Amplifica raised $26 million with Eli Lilly participating, and AMP-303 is being called the hairy mole drug. Its own patent says the study compound is 0.3% hyaluronic acid. Here is what the Phase 1 data shows, what it leaves out, and why the molecular weight matters.
On 12 August 2026, Amplifica closed a $26 million Series B led by Tasso Partners, with Eli Lilly among the participants. Coverage of the round leaned on the company's origin story: a 2023 Nature paper explaining why some moles sprout thick hair, and a signalling protein called osteopontin that turned out to be driving it.
That paper is good science. It is also not what Amplifica has been injecting into people.
The short answer: AMP-303 is hyaluronic acid. Amplifica's own patent application describes the study compound as 0.3% non-crosslinked hyaluronic acid in phosphate buffered saline, with an average molecular weight of up to about 100 kDa. Osteopontin, the molecule from the hairy-mole Nature paper, is a separate preclinical program (AMP-203) that no person has received. The Phase 1 result being quoted, a 15% increase in hair count, is a mean among undefined "responders" inside one of two subgroups, measured against each man's own baseline rather than against placebo, in a trial that enrolled only mild-to-moderate hair loss and was never registered on any clinical trial registry.
AMP-303 is an intradermal injection being developed by Amplifica Holdings Group for androgenetic alopecia. It is investigational, it is not approved, and it is not available.
For two years the composition was described publicly only as "a novel polysaccharide," which is the phrasing used in the company's 2025 conference abstract [1]. That wording is technically accurate and practically uninformative, since it covers everything from cellulose to heparin.
Amplifica's patent filings resolve it. WO2024211282A1, Example 1, states plainly:
"The study compound is hyaluronic acid (also herein AMP-303) for intradermal injection at a concentration of 0.3% ±0.1%."
and
"AMP-303 for intradermal injection consists of a non-crosslinked hyaluronic acid and phosphate buffered saline."
Claim 1 of the same filing caps the average molecular weight at "up to about 100 kDa," and the specification describes the material as low average molecular weight, with one embodiment at about 60 kDa [2].
So AMP-303 is dilute, low-molecular-weight, non-crosslinked hyaluronic acid, injected into the scalp. That is a meaningful thing to know before evaluating anything else about it.
No. This is the most common misconception about the program, and Amplifica's own pipeline contradicts it.
Osteopontin is designated AMP-203 and sits in preclinical development. AMP-601 is SCUBE3, licensed from UC Irvine. AMP-303, the only candidate that has been in humans, is the hyaluronic acid formulation.
The confusion is understandable. Amplifica's earlier granted patent, US 11,337,993 B2, does mention osteopontin in its abstract: "Compositions for stimulating hair growth include two or more of hyaluronic acid, osteopontin, and another CD44 binding ligand." But independent claim 1 of that patent requires only hyaluronic acid at 1 to 250 µg/mL. Osteopontin enters solely through dependent claims 16, 17, 28 and 29, as an optional addition [3]. And in the later filing that actually describes what went into trial subjects, the word osteopontin appears in none of the 34 claims.
The shared thread between the two molecules is the receptor. Both hyaluronic acid and osteopontin bind CD44. That is a real biochemical relationship, and it is also the entire bridge between the Nature paper and the product.
The underlying work is Wang et al., Nature 2023 [4]. Melanocytic nevi that grow terminal hair contain senescent pigment cells, and those cells secrete osteopontin, which acts on CD44 on hair follicle epithelial cells and drives follicles out of quiescence into anagen.
The genetic evidence is the strong part. Knocking out osteopontin abolishes the precocious anagen phenotype. Knocking out CD44 does the same. Conditional delivery of osteopontin reproduces it. Osteopontin knockout also reduces wound-induced hair neogenesis, which suggests a role there too.
Two caveats deserve more attention than they get.
First, in that same paper, loss of CD44 alone does not compromise bulge stem cell properties. Percentage, EdU labelling and colony-forming potential are all unchanged. CD44 appears permissive rather than instructive for normal hair cycling.
Second, and more important for androgenetic alopecia: hairy nevi are body hair. Body follicles and scalp follicles respond to the same hormones in opposite directions. DHT is trichogenic on the body, converting vellus to terminal hair, while on the scalp it does the reverse. Estradiol is the inverse again: it is the most potent known stimulator of vellus-to-terminal conversion in frontal follicles, yet it retards body hair growth [5].
Whatever drives ectopic hair growth on someone's back has to earn its place separately on a balding hairline. It does not transfer for free.
There is a further wrinkle. In male frontotemporal human follicles treated with 17β-estradiol ex vivo, osteopontin is the second most downregulated gene. That could be negative feedback from upregulation of its CD44 receptor. But if so, it also implies that supplying more of the ligand is insufficient.
There is one paper that makes a serious mechanistic case, and it is worth reading carefully because it is also where AMP-303's biggest problem comes from.
Liu et al. (2026) report that medium-molecular-weight hyaluronic acid promotes hair growth through CD44-mediated ROS induction, phosphorylation of AKT, and subsequent stabilisation of β-catenin [6]. AKT stabilises β-catenin by inactivating GSK-3, and facilitates nuclear entry by direct phosphorylation at Ser552 [7]. Pretreatment with NAC prevents both the ROS rise and the β-catenin rise, indicating that β-catenin upregulation depends on ROS. AKT inhibition abrogates ROS generation.
All molecular weights of HA stimulated VEGF and inhibited TGF-β1 in vitro and in vivo. Medium and high molecular weight also increased alkaline phosphatase and dermal papilla cell migration.
Three honest limits on that enthusiasm.
The Wnt effect is narrower than it sounds. GSK-3-mediated regulation of β-catenin is largely AKT-independent, so this route is more limited than canonical Wnt activation and insufficient to activate LEF1 without Wnt ligands [8].
The AKT arm may already be maxed out in bald scalp. Phosphorylated AKT and SGK1 are upregulated in balding dermal papilla cells, possibly compensating for Wnt downregulation. It fails to fully compensate, and may worsen dermal papilla cell stress [9].
And the functional experiments were done in dermal papilla cells, while the anagen-inducing site for osteopontin-CD44 in vivo is epithelial.
Yes, and this is the crux.
In Liu et al., the optimum for hair growth is 268.1 kDa. Medium-molecular-weight HA produced the most potent modulation of every growth factor measured and the most robust hair growth in mice: earlier anagen induction, greater length, largest diameter. The amount of hair growth correlated with β-catenin expression in mouse skin. Low-molecular-weight fractions performed worst.
AMP-303 is capped by its own patent claim at 100 kDa, with the specification calling it low average molecular weight and one embodiment at about 60 kDa. That is roughly four times below the published optimum, in the worst-performing band.
One caveat on the mouse data, in fairness to no one in particular: measurements were taken at day 14, when treated follicles were in full anagen VI while controls had only just begun entering anagen. The diameter difference may therefore be a timing artifact rather than a real thickening effect.
Less than a day.
Hyaluronan turns over rapidly. Half the body's hyaluronic acid is in the skin, and its half-life there is under 24 hours [10]. Crosslinking is what gives dermal fillers their months-long residence, and AMP-303 is explicitly non-crosslinked.
This creates a timing problem that has nothing to do with trial design. Amplifica reports its peak effect at 60 days after a single treatment cycle. By then roughly sixty half-lives have elapsed. Nothing injected remains in the tissue.
That leaves two possibilities. Either one exposure flipped a durable switch, or the day-60 measurement reflects something other than the drug.
The reported shape argues against the switch. Human scalp anagen lasts two to six years, so a follicle genuinely pushed into growth should stay there. The AMP-303 effect peaks at 60 days and is smaller by 150 days, where follow-up ends. A decaying curve is what a transient response looks like.
It is also worth noting that claim 1 of Amplifica's own filing claims administration "for at least 2 times a week," and the protocol in that document doses on days 0, 1 and 2. The "single treatment" framing is doing some work.
In 2024 Amplifica announced Phase 1 data in which some subjects showed a greater than 15% increase in terminal hair count after one session [11]. This is the number that circulates.
It has several layers of selection built into it.
The threshold is relative to baseline, not to placebo. A similar number of placebo subjects could have landed just below the same cutoff. Comparing responders in the treatment group to responders in the placebo group would be the meaningful comparison, and it was not reported.
Responder is never defined, and the count is never given. We do not know how many responders there were in either arm, or what baseline density they started from. The lower the baseline, the less a 14.5% change means in absolute hairs.
Singling out responders suggests the whole group missed. If the treatment population as a whole had separated from placebo, that is the number a sponsor reports. The conference abstract's own word for the full-population result is "a trend."
Two different 15% figures are quoted interchangeably. The press release describes a statistically significant proportion of subjects clearing a 15% threshold (P<0.001). The abstract describes a mean of 14.5% among Group 1 responders (p<0.005). These are different quantities with different tests.
Severity was capped at entry. Enrolment was limited to men aged 18 to 45 with mild-to-moderate androgenetic alopecia. Nobody with advanced loss was studied.
Half the trial has never been reported. Subjects were stratified by duration: Group 1 at 3 to 5 years (n=32) and Group 2 at 10 or more years (n=29). Every efficacy figure released comes from Group 1. Fifteen months after presentation, Group 2 has no published result.
A cohort disappeared. A third group of post-transplant patients was described when the trial opened in 2023 and is absent from the published abstract.
No absolute hair counts exist. Every efficacy figure is a percentage. Hair was counted from a single 1 cm² target area per side, inside an injection field roughly ten times that size.
Because AMP-303 was given as a split-scalp study: drug into one side of the frontotemporal scalp, saline into the other side of the same head, roughly 20 injections per side.
Injected saline is not an inert control in a scalp.
In a split-scalp trial of platelet-rich plasma, the saline half gained 15.68 hairs/cm², a significant within-arm change, while the difference between PRP and saline was not significant [12]. In a triple-blind, randomised, split-scalp trial of botulinum toxin, the saline half gained 7.9 hairs/cm² and was the only side in the study with a significant change [13].
This is a measurable design effect, not an anecdote. In a systematic review of platelet-rich plasma trials, placebo-arm hair density at six months was about 37 hairs/cm² higher in split-scalp studies than in whole-head studies [14]. For contrast, the placebo arm in the finasteride registration trials lost hair, which is what untreated androgenetic alopecia does.
The single number that would resolve this for AMP-303 is the saline side's own change from baseline. It has never been published.
Two arguments that sound persuasive here should be avoided, because they do not hold. Carry-across between scalp halves biases toward the null and cannot manufacture a positive result. And wound-induced hair neogenesis is not demonstrated in humans at all; the mouse threshold requires wounds several centimetres square, orders of magnitude beyond a needle puncture.
No. Searches of ClinicalTrials.gov, the WHO ICTRP, EU CTIS, EudraCT and ISRCTN return no Amplifica records.
A first-in-human safety study is generally not an "applicable clinical trial" under FDAAA 801, so this is permissible rather than a violation. The consequence is still real: there is no pre-specified endpoint or analysis plan against which the reported result can be checked, and there is no peer-reviewed publication of the data. The entire clinical record is one conference late-breaker and a set of press releases.
Yes, and it is the most relevant precedent nobody cites.
The Swedish company Follicum developed FOL-005, an osteopontin fragment engineered to prevent integrin binding while retaining hair-growth activity, which in preclinical testing showed a stronger effect than native osteopontin [15]. Their clinical results, posted in 2020, were modest: after three months, total hair count increased by 1.08/cm² over placebo, with total anagen hairs up 8.16/cm² over placebo [16].
A larger topical Phase 2a in 209 men followed and missed. The program was discontinued in May 2021, every project was shelved that September, and the company was absorbed by Coegin Pharma and delisted in February 2022.
Two things are worth saying in Amplifica's favour here. FOL-005 was a 15-amino-acid fragment rather than the intact protein, so it was never a clean test of what full-length osteopontin does through CD44. And the trial that killed it was a daily topical cream, which is a maintenance schedule applied to what may be a switch.
It is also worth saying that the only peer-reviewed study of FOL-005 on human scalp tissue found it induced premature catagen via reduced FGF-7, and its authors proposed testing it for unwanted hair growth [17].
AMP-203 is osteopontin, preclinical. Given that AMP-303 is hyaluronic acid and both are CD44 ligands, the pairing of AMP-203 with AMP-506 in the pipeline may indicate a combination product, possibly with a third CD44 ligand such as serglycin or IGFBP4 based on the patent filings.
AMP-601 is SCUBE3. UC Irvine's work shows it is secreted from the dermal papilla and activates TGF-β receptors in epithelial cells, and that TGF-β inhibition abolishes the hair growth effect [18]. Presumably it relieves the SMAD1-mediated quiescence program in follicle stem cells, which would trigger anagen without reversing miniaturisation [19]. Genetic studies do support a causal role for TGF-β in androgenetic alopecia [20], but the driver appears to be overactivation in the dermal papilla. SCUBE3-null mice show only a modest delay in anagen entry.
Since TGF-β is a potent catagen-promoting factor upregulated by DHT, using a TGF-β activator in AGA is a harder sell than hyaluronic acid, which suppresses TGF-β1. SCUBE proteins also chaperone Sonic hedgehog from producing to target cells [21], which makes SCUBE3 a rate-limited and therefore safer way to potentiate Hedgehog than a smoothened agonist. It is also, for the same reason, unlikely to shift pathway activation much in AGA.
No, and there is nothing to wait for yet in any practical sense. There is no registered Phase 2, no published paper, and no absolute efficacy number.
The base rates are unkind. Across all modalities, Phase 1 to approval runs about 7.9%, and Phase 2 to Phase 3 about 28.9%. The androgenetic alopecia graveyard is specifically full of programs with positive-sounding Phase 2 data that never published a Phase 3.
None of that makes Amplifica a bad-faith operation. Maksim Plikus is a serious scientist with an h-index in the fifties, no retractions and no corrections on the founding paper, and the nevus work is a genuinely original way into follicle biology. The team behind him built and sold Bonti to Allergan.
The gap is between the science and the syringe. The osteopontin biology that makes Amplifica interesting is preclinical. What has been in humans is low-molecular-weight hyaluronic acid, at roughly a quarter of the only published optimal molecular weight, with a sub-24-hour tissue half-life, evaluated against saline injected into the other side of the same head, reported as a mean among an undefined subset of the milder half of a mild cohort, in a study nobody registered.
The right response is neither excitement nor dismissal. It is a specific request: publish the saline side's change from baseline, publish Group 2, and register the next trial.
AMP-303 is Amplifica's investigational intradermal injection for androgenetic alopecia. Its own patent application describes the study compound as 0.3% non-crosslinked hyaluronic acid in phosphate buffered saline, with an average molecular weight capped at about 100 kDa. It is not approved and is not available.
No. Osteopontin is a separate Amplifica program designated AMP-203, still preclinical, that no person has received. AMP-303 is hyaluronic acid. Both bind the same receptor, CD44, which is the source of most of the confusion.
Amplifica reported that some subjects showed a greater than 15% increase in hair count at 60 days. That figure is a mean among undefined responders in one of two subgroups, measured against each man's own baseline rather than against placebo. The abstract's own description of the full-population result is 'a trend,' and no absolute hair counts have ever been published.
The one study linking hyaluronic acid to hair growth through CD44 found an optimum at 268.1 kDa, with low-molecular-weight fractions performing worst (Liu 2026). AMP-303's patent caps its average molecular weight at about 100 kDa and describes it as low molecular weight, roughly four times below that optimum.
Its half-life in skin is less than one day, and AMP-303 is explicitly non-crosslinked, which is the fastest-clearing form. Crosslinking is what makes dermal fillers last months. Amplifica reports its peak effect at day 60, by which point nothing injected remains in the tissue.
No. Searches of ClinicalTrials.gov, the WHO ICTRP, EU CTIS, EudraCT and ISRCTN return no Amplifica records. A first-in-human safety study is generally not an applicable clinical trial under FDAAA 801, so this is permissible, but it means no pre-specified endpoint or analysis plan can be inspected.
Yes. Follicum's FOL-005 was an engineered osteopontin fragment. Its posted results showed about 1.08 more hairs per cm2 than placebo, a larger topical Phase 2a in 209 men missed, the program was discontinued in 2021 and the company delisted in 2022.
Not automatically. Hairy nevi are body hair, and body and scalp follicles respond to hormones in opposite directions: DHT thickens body hair while thinning the scalp, and estradiol does the reverse. A signal that drives ectopic body hair growth has to be demonstrated separately in balding scalp.
This article expands on an analysis first published by Papa Smurf in April 2026. AMP-303 is an investigational product that has not been approved by the FDA for any use. Nothing here is medical advice.