The Book of PeptidesThe Book of Peptidesthe painkiller that missed
2004 (first SS-31 publication, out of a 1999-2003 opioid-analgesic program); 2013-2014 (cardiolipin mechanism reframe); 2025 (first approval)
Elamipretide began as a failed painkiller, spent a decade filed in the literature as a free-radical scavenger before the mechanism work reattributed its action to a lipid it binds, and reached the US market in September 2025 on a twelve-patient trial that had missed both of its primary endpoints.
Neither researcher behind the Szeto-Schiller peptides was looking for a mitochondrial drug. Peter Schiller, a peptide chemist in Montreal, built analogs of a polar opioid tetrapeptide; Hazel Szeto, a pharmacologist at Weill Cornell, tested them in animals. Their lead compound, SS-02, behaved in a way it had no business behaving: a molecule with a 3+ net charge should not cross the blood-brain barrier, yet it produced central analgesia in rodents. Tagged with a fluorophore, it turned out to enter cells unaided and pool inside mitochondria. So they rebuilt it to keep the address and drop the drug: SS-31 is the same four residues resequenced to move the dimethyltyrosine off the N-terminus, where the opioid pharmacophore has to sit. Opioid affinity fell away; the targeting survived.
The compound was licensed to what became Stealth BioTherapeutics and reached the clinic as elamipretide. For a decade it was filed everywhere as a mitochondria-targeted antioxidant. Birk and colleagues, in 2013 and 2014, showed instead that it binds cardiolipin, the signature phospholipid of the inner mitochondrial membrane — a targeted lipid interaction rather than the generic radical-scavenging the earlier framing implied. The approved US label calls it a "mitochondrial cardiolipin binder," with no antioxidant claim attached; a 2024 CRISPR screen has since nominated phospholipid scramblase 3 as a further target.
Then came a decade of expensive misses. EMBRACE STEMI, a 300-patient Phase 2a reported in 2016, missed its infarct-size endpoint. MMPOWER-3, the Phase 3 in primary mitochondrial myopathy, enrolled 218 patients and was terminated because, per the registry, its double-blind portion did not meet the primary endpoints. A 2024 post hoc analysis put the effect in one group — patients whose disease arises from nuclear genes governing mtDNA maintenance — where the six-minute walk difference reached 25.2 m against 2.0 m on placebo (p = 0.06), and 37.3 m against -8.0 m (p = 0.0024) in the narrower replisome-plus-CPEO subset. Both are post hoc subgroups of a trial that missed its primary endpoint — hypothesis-generating, not evidence.
The core is real, and serendipity is the researchers' own word for it — but three details get stretched. The accident belongs to the parent compound: the opioid SS-02 reached mitochondria on its own, and SS-31 was <i>deliberately</i> engineered to keep that targeting and delete opioid activity. "Mitochondrial antioxidant" is outdated; the label says cardiolipin binder. And "FDA-approved mitochondrial drug" oversells it: an accelerated approval on an intermediate endpoint, from a twelve-patient trial that missed both primaries.
Elamipretide reached the US market on 19 September 2025 as FORZINITY, an accelerated approval for muscle strength in Barth syndrome at 30 kg and above — a floor tracking TAZPOWER's own eligibility, which leaves the youngest patients outside it. The label carries a benzyl alcohol warning and states that approval rests on knee extensor strength as an intermediate clinical endpoint, the confirmatory trial due in 2029. Nothing here is approved for aging. That interest traces to a 2021 randomized trial in 39 adults aged 60-85 where a single infusion moved muscle ATPmax on percentage change (p = 0.045) though not absolute (p = 0.055), gone by day seven, with no improvement in fatigue resistance. SS-31 meanwhile sells through research-chemical channels, not for human use, on a longevity story its label does not make.
The indication that landed was the rarest one. TAZPOWER enrolled twelve patients with Barth syndrome, a TAFAZZIN disorder of the exact lipid SS-31 binds. Per the FDA label it was not superior to placebo on either primary endpoint, six-minute walk and Total Fatigue Score, and knee extensor strength did not separate in the controlled phase either: a median +4 newtons on drug against -5 on placebo, ranges overlapping. The +63 newtons at week 168 came from the uncontrolled open-label extension, eight of the twelve still enrolled. What followed is documented by the patient community, not any public FDA file: per the Barth Syndrome Foundation the application was refused for filing in 2021 and refused again by complete response letter in May 2025, before a third submission that August.