When a salamander loses a leg, the stump does not simply heal over. Within days, cells at the wound site begin dividing, forming a mound of tissue called a blastema that somehow knows whether to build an elbow or a set of fingers. For decades, biologists called this ability “positional memory” without being able to explain how it worked at the molecular level. A cluster of studies published in 2025, now being built upon in ongoing research as of June 2026, has finally pinned that memory to specific genes, chemical gradients, and even the physical packaging of DNA inside cells.
The results, drawn from experiments on axolotls and zebrafish using CRISPR gene editing, chromatin profiling, and targeted drug inhibitors, offer the most detailed picture yet of how regeneration is genetically orchestrated. They also make painfully clear how far medicine remains from harnessing any of it for people.
The molecular address system inside a salamander’s leg
The centerpiece finding came from a team led by Tobey Trofka and colleagues, whose paper in Nature in 2025 showed that a transcription factor called Hand2 acts as a rear-facing compass needle in regenerating axolotl limbs. Using lineage tracing and CRISPR-based knockouts, the researchers demonstrated that connective tissue cells on the posterior (back) side of the limb ramp up Hand2 expression after amputation, which in turn switches on the signaling molecule Sonic hedgehog (Shh). Shh then patterns the new limb from front to back, ensuring that a thumb-equivalent ends up on one side and a pinky-equivalent on the other.
“It’s essentially a ZIP code system,” is how regeneration biologists have described the concept. Each cell carries molecular markers that tell it where it sits along the limb’s axes, and those markers persist even after the limb is severed. The Hand2 study provided the first direct functional proof that a single transcription factor encodes one of those coordinates and controls the downstream signals that execute the building plan.
How cells know shoulder from fingertip
A separate axis of identity, running from shoulder to fingertip (the proximodistal axis), turns out to depend on how cells metabolize retinoic acid, a derivative of vitamin A. A 2025 study published in Nature Communications showed that the enzyme CYP26B1 breaks down retinoic acid in distal (fingertip-level) blastemas, keeping concentrations low. When researchers blocked CYP26B1 with the drug talarozole, distal blastemas were reprogrammed toward proximal identity. Cells that should have built fingers began behaving as though they were rebuilding a shoulder. Bulk RNA sequencing data, deposited in the NCBI Gene Expression Omnibus under accession GSE272731, confirmed these gene-expression shifts across multiple amputation levels and drug doses.
This finding dovetails with earlier work on a gene called Tig1 (also known as Rarres1), a retinoic-acid-responsive factor that is graded along the proximodistal axis. A study published in Nature Communications showed that Tig1 can reprogram blastema cells toward proximal fates, meeting functional criteria for a proximal determinant. Together, the CYP26B1 and Tig1 results describe a chemical gradient system: high retinoic acid signaling near the body means “build something proximal,” while rapid breakdown of retinoic acid at the limb tip means “build something distal.”
Instructions written in DNA packaging
Genes alone do not tell the whole story. A study in Developmental Cell used genome-wide chromatin profiling across different axolotl limb segments and found that a histone modification called H3K27me3 varies systematically at homeoprotein gene loci depending on position. In plainer terms, the instructions for “build a wrist” versus “build an elbow” are written not only in which genes are switched on but in how tightly the DNA is wound around its protein spools in connective tissue cells. Loosen the packaging at one set of genes and you get wrist; tighten it and you get elbow.
This epigenetic layer of positional memory may explain why connective tissue, rather than muscle or nerve, appears to be the primary carrier of location information in regenerating limbs. It also raises a pointed question for mammalian biology: do human connective tissue cells retain any of these position-specific chromatin marks after injury, or do they lose them as scar tissue forms?
Zebrafish reveal what the wound surface must do
While axolotl studies have focused on the internal blastema, zebrafish research has clarified what happens at the wound’s outer layer. A study in PLOS Genetics established that fin regeneration requires a specialized wound epidermis and identified the transient appearance of the protein laminin beta 1a as a critical early step. Without this extracellular matrix component showing up at the right moment, the wound epidermis cannot relay the signals that initiate regrowth.
Cross-species work has strengthened the case that some of these wound-surface mechanisms are shared. A 2025 paper in PNAS reported that SP-family transcription factors are active in the regenerating epidermis of both axolotls and zebrafish, suggesting a conserved regulatory module that helps organize new tissue growth. The conservation is intriguing because SP-family genes also exist in mammals, though no one has yet shown they can be activated therapeutically in a species that does not naturally regenerate.
The enormous gap between salamander and human
A News and Views commentary in Nature, published alongside the Hand2 findings, offered a measured assessment of what these discoveries do and do not mean for medicine. The commentary noted that while the 2025 studies clarified the molecular basis of positional memory with unprecedented precision, no experiments have tested whether activating Hand2, Shh, or CYP26B1 pathways in mammalian wound tissue can trigger blastema formation. The same genetic switches, the authors cautioned, may behave very differently in tissue that is primed for scarring rather than regeneration.
That scarring response is the central obstacle. When mammals suffer major tissue loss, fibroblasts flood the wound and lay down collagen in a dense, disorganized mat. This fibrotic reaction seals the injury quickly, an evolutionary advantage for avoiding infection, but it walls off the very cells that would need to de-differentiate and communicate in order to form a blastema. None of the 2025 papers addressed how to overcome this barrier.
Readers familiar with regeneration headlines may recall the 2022 work by Murugan and colleagues, published in Science Advances, in which a wearable bioreactor delivering a cocktail of five drugs triggered partial leg regrowth in African clawed frogs (Xenopus). That study demonstrated that some regenerative capacity can be unlocked in an animal that does not normally regrow limbs, but the resulting structures were imperfect and the frogs are still far closer to salamanders on the evolutionary tree than humans are. The gap remains vast.
Safety adds another layer of complexity. Retinoic acid signaling is active throughout the human body, influencing embryonic development, immune function, and reproductive tissues. Drugs like talarozole that modulate this pathway carry risks of off-target effects, and no regulatory assessment of CYP26B1 inhibition for regenerative purposes in humans has been reported. Even local delivery to a wound site could alter neighboring cells in unpredictable ways.
What would real progress look like
The 2025 findings rest on rigorous primary experiments, but they were conducted in genetically tractable laboratory animals raised under tightly controlled conditions. Axolotls in research facilities are inbred strains housed in filtered water at stable temperatures. Zebrafish fin amputations are performed on healthy adults in standardized tanks. Human injuries happen in chaotic settings and are complicated by age, infection, diabetes, vascular disease, and medications that alter healing.
For non-specialist readers trying to evaluate future headlines, a few benchmarks are worth keeping in mind. Genuine progress toward human limb regeneration would involve direct manipulation of the pathways identified in axolotls and zebrafish, not just observation of gene expression. It would demonstrate true structural regeneration, meaning bone, muscle, nerve, and vasculature organized into a functional architecture, not just a lump of new tissue. And it would include rigorous long-term assessment of both function and safety in a mammalian model before any talk of clinical trials.
None of that has happened yet. What has happened is that positional memory, once a vague and almost mystical concept in developmental biology, now has a molecular parts list. Scientists can point to Hand2 as a posterior address marker, to retinoic acid gradients as a proximal-distal coordinate system, to H3K27me3 chromatin marks as segment-specific instructions, and to conserved wound epidermis programs that may represent ancient regulatory circuits shared across vertebrates.
That parts list is not a therapy. But it is the kind of foundation on which therapies are eventually built. The challenge ahead is to determine which components of this machinery can be reactivated in mammals without triggering tumors, fibrosis, or developmental havoc, a project that will take years of careful work in mouse and primate models. For now, the axolotl and the zebrafish have given scientists something they did not have before: a clear set of molecular targets and a map detailed enough to start asking the right questions about why humans heal with scars instead of new limbs.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
