Insulin Analogs Explained: How a Few Amino Acids Reset the Clock

Peptide engineering

Insulin analogs are the most consequential peptide engineering project ever completed, and almost all of it comes down to moving, deleting or decorating a handful of amino acids. The receptor pharmacology barely changed. What changed was the clock — how fast the molecule leaves the injection site and how long it lingers — and that turned out to be the thing that mattered.

The molecule insulin analogs start from

Human insulin is a 51-residue, two-chain peptide: a 21-residue A chain and a 30-residue B chain, held together by two interchain disulfide bonds (A7–B7 and A20–B19) plus one internal bridge inside the A chain (A6–A11). Its formula is C₂₅₇H₃₈₃N₆₅O₇₇S₆ and its molecular weight is 5,808 daltons. It is made as proinsulin and cut into shape inside the secretory granule, releasing C-peptide as the offcut.

The engineering story turns on what insulin does to itself. Two molecules associate into a dimer through an antiparallel beta-sheet formed by the C-terminal end of each B chain — residues B24 to B28, with proline at B28 sitting right in that interface. Add zinc, and three dimers assemble around two zinc ions into a hexamer, each ion held by three histidines at position B10.

Only the monomer binds the insulin receptor. So a vial of insulin is a stable hexameric store that has to come apart in the tissue before it can do anything. That disassembly step became the design target for every rapid-acting analog.

Making it fast: the rapid-acting analogs

All three rapid-acting analogs share one idea — disturb the surfaces that hold insulin together, so it disperses into monomers sooner after injection.

• Lispro (Humalog). The proline at B28 and the lysine at B29 are simply swapped. Nothing is added or removed: the label notes the formula and molecular weight are identical to human insulin. Peak serum levels arrive 30–90 minutes after injection, against 50–120 minutes for regular human insulin.
• Aspart (NovoLog). A single substitution — proline at B28 becomes aspartic acid, putting a negative charge into the dimer interface. Median time to peak is 40–50 minutes versus 80–120 for regular human insulin; maximum glucose-lowering effect at 1–3 hours, duration 3–5 hours.
• Glulisine (Apidra). Two changes, both away from B28: asparagine at B3 becomes lysine, and lysine at B29 becomes glutamic acid. Apparent subcutaneous half-life is 42 minutes against 86 for regular human insulin. It is also formulated without zinc, using polysorbate 20 and tromethamine instead — a formulation answer to the same problem.

Faster analogs since then have stopped changing the peptide at all and changed the excipients. Fiasp is chemically the same molecule as NovoLog with niacinamide and L-arginine added; insulin appears in circulation about 2.5 minutes after injection.

Making it slow: three different tricks

The long-acting analogs are more interesting, because they solve the same problem three completely different ways.

Glargine — make it insoluble at body pH. Two arginines are added to the C-terminus of the B chain, shifting the isoelectric point from about 5.4 to about 6.7. The vial is formulated at pH 4, where the molecule is fully soluble. Injected into tissue at neutral pH, it lands near its isoelectric point and precipitates into microcrystals that redissolve slowly. The A21 asparagine-to-glycine change is often misread as part of the protraction; it is not. It exists because asparagine deamidates in an acidic formulation, and the molecule had to survive the pH 4 vial. End of effect comes at a median of 24 hours, against 14.5 hours for NPH.

Detemir — hitch a ride on albumin. The B30 threonine is removed and a 14-carbon fatty acid is attached to the lysine at B29. The result is neutral in the vial, and more than 98% albumin-bound in the bloodstream. Slower absorption from self-association at the depot plus slow release from albumin gives a terminal half-life of 5–7 hours.

Degludec — build a chain of hexamers. B30 is again removed, and a 16-carbon diacid is attached at B29 through a glutamic acid spacer. In the presence of phenol and zinc the molecule forms long multi-hexamer chains at the injection site that disassemble from the ends, molecule by molecule. Half-life at steady state is about 25 hours, independent of dose, and glucose-lowering effect persisted at least 42 hours after the last of eight daily injections.

The weekly one, and why the label matters

Insulin icodec pushes the same logic to its limit. Its full designation names three point substitutions — tyrosine to glutamate at A14, tyrosine to histidine at B16, phenylalanine to histidine at B25 — plus deletion of B30 and a 20-carbon diacid attached at B29 through a glutamic-acid and two-unit linker. The substitutions stabilise the molecule and deliberately weaken receptor binding, which slows the receptor-mediated clearance that normally destroys circulating insulin. The result is a human half-life of about 196 hours — roughly eight days.

The clinical record is split by diabetes type, and this is where most write-ups are out of date. In ONWARDS 1, 984 insulin-naive people with type 2 diabetes had a slightly greater HbA1c reduction on weekly icodec than on daily glargine (−1.55 vs −1.35 percentage points, superiority p = 0.02). In ONWARDS 6, in type 1 diabetes, HbA1c was non-inferior but combined clinically significant or severe hypoglycaemia ran at 19.93 versus 10.37 events per patient-year — close to double.

Regulators split accordingly. The EU authorised icodec in May 2024 for diabetes in adults including type 1, with an explicit caution that people with type 1 should only start it where there is clear benefit. The FDA issued a complete response letter in July 2024, and approved it in March 2026 for type 2 diabetes only. A weekly injection with an eight-day half-life is a very different safety proposition when the dose cannot be walked back.

What the pre-analog era already knew

NPH insulin is worth keeping in view, because it makes the same point using no engineering at all. It is ordinary human insulin co-crystallised with protamine, a cationic protein originally isolated from fish sperm. The crystals dissolve slowly, converting a fast insulin into an intermediate one with maximum effect at a median of 6.5 hours. The costs are visible in the vial: it is a cloudy suspension that must be resuspended before every dose, and it has a pronounced peak.

The premixed products then reapplied the same trick to the engineered molecules — a 75/25 mix is 75% insulin lispro protamine suspension and 25% soluble insulin lispro. Formulation and sequence engineering were never really alternatives.

Concentration changes kinetics too

Here is the cleanest proof that sequence is not the only lever. Humulin R U-500 is unmodified regular human insulin at five times the usual concentration. Its label reports onset under 15 minutes but a duration of action of 21 hours, with both prandial and basal character, and attributes that behaviour explicitly to the high concentration of the preparation. Identical peptide, completely different clinical profile.

The effect is not automatic. Humalog U-200 is bioequivalent to U-100 at a 20-unit dose, with the same median time to peak. But Toujeo, which is glargine at 300 units/mL with triple the zinc of Lantus, behaves measurably differently from the U-100 product: onset develops over about six hours, and at steady state the 24-hour glucose-lowering effect of a 0.4 U/kg dose is roughly 27% lower than the same dose of Lantus. Concentration is a formulation variable that can reshape the depot.

None of this changes what a unit means. If you are working through unit and volume conversions, our explainer on why units are not a dose and the U-100 versus U-40 syringe guide cover the arithmetic, and the reconstitution calculator and concentration converter handle the mg, mL and mg/mL relationships.

AspB10: the analog that got cancelled

The most instructive insulin analog is the one you cannot buy. Insulin X10 replaced the histidine at B10 with aspartic acid — a rational choice, since B10 is exactly where zinc binds the hexamer together, and destabilising it should mean faster absorption. It did.

It also did something else. AspB10 binds the insulin receptor with about 251% of human insulin’s affinity, dissociates from it more slowly, and — the decisive number — binds the IGF-1 receptor at about 492%. Mitogenic potency ran several-fold higher in multiple cell lines. Development was discontinued after supra-pharmacological doses increased mammary tumour incidence in female rats in a chronic toxicity study.

The lesson generalises well beyond insulin: in peptide engineering, pharmacokinetics and receptor pharmacology are not independent dials. The residue that controls how fast a molecule disperses may also sit on the surface that decides what it binds and how long it stays bound. Every insulin programme since has been designed around that discovery, and the same logic drives half-life extension work across the peptide field.

From drugs to biologics

One regulatory note that changed the market. On 23 March 2020, under the Biologics Price Competition and Innovation Act, approved applications for protein products including insulin were deemed to be biologics licences rather than new drug applications. That transition is what made true biosimilar and interchangeable insulins legally possible. In July 2021 insulin glargine-yfgn became the first biological product to receive an interchangeability determination against its reference product — meaning, for the first time, a pharmacy could substitute it directly.

Frequently asked questions

Are insulin analogs stronger than human insulin?

No. A unit is defined by biological activity, so a unit of analog and a unit of human insulin lower glucose comparably. What differs is when. The analogs were designed to change the shape of the curve, not its area.

Why does glargine come at pH 4 when everything else is neutral?

Because its protraction mechanism requires it. Glargine’s isoelectric point was deliberately shifted to about 6.7, so it is only fully soluble at acidic pH. The acidic vial is not a preservative choice — it is the storage condition that keeps the drug dissolved until tissue pH precipitates it.

Do fatty-acid-modified insulins work the same way as each other?

Not quite. Detemir’s 14-carbon chain works mainly through albumin binding. Degludec’s 16-carbon diacid works mainly by building multi-hexamer chains in the depot that disassemble slowly, with albumin binding as a secondary contribution. Icodec extends the same idea with a 20-carbon diacid plus substitutions that deliberately slow receptor-mediated clearance.

Is weekly insulin approved everywhere for everyone?

No, and the difference matters. The EU authorisation covers adults with diabetes including type 1, with an explicit caution about starting type 1 patients on it. The FDA approval, granted in March 2026, covers type 2 diabetes only. The hypoglycaemia signal in the type 1 trial is the reason for the split.

References

Informational only — not medical advice · 21+. Nothing here is dosing guidance. Insulin is a high-risk medicine; any decision about it belongs with a qualified healthcare professional.