from its transcriptome alone was 26% less accurate than predicting it from a sister cell's observed fate — proof that decisive information is hidden from RNA sequencing. And Pseudotime Trajectory Inference from Single-Cell Snapshots reminds us that trajectory orderings reconstructed from destroyed cells are inferences, not observations. So the second component is longitudinal origin identification by heritable genetic barcoding, following the LARRY approach of that same lineage-tracing work, delivered to the progenitor compartment before injury so that the founding cells of every repair clone are permanently marked; Genetic Lineage Tracing for Mapping Cellular Ancestry motivates this because plastic and transitioning cells shed their canonical surface markers. Each recovered clone then gets three annotations: its barcode-defined family size, its docking coordinate $s(c)$ on the atlas, and an aging score computed from the 330 Global Aging Genes identified in Mouse aging cell atlas analysis reveals global and cell type-specific aging signatures, which analyzed more than 110,000 cells across 23 tissues and 76 tissue-cell types from mice aged 3 to 24 months.

The third component is the intervention. Age-series injury experiments with four arms — vehicle, transient EMT inhibition in the early post-injury window, pre-injury rejuvenation, and the combination — with the origin map as the primary readout and tissue-level endpoints secondary. The rejuvenation arm uses the microRNA miR-145-5p, which declines with age MicroRNA-145-5p in aging stem cells and whose re-expression restored proliferation and migration in aged adipose-derived stem cells and accelerated cutaneous wound closure in MiR-145-5p overexpression rejuvenates aged adipose stem cells and accelerates wound healing. Before committing to the heart, we run the whole pipeline as a pilot in skin wound healing — the one system where the file's rejuvenation result was actually obtained — to calibrate dosing, timing, and the barcode readout, then port it to the cardiac injury model.

Baselines

The "do nothing" reference point is spontaneous repair under vehicle, established in our own young-adult animals: scar fraction, shortening fraction, and the origin-shift statistic $\Delta$. No reported result found on Lacuna for a quantified spontaneous-repair benchmark in cardiac injury, so this arm must generate its own reference numbers, and every intervention is judged against it in the same apparatus, same age, same injury size.

The strongest existing method for the origin-identification component is transcriptome-only atlas docking — the established approach of the cancer-origin literature, as practiced in A spatiotemporal single-cell atlas of the developing human spinal cord and Sonic Hedgehog medulloblastoma subtypes driven by developmental stage at the time of maturation arrest. It is the strongest competitor because it is the field's default and it requires no barcoding infrastructure. To make the comparison fair, we run it on exactly the same reference atlas, the same recovered cells, and the same compute, and evaluate both pipelines on held-out clone fates, where the barcodes supply ground truth. The number to beat comes from Lineage tracing on transcriptional landscapes links state to fate during differentiation: a neural network predicting fate from the transcriptome alone was 26% less accurate than prediction from a sister cell's fate. Our multimodal-plus-barcode pipeline counts as succeeding only if it shrinks that accuracy gap on repair clones; if transcriptome-only docking matches it, the barcoding apparatus was unnecessary and we will say so.

Two ablations isolate the proposal's key ingredients. Docking against a young-only reference atlas — dropping the age dimension — tests whether the age-series atlas is what actually buys the aging prediction, or whether any developmental atlas would do. And the EMT-inhibition-alone arm in old animals isolates niche availability: if redirection fails there while succeeding in middle-aged hearts, the depletion constraint is confirmed; if it succeeds there too, the constraint story is wrong and the simpler intervention wins.

Why not just…?

Why not just count progenitors with surface markers in young versus aged hearts? Because bulk and marker-based counts are known to be blind to exactly the change we care about. In Aging disrupts cell subpopulation dynamics and diminishes the function of mesenchymal stem cells, viability, frequency, and proliferation of adipose-derived stem cells from 3-month-old and 21-month-old mice were indistinguishable at the population level, while single-cell clustering revealed the selective destruction of a highly active pro-vascular subpopulation enriched for Hif1a, Vegfa, and Cxcl12, alongside collapsed antioxidant defenses (Sod2, Sod3). And MicroRNA-21 maintains hematopoietic stem cell homeostasis through sustaining the NF-κB signaling pathway in mice showed that phenotypic stem-cell counts can rise while long-term reconstituting ability falls. Counting cannot see which subpopulation remains, let alone where repair launches from.

Why not just run the EMT-inhibitor trial after injury and measure heart function? This is a real threat, and we do not argue around it: if the only question were whether EMT inhibition helps young adults, that trial alone would settle it, and our design includes it as one arm. But two things make the bare trial insufficient for the aged case, which is the case that matters clinically. First, mesenchymal-transitioned cells also build the vascular support structure of repair — the depleted pro-vascular subpopulation in Aging disrupts cell subpopulation dynamics is defined by Vegfa and Cxcl12, and Immunoregulatory Roles of Tumor‐Originated Pericytes Identified by Single‐Cell Analysis in Glioblastoma shows that stromal cells traced to distinct origins carry distinct, consequential functions — so blanket inhibition in a depleted aged niche could remove the only remaining repair source. Second, no reported result found on Lacuna for cardiac EMT-inhibition outcomes, so a null result in old animals would be uninterpretable without the origin map: one could not tell whether the mechanism is wrong or the launching cells were simply unavailable. The map is what converts a single efficacy number into a rule for who benefits and when.

Why not just rejuvenate with miR-145-5p and skip the origin question? Rejuvenation and redirection are different operations. miR-145-5p overexpression demonstrably restored proliferation and migration in aged adipose stem cells and accelerated wound closure MiR-145-5p overexpression rejuvenates aged adipose stem cells and accelerates wound healing, but if aged repair launches from mesenchymal-transitioned cells, then making those founding cells more proliferative without changing which cells found the repair could enlarge the scar-producing clone pool rather than the regenerative one. The combination arm tests this directly: if rejuvenation without EMT inhibition leaves $\Delta$ unchanged or worsens fibrosis while the combination repairs, redirection — not proliferation — is the operative mechanism.

Proposed Experiments and Decisive Checks

The primary metrics are the origin-shift statistic $\Delta(a)$ across the age series, the Global Aging Gene score of dominant repair clones, and the fraction of repair tissue attributable to regenerative-origin clones, with scar fraction and shortening fraction as tissue-level endpoints. The hypothesis is supported if $\Delta$ rises monotonically with age while clone-level aging scores track it, and if EMT inhibition within the early window shifts the docking distribution of newly founded clones toward earlier coordinates in young and middle-aged hearts. It fails informatively if aged hearts still found repair from regenerative-origin clones yet scar anyway — that would localize the block downstream of origin choice, pointing next at senescence accelerators like the long non-coding RNA SAN, which drives senescence through the miR-143-3p/ADD3 axis Hewei Xiong et al., or at telomere attrition, flagged as a primary determinant of age-related fibrotic disease in Telomere Dysfunction in Idiopathic Pulmonary Fibrosis: A Comprehensive Review. There is an instructive precedent for the aging logic itself: in Defining the marker and developmental trajectory of myeloid-derived suppressor cells in aging by single-cell transcriptomics, aged cells were traced not to arrested immature precursors, the textbook expectation, but to terminally differentiating mature neutrophils and monocytes, with a tenfold surge in CD300c+ cells in aged bone marrow — aging changed the origin, not just the abundance. That is precisely the kind of change that population-level assays miss and that clone-resolved, barcode-grounded docking is built to catch.

The decisive check is a swap experiment: take the age-series origin map, predict which hearts will respond to early EMT inhibition from the pre-injury composition of their progenitor compartment alone, and test those predictions prospectively in animals the map was not built on. If the map's stratification predicts response better than progenitor counts or chronological age, then repair having a cell of origin is not a metaphor but a usable clinical variable — and the cancer field's hardest-won lesson becomes regenerative medicine's most practical one.