Freeze-Thaw Cycles and Aliquoting: What the Stability Literature Measures

The instruction "avoid repeated freeze-thaw cycles" appears on nearly every peptide datasheet. It is sound advice doing a great deal of work with very little explanation. Freeze-thaw damage is not one mechanism. It bundles together at least four physically distinct events — interfacial denaturation, cryoconcentration and the pH shifts that accompany it, cold denaturation, and adsorption to whatever the solution is sitting in — each with its own dependence on rate, concentration, buffer and container. This guide separates them, states what the stability literature has measured for each, and is explicit about which numbers come from protein experiments extrapolated to peptides.

That last caveat carries weight. Almost the entire mechanistic literature on freezing damage was built on model enzymes and monoclonal antibodies, because those molecules have readouts: residual catalytic activity, a size-exclusion aggregate peak, a folded structure available to lose. A short synthetic peptide has none of those. The mechanisms below are well characterized in their own systems; whether they act with the same force on a short, largely unstructured peptide is a question the published record has mostly not asked.

The ice-water interface

The dominant explanation for freezing damage is that ice creates surface area, and surfaces unfold molecules. The clearest early evidence for this came from a comparison of several proteins of differing freeze sensitivity: the tendency of a protein to denature on freezing correlated with its tendency to denature at a surface in unfrozen solution, with a reported correlation coefficient of 0.99, and small amounts of surface-active agent protected against both ^[1]. In the same work, quench cooling by dipping samples in liquid nitrogen produced insoluble precipitates, while freezing protocols involving supercooling caused less precipitation. The authors stated it as a suggestion, and scoped it to proteins: these results, they wrote, suggest that the denaturation of proteins during freeze-thawing can be ascribed primarily to the increase in the area of the ice-water interface during freezing ^[1].

Freezing rate follows directly from that. Fast freezing nucleates many small ice crystals and therefore a large total interfacial area; slow freezing produces fewer, larger crystals. Working with model enzymes on a controlled microcryostage in the absence of cryoprotectants, one study found slow freezing (around 1 °C/min) combined with fast thawing (above 10 °C/min) gave higher activity recovery, while fast freezing with slow thawing caused more severe damage, and — once freezing concentration and buffer pH change had been minimized by using a potassium phosphate buffer — ascribed the pattern to interfacial exposure during freezing plus recrystallization stress during thaw ^[2]. Note that the freezing and thawing limbs behaved as separable, reciprocal effects — the thaw is not simply the freeze run backwards.

Why a surface with no chemical difference from bulk water should be destabilizing is less settled. Molecular dynamics simulation of a model protein indicates the ice-water interface lowers the free-energy barrier for unfolding, and — importantly — that the effect does not require the protein to adsorb directly to the ice. In that work it was mediated by nearby liquid water showing an increased tendency to hydrate nonpolar groups, and was mitigated by glucose acting through preferential exclusion ^[3]. A separate line of work proposes that the destabilizing stress does not originate with the ice-water system at all: as temperature falls, dissolved air is expelled from the growing ice lattice as microbubbles and trapped by the advancing front, creating additional air-water interface. Freezing at approximately 1 K/s produced microbubbles roughly 100–500 µm across, slower freezing produced larger elongated bubbles, and deaerating samples under reduced pressure before freezing reduced subvisible aggregate counts for both a highly surface-active and a low surface-active protein, though by an amount that depended on the freezing technique and was least apparent in conventional lyophilization ^[4]. The authors present this as a hypothesis tested in specific freezing techniques, not as a replacement for the interfacial model.

A 2020 review of the field is candid that the subject is not yet fully understood and that several ice-induced denaturation mechanisms remain under active proposal ^[5]. A bench researcher should read freezing-rate advice in that light: the direction of the effect is not fixed. A design-space analysis identified two opposing scenarios, which the authors say suggest that the ice-water interface is the dominant cause of denaturation for proteins with high bulk stability, whereas the duration of the freezing process is the key parameter to control for proteins susceptible to cold denaturation — and supported this with experiments in which slow freezing was optimal for lactate dehydrogenase and the opposite held for myoglobin ^[6]. "Freeze quickly" and "freeze slowly" are each defensible, depending on the molecule.

Cryoconcentration: what the unfrozen phase becomes

Water leaving solution as pure ice does not take solutes with it. Everything dissolved is excluded into a shrinking unfrozen phase, so concentration in that phase rises steeply as freezing proceeds. The arithmetic is worth doing once. Take a 1.0 mL aliquot at 2.0 mg/mL, containing 2.0 mg of peptide. If 90% of the water has crystallized, the solute is confined to the remaining 0.10 mL: 2.0 mg ÷ 0.10 mL = 20 mg/mL, a tenfold rise. At 99% ice, the same 2.0 mg occupies 0.010 mL — 200 mg/mL. That is illustrative arithmetic assuming complete solute exclusion, not a measurement — real freeze-concentrated phases are heterogeneous. The point survives: for part of every freeze, the peptide sits at a concentration nothing on the label describes.

The buffer is concentrated alongside it, and buffers do not all stay in solution. In sodium phosphate the disodium species crystallizes preferentially, stripping base from the remaining liquid. Measured in situ during cooling to −10 °C at 0.3–0.5 °C/min, precipitation of Na2HPO4·12H2O caused abrupt pH decreases beginning between −0.5 and −4.0 °C; from an initial pH of 7.4, 50 and 100 mM solutions reached pH 4.2 ± 0.1 at −10 °C while 8 mM solutions reached 5.2 ± 0.2 ^[7]. Lower initial buffer concentration and lower initial pH both produced smaller shifts. A broader survey across common formulation buffers found sodium phosphate showed the greatest pH change from +25 to −30 °C, that histidine hydrochloride, sodium acetate, histidine acetate, citrate and succinate each changed by less than 1 pH unit over that range, and that Tris-hydrochloride increased by approximately 1.2 units ^[8].

It would be easy to conclude from this that pH shift is the thing to worry about. The same survey argues against that inference for its own system: in a formulated IgG2 antibody, soluble aggregate formation after repeated freeze-thaw cycles was not governed by the pH change but by the presence or absence of noncrystallizing cryoprotective excipients, with longer phase-transition times producing more aggregation when no cryoprotectant was present ^[8]. Cryoconcentration changes several variables at once, and which one does the damage is molecule-dependent.

Scale is the other honest caveat. The cryoconcentration literature is largely about large containers, where slow uncontrolled freezing produces macroscopic zones of differing composition. One long-term study varied the antibody-to-buffer ratio to mimic those zones and stored samples at −80 °C, −20 °C and −10 °C for 6 weeks, 6 months and 12 months: storage below the glass transition temperature of the maximally freeze-concentrated solution (Tg′) did not lead to aggregation, while above Tg′ — at −20 °C and −10 °C — small changes in protein and buffer concentration markedly affected stability ^[9]. A 200 µL aliquot in a microcentrifuge tube is not a multi-liter bag, and the spatial gradients are far smaller. The Tg′ result, however, is the one mechanistic argument in this literature for a colder storage setpoint that does not rest purely on convention.

Concentration is a variable, not a fixed property of the experiment

How much peptide is in the aliquot changes how much damage a cycle does, and not by a small margin. Using small-angle neutron scattering, dynamic light scattering and low-temperature X-ray diffractometry together, one study found lactate dehydrogenase at 1000 µg/mL in 10 mM buffer showed a pronounced self-stabilizing effect and did not aggregate after five freeze-thaw cycles, whereas at 10 and 100 µg/mL irreversible aggregation was avoided only with an appropriate buffer selection ^[10]. In the same work, selective crystallization of disodium hydrogen phosphate during freezing caused a pronounced pH shift while histidine-buffered solutions showed none ^[10].

This cuts against a naive reading of aliquoting practice. Splitting a stock into very small, very dilute portions reduces the number of cycles each portion sees, which is the goal — but dilute solutions have more interfacial area and more container surface per molecule of peptide. The two effects push in opposite directions, and the published data quantifying that trade-off is for enzymes, not peptides.

Adsorption to container surfaces

Peptide that is no longer in solution has not necessarily degraded. It may simply be on the wall of the tube. A systematic study of 50 peptides obtained by enzymatic digestion of bovine serum albumin classified them by response and adsorption behavior, then tracked the worst-affected subset across three tube types and three vial types. Modified ("low-bind") polypropylene gave recoveries above 80% over 24 hours during storage and sample preparation, and the parameters that ought to predict adsorption — an HPLC index and peptide length — were significantly correlated with adsorption degree (p < 0.05) but explained the behavior of only about 30% of the peptides ^[11]. The same work reduced instrument carryover to roughly 1/150 of its starting value through combined chromatographic optimization ^[11].

Two things follow. Adsorptive loss is large enough to matter within a single working session, not only over long-term storage; and it is not reliably predictable from sequence. If a peptide's recovery in a particular container matters to a result, it has to be measured for that peptide, because the available descriptors do not settle it. Adsorption is also not independent of freezing: each thaw re-exposes the peptide to the container wall, and cryoconcentration raises local concentration at that wall during the freeze.

What the stability literature actually measures

It is worth being precise about the endpoints, because they determine what these papers can and cannot tell you. The freeze-thaw literature overwhelmingly reports one of: residual enzymatic activity ^[2], insoluble precipitate formation ^[1], higher-molecular-weight species by size-exclusion chromatography and subvisible particle counts by flow imaging ^[9], particle size by dynamic light scattering or conformation by neutron scattering ^[10], Tg′ by differential scanning calorimetry ^[9], or analyte recovery by liquid chromatography-mass spectrometry ^[11]. Freeze-thaw also appears as one stress arm among several in peptide analytical method development — a size-exclusion LC-UV/HRMS method for a synthetic GLP-1 analog was developed against photolytic, thermal, freeze-thaw and shaking stress in order to separate and identify aggregates ^[12].

Notice what is missing. Activity assays require a folded catalytic protein; aggregate peaks require an aggregate large enough to resolve from monomer. For a short synthetic peptide with no enzymatic readout and limited tertiary structure, the most sensitive tools here do not apply, and what remains is chromatographic purity and recovery — which registers loss but does not always distinguish adsorption from degradation from aggregation.

The literature also contains results that cut against the standard warning. Synthetic GLP-1 and glucagon spiked into pooled human plasma with protease inhibitors added showed recoveries that were unaffected by up to three thawing-refreezing cycles, and GLP-1 was stable for at least one year at −20 °C and −80 °C. In the same experiments glucagon recovery was reduced by almost 50% by freezing compared with immediate analysis, regardless of storage time ^[13]. Read carefully, that is not a license to refreeze: it says the loss for one of the two peptides arrived with the first freeze and did not accumulate with further cycles, and that two peptides in the same assay behaved differently from one another. It is a matrix experiment with inhibitors present, not a neat-buffer stock, and should not be generalized past that — but it does show that cycle count is not always the variable governing recovery.

What this means for aliquoting

The practical logic of aliquoting is not that freezing is uniquely harmful — it is that the number of times a molecule is subjected to any of the above should be a design decision, not an accident of how often someone opened the freezer. Size aliquots to one experiment's consumption and the count follows from arithmetic: a 2.0 mL reconstituted stock supporting an assay that uses 150 µL gives 2.0 mL ÷ 0.150 mL = 13.3, so 13 full aliquots with 50 µL left over — better handled as a fourteenth short tube than as a repeatedly disturbed remainder.

Beyond that, most of what is recommended is convention supported by mechanism rather than by controlled peptide studies, and it is more useful to say so than to imply otherwise. Low-binding tubes are supported for peptide recovery specifically ^[11]. A buffer that does not undergo selective salt crystallization is supported for pH stability on freezing ^{[7, 8, 10]}. Storage below Tg′ is supported for aggregation, in an antibody system ^[9]. The choice between −20 °C and −80 °C for a particular research peptide, the practice of thawing on ice rather than at room temperature, and the avoidance of auto-defrost freezers are reasonable extrapolations from the mechanisms above, not conclusions from experiments on that peptide.

What is not established

Several things a datasheet states with confidence are not settled in the literature:

• A cycle limit. No published work establishes a general threshold — two cycles, five cycles — beyond which a research peptide is compromised. Where cycle counts have been tested, the outcome depended on concentration and buffer ^[10] or showed no cycle dependence at all ^[13].
• Which mechanism dominates. Ice-surface denaturation, entrapped air interfaces, cold denaturation and buffer-driven pH shift are all supported by experiment, and which one controls a given system is molecule-dependent ^{[4, 6, 8]}.
• The direction of the freezing-rate effect. Slow freezing was optimal for one model protein and fast freezing for another in the same study ^[6].
• Adsorptive loss for a named sequence. The HPLC index and peptide length correlate with it but explain a minority of cases ^[11].
• Whether protein findings transfer to short peptides. This is the largest gap. The mechanisms are established on molecules with folded structure and functional readouts; the compounds in a peptide freezer mostly have neither.

None of this argues for handling material carelessly. It argues for recording what you did — cycle count, storage temperature, buffer, container type, date — so that when a stock behaves differently at week twelve than at week one, the record can distinguish a handling artifact from a result. For research use only. Not for human or veterinary use.