The air–liquid and solid–liquid boundaries of a vial are the only regions of a peptide sample where a hydrophobic face can bury itself against something other than another peptide molecule, and that is where interfacial damage begins. Surface-active excipients are typically added in protein formulations to prevent undesired interactions of proteins on surfaces and the particle formation and aggregation that follow [1]. The question that decides whether one will help your sample is mechanistic: is the surfactant occupying the interface before your peptide reaches it, or is it binding the peptide in bulk? Polysorbate 20, polysorbate 80 and poloxamer 188 do not answer that the same way.
Occupying the interface
At the air–water interface the two mechanisms leave different mechanical signatures. Adsorption of a monoclonal antibody from surfactant-free solution created a monolayer with significant viscoelasticity — which, the authors note, can influence measurements of bulk antibody solution viscosity — whereas polysorbate 20 adsorption produced an interface with negligible interfacial viscosity that protected the air–water interface from antibody adsorption. The comparison combined surface tensiometry, interfacial rheology, capillary viscometry and neutron reflectometry [2]. That is an easily missed consequence for anyone measuring bulk viscosity on a surfactant-free solution, and it belongs alongside the concentration and shear-rate effects covered in the viscosity guide.
How much surfactant it takes to hold that interface is not a fixed number. Pendant bubble tensiometry and X-ray reflectivity on polysorbate 80 with a monoclonal antibody showed that as antibody concentration was raised across 0.5–2.5 mg/mL, an increased polysorbate 80 concentration — still below its critical micelle concentration — was required to prevent the antibody from co-existing with surfactant in the surface layer [1]. The same study compared polysorbate 80 against a different amphiphile, N-myristoyl phenylalanine-N-polyetheramine diamide (FM1000), and found that FM1000 saturated the interface faster and blocked a larger percentage of interfacial area, so a lower bulk concentration of it sufficed [1].
The solid–liquid interface is a different competition
Optical waveguide lightmode spectroscopy at hydrophilic silica–titania surfaces was used to follow the adsorption of poloxamer 188, polysorbate 80 and polysorbate 20, with and without chicken egg white lysozyme and, in separate experiments, recombinant human granulocyte colony-stimulating factor, in both sequential and competitive modes. Polysorbates 80 and 20 inhibited protein adsorption only through preferential location at an interface for which they showed sufficient affinity, and not by forming less surface-active protein–surfactant complexes. Poloxamer 188 did the reverse: it inhibited adsorption by entering into protein–surfactant complexes of low adsorption affinity, and not by preferential location at the interface [3].
Those are not interchangeable troubleshooting moves. If material is disappearing onto labware — the loss problem treated in the adsorption guide — a polysorbate can only help if it reaches and holds that particular surface, and affinity for a hydrophilic silica-like surface is not affinity for polypropylene. A poloxamer, on that evidence, works through a solution-phase complex instead, so which surface is involved is less of an objection.
The CMC, and why it is a weaker rule than it reads
Polysorbates are not single compounds, so the critical micelle concentration is a population property. Small-angle neutron scattering of polysorbate 20 and polysorbate 80 and of their monoester and diester fractions at 4, 22, 40 and 50 °C returned aggregation number, radius of gyration, core radius, critical micelle concentration, shell thickness and shell hydration for each. The critical micellar concentration of the polysorbate 20 monoester fraction was significantly higher at 4 °C than that of the diester fraction or of the original all-laurate polysorbate 20 material, and the work was framed around identifying which components drive the temperature-dependent behaviour of these surfactants in formulations [4]. A CMC quoted for "polysorbate 20" without a temperature and without an ester profile is an average over species that do not behave alike.
Working concentrations sit low. A 60 MHz benchtop quantitative NMR method was developed to quantify polysorbate 20, polysorbate 80 and poloxamer 188 at and below target concentrations of 0.025% w/v, directly in formulations containing histidine and sucrose [5]. The familiar rule of thumb is to set that working concentration above the CMC. Two results complicate it. The polysorbate 80 titration described above ran below the CMC throughout, and its authors concluded that the CMC is not the critical factor indicating a surfactant's ability to inhibit protein adsorption, particularly across chemically different surfactants [1].
The second used agitation stress with two particle-sizing routes: three-dimensional homodyne light detection for submicron particles and flow imaging for subvisible particles. The surfactant concentration needed to suppress submicron particle formation was generally higher than that needed to suppress subvisible particles and, in contrast to previous studies as the authors put it, that concentration did not correlate with the CMC. The thresholds depended on the specific protein–surfactant combination, and the authors suggest that factors other than CMC, potentially including interfacial adsorption and container-surface interactions, may contribute under this stress model [6]. Neither result makes the CMC irrelevant, but together they mean the working concentration is something you titrate against the stress you actually apply.
What has been shown with peptides
Most of this literature is antibody literature; two peptide results are worth holding onto. In a thioflavin T assay on a GLP-1-like peptide designated G48, run from 0 to 600 µM at pH 3.7, 7.4 and 8.5, the aggregation profile was pH-dependent: at pH 3.7, below the peptide's pI of about 5, the kinetics were those of amorphous aggregates forming by a nucleation-independent mechanism, while at pH 7.4 and 8.5 they were nucleation-dependent. Incorporating polysorbate 80 suppressed fibrillation of G48 at all three pH values and maintained the native peptide conformation, whereas a phenolic co-formulant, ferulic acid, had minimal effect on fibril growth [7].
The second is a caution. Human insulin aggregation was induced at low pH in the presence of varying polysorbate 80 concentrations. The surfactant slowed but did not prevent aggregation; amyloid spherulites formed under all conditions, and above the surfactant's critical micelle concentration they contained a higher content of intermolecular β-sheet and showed denser packing, giving a more stable aggregate. Molecular dynamics simulations in the same work suggested a tendency for insulin to form dimers in the presence of the surfactant, indicating a change in protein–protein interactions [8]. Surfactants, the authors conclude, not only alter aggregation kinetics but also affect the physicochemical properties of whatever aggregates do form — so flattening an aggregation curve is not by itself evidence that the sample is in better shape.
The surfactant is the least stable thing in the tube
Hydrolysis of the ester bond is often enzymatic, and in recombinant material the enzyme is a residual host-cell protein. In monoclonal antibody drug products containing multiple lipases, complete inhibition of polysorbate 20 degradation by the lysosomal acid lipase inhibitor lalistat2 identified lysosomal acid lipase, rather than lipoprotein lipase, as the cause; several other antibodies carrying comparable lipoprotein lipase levels of 0.5–1.5 ppm showed no polysorbate 20 degradation in the absence of lysosomal acid lipase [9]. That root cause is specific to expressed material carrying host-cell protein and does not transfer to a synthetic peptide preparation. Oxidation does.
What hydrolysis produces is the problem. Polysorbate 20 degraded over roughly 18–36 months of storage at 2–8 °C in two monoclonal antibody formulations; the released free fatty acids accumulated and ultimately precipitated to form particles, and a qualified method quantified capric, lauric, myristic, palmitic/oleic and stearic acids in both the soluble and the insoluble fraction of the formulations. Tracking free fatty acid on real-time storage appeared to aid in identifying batches with a higher probability of particulate formation on extended storage at 5 °C [10].
Oxidation attacks the fatty acid rather than the ester bond. Linoleic acid accounts for up to 18% of all fatty acids in multi-compendial grade polysorbate 80, and oxidative degradation of polyunsaturated fatty acids in polysorbate was shown to give α,β-unsaturated carbonyls. 4-hydroxynonenal was confirmed as a degradant of oxidatively stressed polysorbate 80; it was not directly detectable in the polysorbate 80 matrix by UV or MS, so quantification required derivatisation with 2,4-dinitrophenylhydrazine, salting-out assisted liquid–liquid extraction of the hydrazone, and multiple reaction monitoring. Butylhydroxytoluene prevented its formation under oxidative stress [11]. The reason that degradant was pursued at all is that 4-hydroxynonenal is highly reactive towards nucleophilic amino acids, and so could potentially modify the molecule the surfactant was added to protect [11].
The two routes are not equivalent in consequence. Esterase-hydrolysed and oxidised polysorbate 80 were compared by mass spectrometry and by Langmuir trough adsorption isotherms: hydrolysed polysorbate 80 showed slower surface adsorption rates while oxidised polysorbate 80 showed increased surface activity, and the critical micelle concentration remained unchanged in that comparison. Formulated with a monoclonal antibody and shaken, hydrolysed polysorbate produced a large increase in particle formation, with monomer loss by size-exclusion chromatography comparable to formulations containing no polysorbate at all, whereas oxidised polysorbate 80 was still protective. The free fatty acids released by hydrolysis formed insoluble particles during mechanical agitation, and those particles stimulated protein aggregation [12].
Which ester species is lost also matters. Two immobilised, hydrolytically active lipases were used to generate partially degraded polysorbate 20, one targeting mainly monoester species and the other higher-order esters. The CMC increased slightly on monoester degradation and significantly on higher-order ester degradation; the degradation pattern significantly affected the dynamic surface tension of an antibody formulation, while changes in equilibrium surface tension were mainly caused by adsorption of free fatty acids onto the air–water interface; and in an agitation protection study, monoester degradation produced soluble antibody aggregates and proteinaceous particles [13]. Read against the unchanged CMC reported for esterase-hydrolysed polysorbate 80 [12], the reconciliation is that whether the CMC moves depends on which ester species the particular enzyme removes.
Measuring it is harder than adding it
Polysorbates have no chromophore, which is the root of the analytical difficulty: they are hard to detect with high sensitivity and simplicity at once. Charged aerosol detection gives a highly sensitive response to non-volatile and semi-volatile species — the esters, acids, oxidised aldehydes and contaminant ions present in polysorbate — which is why liquid chromatography coupled to charged aerosol detection is reviewed as a route to the qualitative and quantitative analysis of polysorbate, its composition profile and the reasons for its degradation [14].
Protein in the sample is the second problem. The common mixed-mode ion-exchange reversed-phase HPLC method with evaporative light scattering detection can be adversely affected at high protein concentration through high sample viscosity, gel formation, column clogging, interfering peaks and loss of accuracy; an EDTA-mediated ethanol protein precipitation step was developed to remove the protein first, and was applied to antibody formulations across 10–250 mg/mL [15]. The benchtop qNMR route noted above sidesteps the separation entirely and is non-destructive, quantifying the three surfactants directly in formulations containing histidine, sucrose and one of a monoclonal antibody, an antibody–drug conjugate or an Fc-fusion protein [5].
Every one of those methods was developed against a protein matrix. A concentrated peptide is a different interference problem: small, often highly soluble, and not necessarily removed by a precipitation chemistry tuned to drop an antibody out of solution. Transfer any of them to a peptide sample only with your own recovery and spike-return data.
SDS is a different reagent doing a different job
Six surfactants — sodium dodecyl sulphate, dodecyl maltoside, polysorbate 20, and the biosurfactants rhamnolipid, sophorolipid and surfactin — were compared against ZP-GA-1, a stable version of glucagon, using pyrene fluorescence, circular dichroism and isothermal titration calorimetry. All six induced α-helical structure in the peptide, SDS having the biggest impact and polysorbate 20 the smallest. SDS kept ZP-GA-1 solubilised for more than 48 days, against 29 days in dodecyl maltoside, 3 days in polysorbate 20 and 0 days in buffer, and much less SDS than dodecyl maltoside, polysorbate 20 or biosurfactant was needed to redissolve aggregated peptide. Isothermal titration calorimetry confirmed the ordering, with very strong interactions for SDS and very weak ones for polysorbate 20. The authors conclude that simpler surfactant structures promote stronger peptide interactions, and that stronger enthalpic interactions improved absolute solubility, though interaction strength correlated with absolute solubility rather than with the kinetics of precipitation [16].
That is the distinction in one experiment. The polysorbate's interaction with this peptide was very weak [16], consistent with the interfacial-occupancy mechanism established for polysorbates at solid surfaces [3]; SDS binds hard enough to remodel the conformation [16]. That conformational grip is what makes SDS a structural reagent rather than a formulation one. The antimicrobial peptide brevinin-2-related peptide was studied in SDS and in dodecylphosphocholine micelles by fluorescence and circular dichroism: its N-terminal tryptophan interacted with the hydrophobic core of the micelles, interaction with the micelles induced an α-helix conformation, and solution structures determined by NMR in the two micelle types showed significant conformational changes in the residues connecting the N-terminal and C-terminal helices [17].
Same peptide, same laboratory, two structures, and what differed was the micelle. A helical content measured in SDS is a helical content in SDS. It licenses a statement about the peptide's conformational behaviour in an anionic amphiphilic environment; it does not license a statement about the conformation in your aqueous buffer, and if you need both numbers you have to run both samples.
What to record
For any surfactant result: the surfactant and its concentration expressed relative to a CMC measured under your own conditions rather than one read from a table; the interface the stress actually loaded, since agitation, filtration and container contact are not the same experiment; the analyte class, because an antibody result is not a peptide result; and the age and storage temperature of the surfactant-containing solution, because on the timescale of a real-time stability study the excipient is degrading alongside the analyte [10, 13].