Skip to content
type1.science

Glucose-responsive ("smart") insulin

Multiple (Novo Nordisk; University of Bristol/Carbometrics; academic programs)

A class of insulins engineered to sense blood glucose and self-modulate their own activity — turning down when glucose is low to prevent hypoglycemia and up when glucose is high. Two industry candidates have now completed or entered Phase 1 testing in people with T1D — Lilly's LY3938577 and Novo Nordisk's NNC0363-1063 — but no human glucose-response curve or outpatient efficacy result has been disclosed. The published molecular switch, NNC2215, remains animal-only. No product exists yet.

Years awayEarly evidencebasalglucose-responsivemealtime

The scorecard

Onset speed40

Basal-role convention (a basal's job is steady coverage, not a fast meal response): NNC2215 behaves like a long-acting insulin (s.c. half-life ~19 h in pigs) — onset is slow by design; speed is not where its value lies.[1]

Time to peak70

Basal-role convention (flat/peakless scores HIGH for a basal): a long-acting profile is desirable here, and the glucose switch adds extra activity only when glucose rises — a peakless baseline that auto-boosts on demand.[1]

Short tail60

Basal-role convention (long, smooth coverage is good for a basal): protracted action suits once-daily basal dosing; the open question is how cleanly activity falls at true lows in humans.[1]

Consistency35

Entirely preclinical — glucose-sensitivity is demonstrated in rats and pigs but never in people; day-to-day human absorption variability is simply unknown, and a prior smart-insulin (MK-2640) lost its glucose response in human trials.[4]

Exercise flexibility75

A working glucose switch is, in principle, the strongest possible exercise-safety feature — activity would auto-attenuate as glucose falls during exertion — but this is theoretical until shown in humans.[1]

Access & cost5

Access convention (cheaper/more available = better): not approved, not in human trials, no price, no biosimilar — years from any pharmacy. Scored low purely on present obtainability.[5]

Insulins are scored relative to their role peers (see tags: rapid, ultra-rapid, basal, inhaled). A basal insulin's onset score compares it to other basals, not to mealtime insulins.

Editor’s take

This is the molecule the field has chased since the 1970s, and NNC2215 is the most convincing demonstration yet that a reversible glucose switch can be built into insulin itself. But "most convincing yet" still means animals only — and the graveyard of smart insulins (Merck's MK-2640 worked in dogs, then failed the glucose-response test in humans) is the reason we rank this as `research`, not a product. It is distinct from our [sub-20-minute concept](/items/concept-ultrafast-insulin): that target is about speed; this is about a safety floor against lows.

The full picture

Glucose-responsive insulin — often called "smart insulin" — is not one product but a research class: insulins engineered to sense the surrounding glucose and adjust their own potency, dialing down when glucose is low to avoid hypoglycemia and dialing up when it is high.1 The goal has been pursued since the 1970s precisely because the central danger of every insulin in use today is that its action is fixed once injected — a slightly-too-large dose can drive a dangerous low regardless of what blood glucose is actually doing.1 The class has now reached people with T1D, but only just. Novo Nordisk's NNC0363-1063 completed two Phase 1 studies in T1D in March and May 2026,23 and Lilly's LY3938577 is recruiting a Phase 1 in T1D.4 Neither company has disclosed a human glucose-response curve or any outpatient efficacy result, and no results are posted for the completed studies. Nothing is approved and nothing is in Phase 3.

Keep two things separate. The clinical candidates above are proprietary molecules with almost nothing published about them. The published molecule — NNC2215, reported in animals in October 20241 and described in detail below — is a different compound and has never been given to a person.5 Everything in the next three sections is animal and test-tube data about NNC2215; it is the best-characterised glucose-responsive insulin in the literature, not the one furthest along in trials.

How it works (NNC2215). Insulin is fitted with a molecular "switch": a glucose-binding macrocycle attached at one site (B29Lys) and a glucoside at another (B1Phe). At low glucose the glucoside docks into the macrocycle, holding the insulin in a closed, less-active shape that clashes with the insulin receptor; as glucose rises, glucose displaces the glucoside, the switch opens, and the insulin becomes more active.1 The glucose-binding macrocycle itself is a "synthetic lectin" developed over roughly three decades in Anthony Davis's lab at the University of Bristol.6

The real numbers — but read them as a target type, not a label. This is a basal-like molecule, so we score it against other basal insulins (where flat, peakless, long coverage is good). In test-tube assays, NNC2215's insulin-receptor affinity rose 3.2-fold as glucose went from 3 to 20 mM (and 12.5-fold from 0 to 20 mM) — a genuine, reversible glucose response.1 Its half-life was about 1.2 h given intravenously to rats and 1.3 h in pigs; given under the skin in pigs, bioavailability was 73% and the half-life stretched to about 19 h, suggesting a once-daily basal profile.1 There is no human onset, time-to-peak, or duration data — because it has never been given to a person.5

Did the glucose-sensing actually protect against lows? In pigs, when insulin was held constant and glucose was allowed to fall, glucose bottomed out around 4.5 mM with NNC2215 versus below 3 mM with ordinary long-acting insulin — measurable protection against hypoglycemia.1 In a glucose challenge in diabetic rats, the switch added roughly 30% extra insulin effect at high glucose without any change in dose — hinting it could blunt post-meal spikes too.1

Why honest caution matters here. The field has been burned before. Merck's MK-2640 used a different trick — saccharides that make insulin clear faster via the mannose receptor when glucose is low — and looked excellent in dogs and minipigs (a 3-fold therapeutic index).7 But in people with type 1 diabetes it could not demonstrate a glucose-dependent change in clearance, was about 25-fold less potent than human insulin, and the program was stopped.8 That is the bar NNC2215 still has to clear.

Delivery, approvals, access. None apply yet: there is no approved glucose-responsive insulin anywhere, no formulation, no price, and no biosimilar.5 The commercial backdrop is that Novo Nordisk acquired the Bristol spin-out Ziylo in 2018 for up to $800 million in milestones to build exactly this, with the spin-out Carbometrics continuing to optimize the glucose-binding chemistry.9

What's coming. For NNC2215 specifically, the next step is further pharmacological optimization, then — if it holds up — first-in-human studies; as of the 2024 report it had not been tested in humans.5 For the class, the near-term question is narrower and more concrete: Novo has completed two Phase 1 studies of NNC0363-1063 in people with T1D without posting results,23 so the first real test is whether a human glucose-response curve is ever disclosed — and whether it survives contact with the outpatient world. A glucose-responsive insulin would attack the insulin-speed and hypoglycemia-safety gap from the molecule end rather than the algorithm end: a chemical safety floor that works even when a pump, sensor, or meal estimate is wrong. Whether NNC2215's animal-stage glucose response survives the translation to people — the exact step that defeated MK-2640 — is the open question that defines this entire class.

References

  1. Hoeg-Jensen T, Kruse T, Brand CL, et al. Glucose-sensitive insulin with attenuation of hypoglycaemia. Nature (2024). https://doi.org/10.1038/s41586-024-08042-3 ↩ ↩2 ↩3 ↩4 ↩5 ↩6 ↩7 ↩8

  2. A Research Study of a New Medicine NNC0363-1063 in Healthy Participants and Participants With Type 1 Diabetes. ClinicalTrials.gov (Phase 1, n=117; started 12 Nov 2024, completed 18 Mar 2026; no results posted). https://clinicaltrials.gov/study/NCT06685185 ↩ ↩2

  3. A Research Study of a New Medicine NNC0363-1063 in Participants With Type 1 Diabetes. ClinicalTrials.gov (Phase 1, n=36; started 16 Dec 2025, completed 8 May 2026; no results posted). https://clinicaltrials.gov/study/NCT07305805 ↩ ↩2

  4. A Study of LY3938577 in Healthy Participants and Participants With Type 1 Diabetes Mellitus (T1DM). ClinicalTrials.gov (Phase 1, n=118; started 15 May 2024, recruiting as of Aug 2026). https://clinicaltrials.gov/study/NCT06280703 ↩

  5. Novo Nordisk Researchers Engineer Glucose-Sensitive Insulin Switch. Inside Precision Medicine (2024). https://www.insideprecisionmedicine.com/topics/translational-research/novo-nordisk-researchers-engineer-glucose-sensitive-insulin-switch/ ↩ ↩2 ↩3 ↩4

  6. Tromans RA, Carter TS, Chabanne L, et al. A biomimetic receptor for glucose. Nature Chemistry (2019). https://doi.org/10.1038/s41557-018-0155-z ↩

  7. Kaarsholm NC, Lin S, Yan L, et al. Engineering Glucose Responsiveness Into Insulin. Diabetes (2018). https://doi.org/10.2337/db17-0577 ↩

  8. Krug AW, Visser SAG, Tsai K, et al. Clinical Evaluation of MK-2640: An Insulin Analog With Glucose-Responsive Properties. Clinical Pharmacology & Therapeutics (2018). https://doi.org/10.1002/cpt.1215 ↩

  9. Novo Nordisk acquires Ziylo Ltd to accelerate its development of glucose responsive insulins. GlobeNewswire (2018). https://www.globenewswire.com/news-release/2018/08/17/1600408/0/en/Novo-Nordisk-acquires-Ziylo-Ltd-to-accelerate-its-development-of-glucose-responsive-insulins.html ↩

Coming soon

ETA · Phase 1 in T1D. Novo's NNC0363-1063 completed two Phase 1 studies in people with T1D (Mar and May 2026); Lilly's LY3938577 is recruiting a Phase 1 in T1D and is in Phase 2 in T2D. No glucose-response curve or outpatient efficacy result has been published from any of them. The separately published switch molecule, NNC2215, is still animal-only.

  • →Disclosure of a human glucose-response curve from the completed NNC0363-1063 Phase 1 studies in T1D
  • →Completion of the LY3938577 Phase 1 in T1D (NCT06280703), which retains mealtime lispro — the design implies a basal safety buffer, not a total-replacement insulin

Sources

  1. [1]Hoeg-Jensen T, Kruse T, Brand CL, et al. Glucose-sensitive insulin with attenuation of hypoglycaemia. Nature (2024) · peer-reviewed · 2024-10-16 — Primary paper. NNC2215: macrocycle at B29Lys + glucoside at B1Phe; IR affinity +3.2x from 3->20 mM glucose (+12.5x from 0->20 mM); in vivo half-life ~1.2 h (rat IV), ~1.3 h (pig IV); s.c. bioavailability 73%, s.c. half-life ~19 h (pig); hypoglycemia protection in pigs; ~30% extra insulin effect during rat GTT. PMC11499270; Nature 634:944-951.
  2. [2]Tromans RA, Carter TS, Chabanne L, et al. A biomimetic receptor for glucose. Nat Chem (2019) · peer-reviewed · 2018-11-12 — The glucose-binding macrocycle (synthetic lectin) underpinning NNC2215; Ka ~18,000 /M for glucose, ~100x selectivity over other sugars. Davis lab, University of Bristol / Ziylo Ltd. Nat Chem 11:52-56.
  3. [3]Kaarsholm NC, Lin S, Yan L, et al. Engineering Glucose Responsiveness Into Insulin. Diabetes (2018) · peer-reviewed · 2017-11-02 — Merck's MK-2640 (mannose-receptor clearance principle): ~30% reduction in IR availability across 280->80 mg/dL in dogs; therapeutic index 3-fold (vs 1.3-fold for regular insulin) in minipigs. Diabetes 67:299-308.
  4. [4]Krug AW, Visser SAG, Tsai K, et al. Clinical Evaluation of MK-2640: An Insulin Analog With Glucose-Responsive Properties. Clin Pharmacol Ther (2018) · peer-reviewed · 2018-09-30 — The cautionary human result: in T1D crossover study, could NOT demonstrate glucose-dependent change in MK-2640 clearance; potency ~25x lower than human insulin; program terminated. Clin Pharmacol Ther 105:417-425.
  5. [5]Novo Nordisk Researchers Engineer Glucose-Sensitive Insulin Switch (Inside Precision Medicine) · news · 2024-10-16 — States explicitly NNC2215 'has not yet been tested in humans'; preclinical stage; summarizes the three in vivo models.
  6. [6]Novo Nordisk acquires Ziylo Ltd to accelerate its development of glucose responsive insulins (GlobeNewswire) · news · 2018-08-17 — Aug 2018 acquisition of University of Bristol spin-out Ziylo for up to $800M in milestones; Carbometrics spun out to optimize glucose-binding molecules (and pursue CGM applications).
  7. [7]A Study of LY3938577 in Healthy Participants and Participants With Type 1 Diabetes Mellitus (T1DM) · registry · 2024-05-15 — Phase 1, Eli Lilly, n=118, recruiting as of Aug 2026. Comparator arms include insulin degludec and insulin lispro — the retained mealtime lispro implies a basal safety buffer rather than total replacement.
  8. [8]A Research Study of a New Medicine NNC0363-1063 in Healthy Participants and Participants With Type 1 Diabetes · registry · 2024-11-12 — Phase 1, Novo Nordisk, n=117, completed 18 Mar 2026. Single ascending dose, proof-of-principle and meal-test parts in T1D. No results posted.
  9. [9]A Research Study of a New Medicine NNC0363-1063 in Participants With Type 1 Diabetes · registry · 2025-12-16 — Phase 1, Novo Nordisk, n=36, completed 8 May 2026. No results posted.