Lyophilised Peptide Stability and Cold Storage

Water as Reactant, Plasticiser, and Medium

Water is not a passive background in peptide degradation; it occupies three distinct roles. It is a stoichiometric reactant in hydrolytic steps, including the ring-opening of the succinimide intermediate that yields aspartate and isoaspartate from asparagine, and the cleavage of labile backbone amide bonds. It is a plasticiser, lowering the glass transition temperature of an amorphous matrix and thereby raising the rate at which molecules within that matrix can reorient and diffuse. And it is a medium, supplying the proton-transfer network on which general acid-base catalysis depends. Lyophilisation acts on all three roles at once. Ice is removed by sublimation under reduced pressure during primary drying, and a portion of the remaining non-frozen, sorbed water is removed by desorption at elevated shelf temperature during secondary drying. Residual water in a finished solid commonly falls between a few tenths of a percent and a few percent by mass, determined by Karl Fischer titration.

Residual moisture measured gravimetrically and water activity are not the same quantity and should not be used interchangeably. Water activity is the ratio of the vapour pressure of water above the material to the vapour pressure of pure water at the same temperature; it describes the escaping tendency of the water present rather than its amount. Two solids with identical Karl Fischer values can differ substantially in water activity depending on whether the water is dissolved in an amorphous bulk phase, held at crystal surfaces, or bound in a hydrate lattice. The relationship between the two is the moisture sorption isotherm, measured by dynamic vapour sorption. In amorphous solids the isotherm is typically steep and the water is absorbed into the bulk rather than adsorbed as a surface film, which is why a small mass fraction of water can have a disproportionate effect on the whole matrix. Degradation rates in dried peptide solids often track water activity and the mobility it confers more closely than they track total water content.

A note on scope. This article concerns the physical chemistry of a dried peptide solid held in a sealed container, and the analytical methods used to characterise its change over time. The preparation of solutions and any downstream use of the material fall outside that scope and are not treated anywhere below.

The Amorphous Glass: Tg, Tg-prime, and Molecular Mobility

A peptide freeze-dried from an aqueous solution containing a non-crystallising solute generally yields an amorphous solid rather than a crystalline one. The relevant thermal landmarks appear in two stages. During freezing, ice crystallises and the solutes are concentrated into an unfrozen phase that becomes progressively more viscous until it vitrifies; the glass transition of that maximally freeze-concentrated phase is denoted Tg-prime, with reported values near -32 degrees Celsius for sucrose systems and near -29 degrees Celsius for trehalose systems. Closely related is the collapse temperature, the temperature above which the drying cake loses its structure through viscous flow. Primary drying is conducted below the collapse temperature because a collapsed cake retains water inhomogeneously and dries unpredictably. After drying, the solid has its own glass transition temperature, Tg, which is a property of composition and of the residual water it contains.

Water is an unusually effective plasticiser because its own glass transition is very low, commonly cited near 136 K. Mixing rules of the Gordon-Taylor type predict that adding a low-Tg component to a high-Tg matrix depresses the mixture Tg steeply at low mass fractions. Hancock and Zografi, reviewing amorphous pharmaceutical solids, noted that on the order of one percent by weight of water can depress the glass transition of such a system by roughly ten to twenty kelvin. This is the mechanism by which a moisture ingress that looks trivial as a mass fraction can move a solid from a glassy state at its storage temperature to a rubbery one, with a corresponding increase in molecular mobility.

The glass transition is a useful design boundary, not a switch that turns chemistry off. Below Tg the large-scale, cooperative alpha relaxation is arrested on laboratory timescales, but chemical degradation continues. Yoshioka and Aso surveyed the literature correlating molecular mobility with chemical stability in amorphous pharmaceuticals and found the correlation to be system-dependent: in some systems degradation rates track global mobility through structural relaxation time, while in others local, sub-Tg beta relaxations or the mobility of small reactant species such as oxygen and water control the rate. Storage below Tg is therefore a necessary condition for a stable amorphous solid, not a sufficient one.

Two mechanistic hypotheses are conventionally invoked to explain why glass-forming excipients stabilise dried peptides and proteins in the solid state. The vitrification hypothesis attributes stabilisation to kinetic immobilisation in a high-viscosity matrix. The water replacement hypothesis attributes it to hydrogen bonding between the excipient and the solute that substitutes for the hydration shell lost during drying. The two are not exclusive and both appear to contribute. A related and frequently overlooked failure mode is physical: if an excipient present in the amorphous phase crystallises during storage, it expels the water dissolved in it into the remaining amorphous fraction, depressing that fraction’s Tg and accelerating its chemistry. The failure is self-accelerating, and is one reason powder X-ray diffraction belongs in a solid-state stability protocol alongside chemical assays.

Degradation Routes That Persist in the Dried Solid

Deamidation of asparagine is the most-studied covalent route. Under neutral and mildly alkaline conditions the backbone amide nitrogen of the following residue attacks the asparagine side-chain carbonyl to give a five-membered succinimide, which then hydrolyses to a mixture of aspartate and isoaspartate, typically in a ratio near one to three in favour of the isoaspartyl product. Geiger and Clarke established the succinimide route and its coupling to isomerisation and racemisation. The rate is strongly sequence-dependent: the residue on the carboxyl side of the asparagine dominates, with asparaginyl-glycyl the fastest common motif, followed by asparaginyl-seryl and asparaginyl-histidyl. Robinson and Robinson measured a library of several hundred asparaginyl pentapeptides at pH 7.4 and 37 degrees Celsius and found first-order half-times spanning roughly one day to more than a year across sequence context alone. Glutamine deamidates by an analogous six-membered glutarimide route but far more slowly. The mass change is +0.984 Da, which is small enough to require adequate mass resolving power to distinguish from isotopic overlap.

Aspartyl isomerisation proceeds through the same succinimide intermediate approached from the acid side, so aspartate can convert to isoaspartate without any deamidation event, and the planar succinimide permits racemisation to the D isomers. Aspartyl-glycyl is the classic fast motif. Because isoaspartate is isobaric with aspartate, it is invisible to mass measurement alone and must be resolved chromatographically or identified by fragment-ion evidence.

Solid-state hydrolytic chemistry is governed by the ionisation state the material carried out of solution, a phenomenon usually described as pH memory. The apparent protonation state of ionisable groups, and the buffer species that accompanied them, are largely retained through freezing and drying, so the relevant acidity for a solid is that of the last solution it occupied rather than anything measurable on the dry powder. Freezing itself can shift that state substantially. Gomez, Pikal, and Rodriguez-Hornedo showed that in sodium phosphate systems the selective precipitation of disodium hydrogen phosphate dodecahydrate after ice nucleation produces abrupt pH decreases at temperatures between about -0.5 and -4.0 degrees Celsius, a shift of more than a unit in some compositions. Potassium phosphate systems shift in the opposite direction and by less. The buffer present in the pre-freezing solution therefore prints an acidity onto the solid that persists for the whole of its storage life.

Oxidation does not require bulk water and consequently does not slow as sharply on drying as the hydrolytic routes do; in a well-dried solid it is frequently the dominant chemistry. Methionine thioether oxidises to the sulfoxide (+15.995 Da) and further to the sulfone; cysteine thiol proceeds through sulfenic, sulfinic, and sulfonic states and participates in disulfide exchange; tryptophan gives oxindolylalanine, N-formylkynurenine, and kynurenine; histidine and tyrosine are also susceptible, tyrosine additionally through dityrosine cross-linking. Li, Schoeneich, and Borchardt catalogued these mechanisms and their initiators: peroxide impurities carried in by excipients, trace transition metals catalysing Fenton chemistry, molecular oxygen in the container headspace, and light. Tryptophan and tyrosine absorb below roughly 310 nm, which is why photostability is evaluated separately and why the light-transmission properties of packaging form part of a stability argument.

Diketopiperazine formation truncates the peptide from the N-terminus. The unprotonated alpha-amino group of residue one attacks the carbonyl of residue two in an intramolecular aminolysis, expelling the remainder of the chain and leaving a cyclic 2,5-diketopiperazine. Because only the free base form is nucleophilic, the reaction is base-catalysed and its pH-rate profile tracks the ionisation of the N-terminal amine; Goolcharran and Borchardt demonstrated this with model peptides carrying proline at the second position, and also showed general base catalysis by buffer components, meaning buffer identity and concentration are themselves rate variables. Proline at the second position is a well-known accelerant because the cis amide population and ring geometry favour the required conformation. Sequences with glycine or proline near the N-terminus therefore warrant explicit chromatographic attention for a truncated species and its cyclic counterpart.

Several additional routes deserve mention. Aspartyl-prolyl backbone bonds hydrolyse preferentially under acidic conditions. Reducing sugars condense with lysine side-chain amines and the N-terminal amine in the Maillard reaction, which is the principal reason non-reducing disaccharides are preferred as glass formers; sucrose can itself hydrolyse to glucose and fructose under acidic conditions and thereby generate reducing sugars in situ. Under alkaline conditions beta-elimination from cystine and disulfide scrambling become relevant. Physical aggregation can occur in the solid state without any covalent change and requires a size-based method to detect. The overall consequence is that the degradant profile of a dried peptide is usually not a slowed-down copy of its solution profile; the ranking of routes changes, and a method developed only against solution stress may miss the species that actually accumulate in the solid.

Temperature, Cold Chain, and Thermal Cycling

The temperature dependence of these reactions is steep enough that storage temperature is usually the single largest controllable variable. Activation energies reported for peptide degradation routes commonly lie between about 50 and 120 kJ/mol. Taking 83 kJ/mol as an illustrative value, the Arrhenius relation gives a rate roughly threefold higher for each 10 K increase near ambient temperature, about an elevenfold reduction on moving from 25 to 5 degrees Celsius, and a reduction of two to three orders of magnitude on moving from 25 to -20 degrees Celsius. These figures are arithmetic illustrations of the exponential form, not predictions for any particular material; the extrapolation caveats in the final section apply with full force at the subzero end of that range.

A cold chain is a control and documentation regime, not merely a cold room. Its elements are continuous temperature monitoring with calibrated loggers, defined alarm limits, qualified transport packaging with known thermal mass, and a written basis for accepting or rejecting excursions. Mean kinetic temperature, defined in ICH Q1A(R2), is the single derived temperature that would produce the same extent of degradation as a measured non-isothermal history under an assumed activation energy; it weights excursions exponentially rather than arithmetically, which is why a short warm excursion can dominate a long record of nominal storage. Mean kinetic temperature is a summary statistic and depends on the assumed activation energy, so it carries little meaning reported without that assumption.

For a sealed, dried solid, the practical hazard in thermal cycling is usually not a phase change in the peptide but water. A container equilibrated at -20 degrees Celsius has interior surfaces well below the dew point of ordinary laboratory air, and exposure of those surfaces to ambient atmosphere before thermal equilibration produces condensation, which a hygroscopic amorphous solid takes up rapidly. The resulting moisture increase can be large relative to the residual water the drying cycle was designed to achieve, and it is not reversed by returning the container to cold storage. Thermal cycling of an unopened container also redistributes moisture internally, because sorption isotherms are temperature-dependent and the elastomeric closure, any desiccant present, and the cake do not share a single equilibrium water content across a temperature swing. Repeated cycling therefore has a cumulative effect on the moisture state of the solid even when the seal is never broken.

The separate literature on freezing and thawing of aqueous solutions concerns process intermediates and analytical sample handling rather than a stored dried solid, but its mechanisms are worth naming because they explain several of the solid-state observations above. Freezing generates a large ice-water interfacial area at which surface-induced unfolding can occur; it cryoconcentrates solutes in the unfrozen phase to very high levels, raising bimolecular reaction rates; and it shifts the acidity of that phase through selective salt precipitation as described earlier. Damage in those systems accumulates with the number of cycles rather than with the total time frozen, which makes cycle count the informative variable.

Container, Closure, Headspace, and Desiccants

The stability of a dried peptide is inseparable from the package that maintains its water and oxygen environment. Type I borosilicate glass is effectively impermeable to water vapour; the elastomeric closure and the seal interface constitute the ingress path, and the ingress rate is a measurable property of the container-closure system expressed as mass of water per container per year at a defined external humidity and temperature. Container-closure integrity is verified by deterministic methods such as vacuum decay, helium mass-spectrometry leak detection, or high-voltage leak detection, and can be monitored non-destructively over a stability study by laser-based headspace analysis, which measures headspace oxygen, moisture, and pressure through the glass without opening the container.

The closure is also a moisture source, not only a barrier. Elastomeric stoppers absorb substantial water during steam sterilisation and release it slowly into the container interior afterwards. Pikal and Shah measured this transfer at 5, 25, and 40 degrees Celsius into hygroscopic amorphous solids and showed that equilibration of a dried cake with stopper-borne moisture can raise the water content of the dried solid enough to matter for stability, with the rate governed by temperature. Post-sterilisation stopper drying conditions are consequently part of the stability argument for a finished solid, and a rise in Karl Fischer water over the first months of a stability study is often attributable to the closure rather than to external ingress.

Headspace composition governs the oxidative routes. Backfilling with nitrogen or argon lowers the quantity of oxygen available for methionine, cysteine, and tryptophan oxidation; because that quantity is finite in a sealed container, oxidation can appear pseudo-zero-order until the headspace is depleted and then plateau, which complicates rate fitting if only two time points are taken. Light exposure is controlled by amber glass or by opaque secondary packaging, and the relevant absorbance is that of the aromatic side chains rather than of the peptide backbone.

Desiccants are a secondary-package tool with distinct materials and distinct behaviour. Silica gel sorbs across a broad humidity range and releases water as temperature rises, so it buffers humidity rather than fixing it at a low value. Molecular sieves of 3A or 4A pore size hold water tightly at very low relative humidity and are effectively irreversible under ambient conditions, which makes them the aggressive choice. Clay desiccants sit between the two. Sizing is a capacity calculation: total water ingress over the intended storage period plus the water initially present in the package must not exceed the desiccant capacity at the equilibrium humidity of interest. One caveat is well documented in the solid-state protein literature and applies to aggressive drying generally: the relationship between residual water and degradation rate is not always monotonic, and some systems exhibit a minimum degradation rate at a non-zero water content, with rates rising again as the solid is dried below it. Over-drying is a design error, not merely wasted effort.

Measuring Stability: Stability-Indicating Methods and Study Design

A stability-indicating method is one demonstrated to separate the intact peptide from every degradant that accumulates under the conditions of interest, so that a fall in parent content is detected as a fall in parent content rather than being masked by a co-eluting product. Specificity is established by forced degradation: separate thermal, humidity, acidic, alkaline, oxidative, and photolytic challenges, typically targeting five to twenty percent loss of parent so that primary degradants dominate over secondary ones. Photostability challenges follow ICH Q1B. The evidence expected from such a study is chromatographic resolution of each induced degradant, peak homogeneity assessed by diode-array or mass-spectrometric purity, and a mass balance that accounts for the lost parent within a defensible tolerance. Method validation follows ICH Q2, and a method that has not been challenged against solid-state stress specifically cannot be assumed to be stability-indicating for a dried solid.

Reversed-phase HPLC with ultraviolet detection remains the workhorse, but two of the routes above resist it. Isoaspartate is isobaric with aspartate, so it must be resolved chromatographically, which frequently requires shallow gradients, low-pH mobile phases, or an orthogonal ion-exchange separation; it can also be identified by electron-transfer dissociation, which yields diagnostic fragment ions characteristic of the isoaspartyl backbone, or quantified enzymatically using protein isoaspartyl methyltransferase. Deamidation itself carries only a +0.984 Da shift, which demands sufficient resolving power to separate it from the isotopic envelope of the parent. Oxidation products are easier by mass, at +15.995 Da for the methionine sulfoxide and characteristic shifts for the tryptophan products, but positional assignment still requires peptide mapping.

Chemical assays alone do not describe a solid. A complete solid-state stability protocol pairs them with physical measurements: Karl Fischer titration for residual water, modulated differential scanning calorimetry for the glass transition and for its movement over time, powder X-ray diffraction for crystallinity change in the matrix, dynamic vapour sorption for the moisture isotherm, second-derivative infrared analysis of the amide I band for solid-state conformation, size-exclusion chromatography for soluble aggregate, imaged capillary isoelectric focusing or capillary electrophoresis for charge heterogeneity, and visual assessment of cake structure. A change in Tg or the appearance of diffraction peaks frequently precedes any measurable change in the chemical assay and is the more sensitive early indicator.

Study design determines what the data can support. ICH Q1A(R2) sets the framework of long-term, intermediate, and accelerated conditions with defined time points, and permits bracketing and matrixing designs where justified. Two design choices matter disproportionately. First, an isoconversion approach, in which the time required to reach a fixed extent of degradation is measured at each condition, is generally more robust than fitting rate constants to fixed time points, because it keeps the measured quantity within the well-characterised region of the analytical method rather than near its limit of quantitation. Second, replication and the statistical treatment of the resulting fit govern any extrapolation: ICH Q1E describes the poolability testing and the use of the confidence bound on the regression line, rather than the line itself, when projecting to an acceptance criterion.

Arrhenius Extrapolation and Where It Fails

The Arrhenius relation, expressed as the natural logarithm of the rate constant varying linearly with reciprocal absolute temperature, is the standard basis for projecting a long-term rate from short high-temperature studies. Two practical points about its use recur. It should be fitted over at least three and preferably four temperatures, because a two-point determination provides no residual with which to detect curvature. And the uncertainty in the fitted activation energy propagates exponentially into the extrapolation, so a modest error in slope over a 30 K experimental span becomes a large error when projected across a 40 K gap to a cold storage condition. The confidence interval on the projected rate, not the point estimate, is the quantity that carries the information.

Several failure modes are specific to amorphous solids and are easy to miss. If the experimental temperature range crosses the glass transition of the solid, the system changes from a glassy to a rubbery state and the temperature dependence is no longer Arrhenius; behaviour above Tg is better described by Williams-Landel-Ferry-type coupling to structural relaxation, and a straight-line fit through points on both sides of Tg produces a slope with no physical meaning. If an excipient crystallises at the elevated temperature but not at the storage temperature, the composition being studied is not the composition being predicted. Moisture is likewise not constant across temperatures in a sealed container, because sorption equilibria between cake, closure, and headspace shift with temperature, so accelerated samples may carry a different effective water activity than long-term samples. Oxidation limited by a finite headspace oxygen content changes apparent order as that oxygen is consumed. And a mechanism that is minor at storage temperature but has a high activation energy can dominate the accelerated data, giving a degradant profile that never appears in real time.

Because humidity is a second independent variable with its own effect on rate, accelerated designs for solids frequently use a humidity-modified Arrhenius form in which an empirical term linear in relative humidity is added to the exponent. Waterman and Adami reviewed accelerated ageing methods for pharmaceuticals on this basis, covering hydrolysis, oxidation, reaction with excipient impurities, photolysis, and protein denaturation, and set out the conditions under which such extrapolation is and is not defensible. The consistent conclusion across that literature is worth stating plainly: accelerated data generate hypotheses about which degradation routes matter and roughly how fast, and they generate degradants against which a method can be developed, but real-time data at the intended storage condition remain the evidence. Where accelerated and real-time results diverge, the divergence is information about mechanism, not a nuisance to be averaged away.

References

1. Manning MC, Chou DK, Murphy BM, Payne RW, Katayama DS. Stability of protein pharmaceuticals: an update. Pharmaceutical Research. 2010;27(4):544-575. DOI 10.1007/s11095-009-0045-6.
2. Wang W. Lyophilization and development of solid protein pharmaceuticals. International Journal of Pharmaceutics. 2000;203(1-2):1-60. DOI 10.1016/S0378-5173(00)00423-3; PMID 10967427.
3. Carpenter JF, Pikal MJ, Chang BS, Randolph TW. Rational design of stable lyophilized protein formulations: some practical advice. Pharmaceutical Research. 1997;14(8):969-975. DOI 10.1023/A:1012180707283; PMID 9279875.
4. Hancock BC, Zografi G. Characteristics and significance of the amorphous state in pharmaceutical systems. Journal of Pharmaceutical Sciences. 1997;86(1):1-12. DOI 10.1021/js9601896; PMID 9002452.
5. Yoshioka S, Aso Y. Correlations between molecular mobility and chemical stability during storage of amorphous pharmaceuticals. Journal of Pharmaceutical Sciences. 2007;96(5):960-981. DOI 10.1002/jps.20926; PMID 17455355.
6. Geiger T, Clarke S. Deamidation, isomerization, and racemization at asparaginyl and aspartyl residues in peptides. Succinimide-linked reactions that contribute to protein degradation. Journal of Biological Chemistry. 1987;262(2):785-794. PMID 3805008.
7. Robinson NE, Robinson AB. Molecular clocks. Proceedings of the National Academy of Sciences USA. 2001;98(3):944-949. DOI 10.1073/pnas.98.3.944.
8. Li S, Schoeneich C, Borchardt RT. Chemical instability of protein pharmaceuticals: mechanisms of oxidation and strategies for stabilization. Biotechnology and Bioengineering. 1995;48(5):490-500. DOI 10.1002/bit.260480511.
9. Goolcharran C, Borchardt RT. Kinetics of diketopiperazine formation using model peptides. Journal of Pharmaceutical Sciences. 1998;87(3):283-288. DOI 10.1021/js970325m.
10. Gomez G, Pikal MJ, Rodriguez-Hornedo N. Effect of initial buffer composition on pH changes during far-from-equilibrium freezing of sodium phosphate buffer solutions. Pharmaceutical Research. 2001;18(1):90-97. DOI 10.1023/A:1011082911917; PMID 11336359.
11. Pikal MJ, Shah S. Moisture transfer from stopper to product and resulting stability implications. Developments in Biological Standardization. 1992;74:165-179. PMID 1592166.
12. Waterman KC, Adami RC. Accelerated aging: prediction of chemical stability of pharmaceuticals. International Journal of Pharmaceutics. 2005;293(1-2):101-125. PMID 15778049.
13. ICH Harmonised Tripartite Guideline Q1A(R2), Stability Testing of New Drug Substances and Products, 2003. Source of the mean kinetic temperature definition and of the long-term, intermediate, and accelerated study framework.
14. ICH Harmonised Tripartite Guideline Q1B, Photostability Testing of New Drug Substances and Products, 1996.
15. ICH Harmonised Tripartite Guideline Q1E, Evaluation for Stability Data, 2003.
16. ICH Harmonised Guideline Q2(R1), Validation of Analytical Procedures: Text and Methodology, 2005; revised as Q2(R2), 2023.