Skip to content
NatureHQ

Marine lifemechanism

Antifreeze proteins

The freezing point stopped matching the melting point, and that gap was the fingerprint of a protein that stops ice crystals growing rather than stopping water freezing.

Antifreeze proteins do not lower the freezing point the way salt or glycol does. They stick to the surface of microscopic ice crystals and stop them growing, which leaves a peculiar signature: the temperature at which the fluid freezes stops matching the temperature at which it melts.

The discovery began as an arithmetic problem. Antarctic fish live in water at −1.9 °C, which is below the freezing point their dissolved salts should give them, and yet they are not frozen. In 1969 that gap was measured and named; two years later the substance responsible was isolated, and it turned out to work in a way nothing else in the fish did. Salt, sugar and glycol lower a freezing point colligatively — the effect depends only on how many particles are dissolved, so you need a lot of them. These proteins are present at concentrations far too low for that, and they produce an effect no colligative agent can: they lower the freezing point without lowering the melting point to match. That gap, thermal hysteresis, is the fingerprint of a completely different mechanism. What they do is bind to specific faces of a tiny ice crystal and occupy them, so that further water molecules can only add in the gaps between bound proteins, forcing the ice surface into tightly curved shapes that are thermodynamically unfavourable. Growth stalls. The crystal stays microscopic until the temperature drops far enough to overcome the curvature penalty, at which point ice grows anyway — so these are not a licence to be arbitrarily cold, but a gap of a degree or two that is exactly the gap the fish needs. Two further things make this the best story in the sprint. The fish version was traced back to a digestive enzyme gene, still legible in the sequence: a survival protein that was recruited, not designed. And Arctic cod, at the other end of the planet and unrelated, make virtually the same protein from an entirely different ancestral gene.

Developed coverage · 48% complete · reviewed 2026-09-03

What this page covers

Antifreeze proteins have been found in polar and sub-polar fish, in many insects and other arthropods, in plants, in fungi and in bacteria. The protein families are structurally unrelated to each other, which means the function has been arrived at repeatedly and separately.

Often confused with: Antifreeze in a car radiator, which works by a completely different mechanism; Cryoprotectants such as glucose and glycerol, which do work the car-radiator way; Ice-nucleating proteins, which are their functional opposite and often occur in the same animal

Quick facts

The mechanism
Binding to ice crystal faces and blocking growth — not lowering the freezing point
The fingerprint
Thermal hysteresis: the fluid freezes colder than it melts
Where the fish one came from
A pancreatic digestive enzyme gene, still readable in the sequence
Arrived at twice
Antarctic fish and Arctic cod, from unrelated genes, at opposite poles

Not the same thing as antifreeze

The name is a historical accident, and it points at the wrong mechanism.

Antifreeze proteins do not work like the antifreeze in a car. They bind to tiny ice crystals and stop them growing, which leaves a strange signature: the freezing point no longer matches the melting point.

Established

Specialists would state this without hedging. Multiple independent lines of evidence agree.

Antifreeze proteins act non-colligatively by adsorbing to specific planes of ice crystal surfaces and inhibiting growth, producing thermal hysteresis — a separation between the freezing and melting points — rather than the colligative freezing-point depression produced by dissolved solutes.

Who this applies to
Demonstrated in polar and sub-polar fish, in terrestrial arthropods, and in plants and bacteria, with structurally unrelated proteins in each.
Studied in
Notothenioidei, Boreogadus saida, Insecta, Plantae
Why we rate it this way, and what the caveats are
EstablishedHigh confidence

The mechanism produces a distinctive and easily measured signature that dissolved solutes cannot produce, and it has been reproduced for many independent protein families.

How far it can be extended

Thermal hysteresis has been measured for antifreeze proteins of several unrelated structural families across fish, insects, plants and microbes.

Caveats

  • Antifreeze proteins do not prevent freezing indefinitely. Below the hysteresis gap, ice grows regardless.
  • They also inhibit recrystallisation — the enlargement of small crystals into damaging ones — which may matter as much as the hysteresis in freeze-tolerant animals.

Still unanswered

  • How structurally unrelated proteins arrived at the same ice-binding function, and whether their binding geometries are as different as their folds.

Last reviewed 2026-09-03

The evidence (3 studies)

The short answer

How can a protein stop water freezing at concentrations that low?

It doesn’t stop water freezing. It stops ice crystals growing, by sticking to their surfaces — and blocking growth needs far fewer molecules than changing the freezing point of the whole solution would.

The distinction shows up in a measurement you can do on a bench. Dissolve enough salt in water and both the freezing point and the melting point drop together, by the same amount, because you have changed the thermodynamics of the whole solution. Add an antifreeze protein and the melting point barely moves while the freezing point falls by a degree or more. That gap cannot be produced by a colligative agent at any concentration, which is how the mechanism was identified before anybody could see the protein on the ice.

Check it for yourself

Two ways of dealing with ice, and how to tell them apart
AspectColligative (salt, glycerol, glucose)Antifreeze proteins
What it acts onThe whole solutionThe surface of individual crystals
Concentration neededHigh — grams per litreLow — the effect is per crystal, not per litre
Effect on melting pointFalls, by the same amount as the freezing pointBarely moves
SignatureFreezing and melting points stay togetherThey separate — thermal hysteresis
LimitSet by how much you can dissolveA degree or two, then ice grows regardless

Diagram

The freezing point stops matching the melting point

The gap is the signature of a mechanism no dissolved solute can produce.

The fingerprint: freezing point stops matching melting pointSalt or glycerolmelts and freezes hereBoth points move together, by the same amountAntifreeze proteinfreezesmeltsthe gap — no dissolved solute can do thisThe protein binds ice crystals and blocks their growth.Which needs far fewer molecules than changing the whole solution would.
The same explanation in words

Two temperature lines. On the first, for a dissolved solute such as salt or glycerol, a single point marks where the solution both melts and freezes: the two move together, by the same amount. On the second, for a solution containing antifreeze protein, there are two separate points with a bar between them — the solution freezes at a distinctly lower temperature than it melts. That gap, thermal hysteresis, cannot be produced by a colligative agent at any concentration, and it is how the mechanism was identified: the protein binds to ice crystals and blocks their growth, which needs far fewer molecules than changing the whole solution would.

Where the fish protein came from

A digestive enzyme gene, with the ancestry still legible.

The antifreeze protein of Antarctic fish evolved from a digestive enzyme gene. Parts of the original are still legible in the sequence — a survival mechanism that was recruited rather than invented.

Well supported

Good evidence backs this, though some details remain open.

The antifreeze glycoprotein gene of Antarctic notothenioid fishes derived from a pancreatic trypsinogen gene through amplification of a short internal repeat, with regions of the ancestral gene retained and identifiable in the extant sequence.

Who this applies to
Established for the antifreeze glycoprotein of the Antarctic notothenioid fishes specifically.
Studied in
Notothenioidei
Why we rate it this way, and what the caveats are
Well supportedHigh confidence

The sequence relationship is unusually clear, with the ancestral gene’s remnants directly identifiable rather than inferred from distant similarity.

Caveats

  • Ancestry inferred from sequence similarity is inference, though here the retained regions make it about as direct as this evidence gets.
  • The selective conditions during the transition are reconstructed from the geological record rather than observed.

Still unanswered

  • How the intermediate forms functioned — whether partial antifreeze activity was useful before the repeat had fully amplified.

Last reviewed 2026-09-03

The evidence (2 studies)

The antifreeze glycoprotein is a short sequence repeated many times over — structurally about as simple as a functional protein gets. When it was sequenced and compared against the genome, its origin turned out to be a gene for trypsinogen, a pancreatic digestive enzyme, in which a small internal segment had been amplified into a long tandem repeat. Parts of the original gene are still present at either end, doing nothing, like the frame of a building that was converted to another use. This is the single clearest molecular example NatureHQ holds of the general point that survival machinery is usually recruited rather than invented.

Antarctic fish and Arctic cod make virtually the same antifreeze protein, from completely unrelated genes, in oceans at opposite ends of the planet. They arrived at it separately.

Established

Specialists would state this without hedging. Multiple independent lines of evidence agree.

Antifreeze glycoproteins of Antarctic notothenioids and Arctic cod are near-identical in sequence and function but derive from unrelated ancestral genes, establishing convergent evolution at the molecular level rather than shared descent.

Who this applies to
Demonstrated between two specific fish lineages at opposite poles.
Studied in
Notothenioidei, Boreogadus saida
Why we rate it this way, and what the caveats are
EstablishedHigh confidence

The usual objection to a convergence claim — that similarity might reflect common ancestry — is closed off directly by the ancestral sequences, which are unrelated.

Caveats

  • Two lineages. How often this particular solution has been reached elsewhere is not addressed by this work.

Still unanswered

  • Whether the repeated arrival at the same simple repeated sequence reflects a small number of workable solutions or a mutational bias that makes this one easy to reach.

Last reviewed 2026-09-03

The evidence (1 study)

The convergence result is worth dwelling on because of how completely it closes off the alternative. Whenever two distant species share a striking feature, the first question is whether they inherited it from a shared ancestor. Here the answer is available in full: the two genes descend from unrelated ancestors, and the proteins arrived at near-identical form separately, at opposite ends of the planet, as each ocean cooled. Very few convergence claims can be settled that cleanly.

Diagram

A digestive enzyme gene became an antifreeze

The ancestry is still legible in the sequence.

A digestive enzyme gene became an antifreezeAncestral gene — trypsinogena short internal segmentAntifreeze glycoprotein genethe same segment, repeated over and overParts of the original gene are still there, doing nothing.Arctic cod arrived at nearly the same protein from an unrelated gene.Which rules out shared ancestry, and leaves convergence.Survival machinery is more often recruited than invented.
The same explanation in words

Two gene diagrams. Above, the ancestral trypsinogen gene: a long block, a short highlighted internal segment, and another long block. Below, the antifreeze glycoprotein gene: a short remnant of the original, followed by the same short segment repeated many times over. Parts of the original gene are still present, doing nothing. Arctic cod arrived at nearly the same protein from an entirely unrelated ancestral gene, which rules out shared ancestry and leaves convergence. Survival machinery is more often recruited than invented.

The innovation came ten million years before the radiation

A key innovation is not enough on its own — the opportunity has to arrive as well.

The tidy version of this story is that antifreeze let the notothenioid fishes take over the Southern Ocean while everything else froze out. The dates do not support it. When the group’s phylogeny was built and calibrated, the antifreeze protein turned out to have evolved roughly ten million years before the fishes diversified. The protein was available and nothing much happened with it for a very long time; the radiation followed later environmental change, not the innovation. It is a useful corrective to a common shape of evolutionary story, in which a capability appears and success follows immediately.

  • How did partial antifreeze activity help, before the repeat had fully amplified?

    Why it matters: A protein that only works at full length is hard to evolve gradually. If intermediate forms conferred partial protection, the transition is straightforward; if they did not, something else has to explain how it got there.

    What would settle it: Synthesising reconstructed intermediate sequences and measuring their thermal hysteresis directly.

  • Why has the same simple repeated sequence been reached more than once?

    Why it matters: It could mean there are very few workable ice-binding solutions, or that this particular sequence is unusually easy to reach by mutation. Those imply different things about how predictable molecular evolution is.

Related

  • Freeze tolerance

    Where these proteins do a second, different job

  • Icefish

    The same waters, and a trait that is not an adaptation

  • Adaptation

    Co-option, and why usefulness is not evidence of origin

Antifreeze evolved roughly ten million years before the Antarctic fishes diversified. The innovation sat available and nothing much happened; the opportunity arrived long afterwards.

Emerging evidence

Real findings exist, but too few or too recent to be settled.

Dated phylogenetic analysis of the notothenioid fishes places the origin of antifreeze glycoproteins approximately ten million years before the group’s major diversification, indicating that the radiation followed subsequent environmental change rather than the innovation itself.

Who this applies to
The Antarctic notothenioid radiation specifically.
Studied in
Notothenioidei
Why we rate it this way, and what the caveats are
Emerging evidenceModerate confidence

The relative ordering — innovation before radiation — is robust in the analysis. The size of the gap is a divergence-dating estimate with substantial uncertainty, and fossil calibration for this group is limited.

Caveats

  • Divergence dates carry wide error bars, so ten million years is an estimate rather than a measurement.
  • A gap between innovation and radiation does not establish what filled it; the environmental account is inference from the geological record.

Still unanswered

  • What the notothenioids were doing during the interval, and whether antifreeze was under selection throughout it or only intermittently.

Last reviewed 2026-09-03

The evidence (2 studies)

The research behind this page

7 studies, newest first. Each one has a page explaining what it found and what it could not show.

2012Proceedings of the National Academy of Sciences

Ancient climate change, antifreeze, and the evolutionary diversification of Antarctic fishes

Antifreeze evolved roughly ten million years before the group diversified.

2001Annual Review of Physiology

Antifreeze and ice nucleator proteins in terrestrial arthropods

Antifreeze proteins act by binding to the surface of small ice crystals and preventing their growth, producing a gap between the freezing and melting points; ice-nucleating proteins do the reverse.

1997Proceedings of the National Academy of Sciences

Evolution of antifreeze glycoprotein gene from a trypsinogen gene in Antarctic notothenioid fish

The antifreeze glycoprotein gene evolved from a pancreatic trypsinogen gene — a digestive enzyme — by amplification of a short repeated segment, with parts of the ancestral gene still recognisably present.

1997Proceedings of the National Academy of Sciences

Convergent evolution of antifreeze glycoproteins in Antarctic notothenioid fish and Arctic cod

The two lineages produce essentially the same antifreeze protein from completely unrelated ancestral genes, in oceans at opposite poles, having evolved it independently.

1982Paleobiology

Exaptation — a missing term in the science of form

Current utility and evolutionary origin are separate questions, and a large fraction of useful traits were co-opted rather than built for the job they now do.

1971Science

Glycoproteins as biological antifreeze agents in Antarctic fishes

The glycoproteins lowered the freezing point without lowering the melting point by the same amount — a gap known as thermal hysteresis — showing that they act on ice crystal growth rather than by ordinary colligative depression.

1969Science

Freezing resistance in some Antarctic fishes

The fish remained unfrozen at temperatures below the freezing point predicted from their dissolved salts, indicating an additional and previously unidentified antifreeze substance in the blood.

This page is a stop on a longer route

A guided journey reads several subjects in a deliberate order, with an argument for why one follows another. You can join in the middle.

Where to go from here

Each of these follows from something on this page — a relationship in the evidence, a claim people ask about, or the next mechanism along.

How complete this page is, and what it is still missing

NatureHQ publishes its own gaps. This page is at 48% completeness against what we would call a finished subject, and was last reviewed on 2026-09-03. It carries 4 claims and answers 13 mapped search questions.

  • no research from the last few years is attached — check for newer work
  • more experiments could be explained in plain English
  • no popular claim about this subject has been checked yet
  • Insect and plant antifreeze proteins are structurally different and are mentioned rather than described.
  • The industrial and food-science uses of these proteins are outside the scope of this page.
  • How the proteins recognise particular crystal faces is stated as the mechanism without the structural detail.