Sterile filtration of peptide solutions: membrane chemistry, adsorptive loss, and what passes through

Filtration is usually described by a single number. A 0.22 µm filter is understood as the sterilizing-grade cutoff, a 0.45 µm filter as clarifying, and the choice between them is treated as the whole decision. For peptide solutions this framing is incomplete in a way that has measurable consequences. Pore size determines what particles are retained; the membrane polymer determines what dissolved peptide is retained, what extractables enter the filtrate, and how much of the nominal concentration survives the transfer. At the low concentrations common in analytical and research handling, the polymer choice can account for a larger loss than any other single step in sample preparation.

This post looks at what the common sterilizing-grade membrane chemistries do to peptides in solution, why the loss is concentration-dependent rather than fixed, and what the literature on filter validation says about characterizing it rather than assuming it away.

What “0.22 µm” actually specifies

The sterilizing-grade designation is a performance claim, not a geometric one. A membrane qualifies by demonstrating retention of a defined challenge organism — conventionally Brevundimonas diminuta at a minimum of 10⁷ CFU per square centimeter of effective filtration area — under specified process conditions. The 0.22 µm figure is a nominal descriptor attached to membranes that pass that challenge, not a measured maximum pore diameter. Actual pore size distributions are broad, asymmetric, and depend on the casting process.

Two consequences follow. First, retention is not purely sieving: adsorptive capture on pore walls contributes to bacterial retention, which is why the challenge test is run under defined flow and pressure rather than as a static measurement. Second, the same adsorptive surface that helps retain organisms is available to bind dissolved peptide. The mechanism that makes the filter work is the mechanism that causes analyte loss. These cannot be separated by choosing a different pore rating.

The three common membrane polymers

Sterilizing-grade filters used with aqueous peptide solutions are dominated by three chemistries, and their behavior toward peptides differs substantially.

Polyethersulfone (PES) is the current default for aqueous biological work. It is intrinsically hydrophilic, has high flow rates at a given pressure differential, and shows comparatively low protein and peptide binding without surface modification. Its extractables profile is well characterized. For most research peptide solutions in aqueous buffer, PES is the reasonable starting choice, and studies of protein recovery through PES membranes generally report the smallest losses among unmodified hydrophilic membranes.

Polyvinylidene fluoride (PVDF) is intrinsically hydrophobic and is sold in both untreated and surface-modified hydrophilic forms. The hydrophilic-modified grades are marketed for low protein binding and perform well in that role. Untreated PVDF is a different material entirely for this purpose — hydrophobic peptides partition onto it readily, and it should not be used interchangeably with the modified grade. The distinction is frequently obscured on packaging, where both are labeled simply as PVDF.

Cellulose esters — mixed cellulose ester (MCE) and cellulose acetate (CA) — are hydrophilic and historically common. Cellulose acetate is generally characterized as low-binding for proteins. Mixed cellulose ester binds proteins more substantially and is better suited to microbiological work, where the analyte is the retained organism rather than the filtrate, than to recovering a dissolved peptide quantitatively.

Nylon and polytetrafluoroethylene (PTFE) appear in laboratory filter kits and warrant a note. Nylon has high protein and peptide binding and is a poor choice when filtrate concentration matters. PTFE is used for organic solvents and requires pre-wetting for aqueous solutions; it is not a general aqueous peptide filter.

Why loss is concentration-dependent

Adsorptive loss on a membrane behaves as a surface saturation phenomenon, not as a fixed percentage. The membrane presents a finite number of binding sites. Once those sites are occupied, subsequent peptide passes through largely unretained. This produces a characteristic pattern: percentage recovery is poor at low concentrations and improves as concentration rises, because the same absolute quantity is being lost from an increasingly large total.

The practical implication is that a recovery figure measured at one concentration does not transfer to another. A membrane that shows 98% recovery of a peptide at 1 mg/mL may show substantially lower recovery of the same peptide at 10 µg/mL, and the difference is not experimental error — it is the same absolute adsorption expressed against a smaller denominator. Recovery data reported without the concentration at which it was determined is uninterpretable.

This is the same surface-saturation behavior that governs adsorption to vial walls, pipette tips, and transfer tubing, and it compounds. A dilute peptide solution that passes through a filter, into a vial, and through a pipette tip encounters three sequential adsorptive surfaces. Each removes an absolute quantity, and at sufficiently low concentration the cumulative loss can be a substantial fraction of what was nominally prepared.

Two mitigations appear consistently in the literature. The first is discarding a defined initial volume of filtrate — the sacrificial or pre-saturation volume — so that the membrane’s binding sites are occupied by peptide that is deliberately not collected, and the retained fraction reaches equilibrium before collection begins. The second is inclusion of a carrier protein or a non-ionic surfactant such as polysorbate 20 at low concentration, which competes for binding sites. The second approach is only available where the additive does not interfere with the downstream analytical method, which frequently it does.

Extractables, leachables, and the filtrate

Filtration adds material as well as removing it. Membranes, support layers, housings, and adhesives all contribute extractables — compounds that can be forced out of the device under aggressive laboratory conditions — of which some subset appears as leachables under actual use conditions. For peptide work the relevant concerns are compounds that interfere with analysis or that react with the peptide.

The most frequently discussed class is oxidizing species. Some filter devices, particularly those that have been sterilized by certain methods or stored for extended periods, can release trace peroxides or other oxidants into the first volume of filtrate. For peptides containing methionine, cysteine, or tryptophan, this represents a direct chemical risk, and oxidation of these residues during sample preparation has been documented in protein formulation work. The same sacrificial-volume practice that addresses adsorption also reduces exposure to the highest-concentration leachable fraction, since leachables are most concentrated in the initial filtrate.

A second class is UV-absorbing extractables, which can raise apparent A280 or A214 readings and inflate a concentration determined by UV spectrophotometry. Where UV quantitation follows filtration, a filtrate blank — buffer passed through an identical filter — is the appropriate control, and it is distinct from a buffer blank that has not been filtered.

Characterizing rather than assuming

For work where filtrate concentration is a reported number rather than an operational convenience, the recovery through the specific filter, at the specific concentration, with the specific buffer, is a parameter to be measured. The measurement is straightforward: quantify the solution before filtration, quantify the filtrate, and express the ratio. Doing this once for a given peptide-membrane-buffer-concentration combination produces a figure that can be applied to subsequent runs, provided nothing in that combination changes.

Buffer composition matters more than it might appear. Ionic strength modulates electrostatic contributions to adsorption; pH relative to the peptide’s isoelectric point determines net charge and therefore the sign of any electrostatic interaction with a charged membrane surface. A peptide near its pI, carrying minimal net charge, often shows different adsorptive behavior than the same peptide in a buffer two units away. Recovery data generated in one buffer should not be assumed to transfer to another.

Filter area is a further variable. A larger membrane area presents more binding sites and therefore a larger absolute loss, which is the opposite of the intuition that a larger filter is safer. Where the volume permits, matching the device size to the volume being filtered reduces loss.

Where filtration sits relative to other steps

Sterilizing filtration of a reconstituted peptide solution is a specific operation with a specific purpose, and it is worth being clear about what it does and does not accomplish. It removes microorganisms and particulate matter above the retention rating. It does not remove dissolved degradation products, does not reverse aggregation that has already produced soluble oligomers below the retention threshold, and does not address endotoxin, which passes freely through a 0.22 µm membrane and requires either dedicated endotoxin-removal chemistry or control at the source.

Filtration also interacts with aggregation measurement in a way that is easy to overlook. Filtering a sample before a purity or particulate determination removes exactly the population that the determination was intended to characterize. Where sub-visible particulate matter or large aggregates are the analyte, filtration is a confounder rather than a preparation step, and the sequence of operations determines whether the resulting number means anything.

The membrane is not an inert boundary that solution passes through unchanged. It is a high-surface-area polymer in intimate contact with the peptide, for a defined contact time, under pressure — a set of conditions that would be recognized as an interaction if it were described in any other context. Treating the choice of that polymer as a consumables decision rather than a method parameter is where most of the unexplained loss between a prepared concentration and a measured one tends to originate.