Sterilizing Grade Filtration for Research Peptide Vials: Bacterial Retention and the Integrity Test No COA Reports

Sterilizing grade filtration is the step that removes bacteria from a peptide solution immediately before it enters the vial, and it is qualified by a bacterial challenge test rather than by any measurement made on the finished product. A certificate of analysis reports the chemical purity of the material that came out of that filter. It does not report whether the filter was integral when the batch ran, which membrane polymer was used, or how much peptide the membrane absorbed on the way through.

Research peptides are almost always filled from solution and then freeze dried. Because the solution cannot survive terminal heat sterilization without degrading, the only practical microbial control is passage through a membrane rated to retain bacteria. That single unit operation carries the entire microbiological burden of the batch, and it generates its own records, which sit in the manufacturing file rather than on the certificate the buyer receives.

What Sterilizing Grade Filtration Actually Certifies

A filter earns the description sterilizing grade by passing a standardized bacterial retention test, not by having pores of a measured size. Under ASTM F838, a membrane is challenged with Brevundimonas diminuta ATCC 19146 at a minimum of 10⁷ colony forming units per square centimetre of effective filtration area, and the effluent must be sterile. That corresponds to a retention of greater than seven log, or 99.99999 percent. The organism was selected in the 1970s precisely because it is small, and cultures grown under nutrient limitation produce cells small enough to be a demanding test article.

Two things follow from this that are frequently misread. First, the qualification applies to a filter type and a defined set of process conditions, not to the individual filter that produced any given batch. Second, the 0.2 micrometre designation itself is not a direct measurement. As Lee and colleagues put it in Applied and Environmental Microbiology, the designation rests on physical measurements such as the bubble point, meaning the pressure required to push air through the liquid held in the capillary network of a wetted membrane, combined with mathematical extrapolation. The number describes a hydraulic property, and biological retention is inferred from it.

The Rating Is a Physical Measurement, Not a Biological Guarantee

The published record on retention failure is older and larger than most buyers assume. Sundaram, Eisenhuth, Howard and Brandwein tested eight different 0.2 and 0.22 micrometre rated cartridge types from four manufacturers, all of them claimed to retain B. diminuta at 10⁷ CFU per square centimetre. The filters spanned nylon 6,6 and polyamide, hydrophilic and modified PVDF, asymmetric PES, and cellulose acetate. Challenged with bacteria drawn from a natural water source rather than a laboratory culture, all twenty five integral cartridges showed consistent penetration over an 18 to 24 hour period, at challenge levels of only about 10¹ to 10⁴ CFU per square centimetre. The authors concluded that penetration was qualitatively independent of membrane type and manufacturer, and that naturally occurring waterborne organisms were more penetrative than the validation organism itself.

The mechanism is not simply that some cells are smaller than the pores. Lee and colleagues at the FDA Center for Drug Evaluation and Research made this explicit: none of the organisms studied, including the highly penetrative Hydrogenophaga pseudoflava, are actually physically smaller than 0.2 micrometres. Geometry matters more than hydrodynamic diameter. H. pseudoflava is unusually thin, and it passed 0.2 micrometre rated membranes with log reduction values ranging from four to seven, while three benchmark organisms of similar hydrodynamic diameter but different shape did not pass under the same conditions.

Process conditions modulate this. In the same study, filter discs were challenged at loads from 1 x 10⁸ to 3 x 10¹⁰ CFU per square centimetre at a constant 30 pounds per square inch. For aluminium oxide, cellulose acetate and polyethersulfone membranes, passage was largely unaffected by extremes of viscosity, osmolality and conductivity. For nitrocellulose membranes, all three of those extremes increased retention, and for polysulfone membranes a 50 percent glycerol fluid increased retention, with the differences reaching statistical significance at p less than 0.05. Retention is therefore a property of the membrane, the fluid and the run conditions together, which is exactly why a validation performed on someone else’s fluid does not transfer cleanly.

Key Research Findings

• Sundaram et al., PDA Journal of Pharmaceutical Science and Technology, 2001, 55(2), 65-86: all 25 integral 0.2 and 0.22 micrometre rated cartridges tested, across 8 filter types and 4 manufacturers, showed consistent bacterial penetration over 18 to 24 hours at challenge levels of only 10¹ to 10⁴ CFU per square centimetre.
• Lee et al., Applied and Environmental Microbiology, 2010, 76(3), 695-700: Hydrogenophaga pseudoflava passed 0.2 micrometre rated filters with log reduction values of 4 to 7, at challenge loads of 1 x 10⁸ to 3 x 10¹⁰ CFU per square centimetre and a constant 30 pounds per square inch, with membrane chemistry effects significant at p less than 0.05.
• Kaushal et al., Journal of Industrial Microbiology and Biotechnology, 2013, 40(3), 327-334, authored at the FDA Center for Drug Evaluation and Research: both B. diminuta and Serratia marcescens grew through sterilizing grade membranes of differing polymer composition, stochastically rather than consistently, leading the authors to recommend minimizing processing time and characterizing pre-filter bioburden.
• ASTM F838 defines sterilizing grade as retention of a 10⁷ CFU per square centimetre B. diminuta challenge, equivalent to greater than 7 log or 99.99999 percent removal, measured on a filter type rather than on each production filter.
• Lee, Lee and Kim, PDA Journal of Pharmaceutical Science and Technology, 2002, 56, 99-108: B. diminuta cell size varied with growth agitation rate, and the smallest cells, produced in saline lactose broth shaken at 50 rpm, were not retained by a 0.2 micrometre rated cellulose acetate membrane.
• Zhao and Duong, Agilent Application Note 5991-1308EN: across five proteins at 0.5 mg per millilitre, a PES membrane returned recoveries above 97 percent, while comparator PVDF units produced average recoveries as low as 11.0 percent with relative standard deviations above 100 percent, and non-PES-equivalent PES units ranged from 43.6 to 85.0 percent.

Grow-Through and the Time Variable

Retention is usually discussed as an instantaneous property, as though a cell either passes or is captured. The work of Kaushal and colleagues complicates that. Working at the FDA, they demonstrated that both B. diminuta and Serratia marcescens can grow through sterilizing grade membranes of several polymer chemistries. Passage was not consistent, which is the important detail. It occurred stochastically at basal levels, meaning a filter that retains perfectly in one run may release organisms in another for reasons that no integrity measurement will reveal.

The practical consequence the authors drew is that contact time is a control parameter. A solution held against a membrane for many hours presents a fundamentally different risk than one filtered quickly, and the bioburden entering the filter matters as much as the filter rating. Neither the hold time nor the pre-filter bioburden count appears anywhere on a certificate of analysis. Both are recorded in the batch documentation of a facility that keeps such records, and are simply absent at a facility that does not.

Integrity Testing Is the Only Batch Specific Filtration Record

Because the bacterial challenge is destructive and is performed on representative filters, the link between a validated filter type and the specific filter that produced a batch is the integrity test. Regulatory expectation is that the sterilized filter is verified before use and confirmed again immediately after use, by bubble point, diffusive flow or pressure hold. The post-use test is the one that carries the information, because it detects damage that occurred during the run itself rather than damage present at installation.

A bubble point measurement works because a wetted membrane holds liquid in its pores by capillary action, and the pressure needed to displace that liquid scales inversely with pore diameter. A defect enlarges the effective pore, and the membrane bubbles at a lower pressure than its specification. The limitation is that this is a bulk hydraulic measurement across the entire filtration area. A single oversized channel in a large cartridge contributes a small fraction of total flow, so the sensitivity of the test to a localized defect is finite and depends on filter area, wetting fluid and the specification limit chosen.

This is where the parallel with other physical container tests becomes useful. The same structural problem appears in container closure integrity testing for research peptide vials, where a pass result establishes that no defect above a detection threshold was found, not that the package is defect free. Integrity testing of a filter is evidence of the same kind: informative, necessary, and bounded.

The Membrane Removes Peptide As Well As Bacteria

There is a second reason the filtration step belongs in any serious discussion of peptide quality, and it has nothing to do with microbiology. Membranes bind peptide. In the Agilent evaluation by Zhao and Duong, five proteins spanning 12 to 660 kilodaltons and isoelectric points from 4.5 to 9.6 were filtered through 0.2 micrometre syringe filters and compared against unfiltered controls by size exclusion chromatography. A well made PES unit returned recoveries above 97 percent across all five. Several PVDF comparators did not: average recoveries fell as low as 11.0 percent, with relative standard deviations exceeding 100 percent, and the losses were worst at the lowest concentrations tested. Even among PES units from different suppliers, average recoveries ranged from 43.6 to 85.0 percent.

Two implications follow. At manufacturing scale, membrane binding is saturable and is normally managed by conditioning the filter, so the effect on fill concentration is bounded but not zero. At the bench, where a researcher may filter a small volume of a dilute solution through a fresh membrane, the loss is neither bounded nor small, and it compounds the surface losses described in our analysis of peptide adsorption to labware. A concentration calculated from a labelled vial content and a solvent volume can be substantially higher than the concentration actually delivered to an assay, and the certificate of analysis will not indicate this because the certificate describes the powder, not the handling.

Why None of This Appears on a Certificate of Analysis

A peptide certificate of analysis is an analytical document. It reports identity, chromatographic purity, and in better cases water content, counterion content and microbial attributes. Filtration is a process control, and process controls live in batch records. The filter type, the integrity test values recorded prior to the run and again on its completion, the pre-filter bioburden count, the hold time and the fluid composition are all generated during manufacture and none of them are analytical results on the finished powder.

That distinction is not a criticism of the certificate. It is a boundary. The same boundary explains why cleanroom grade and aseptic process simulation data do not appear on a certificate either, and why a passing sterility result on a finished lot carries a known statistical ceiling rather than a guarantee, as covered in our review of sterility, bioburden and microbial limits. Filtration, environment and finished product testing are three separate layers of evidence. A buyer who reads only the third layer is inferring the first two.

What a Researcher Can Reasonably Ask

The useful questions are narrow. Was the product filtered through a filter qualified as sterilizing grade under ASTM F838, and by whom was that qualification performed. Was a post-use integrity test run on the filter that produced this specific lot, and did it pass its specification. What membrane polymer was used, since polymer identity predicts both retention behaviour under unusual fluids and peptide binding. How long was the solution in contact with the membrane, given that grow-through scales with time.

A supplier who can answer none of these is not necessarily selling poor material. It may simply mean the filling operation does not generate the records, which is itself informative. A supplier who answers them is describing a controlled process, and that claim is separable from, and additional to, the purity number on the certificate. Maple Research Labs publishes its third party analytical documentation on the certificates of analysis page, and treats the analytical certificate as one layer of evidence rather than the whole of it.

Limitations and Open Questions

Most of the retention failure literature was generated with cartridge filters at manufacturing scale, using water borne or nutritionally stressed organisms, and the transferability of those log reduction values to a small disc filter running a peptide solution in acetic acid or bacteriostatic water is not established. The Agilent recovery data were generated on proteins from 12 to 660 kilodaltons, and short synthetic peptides below 5 kilodaltons may bind differently, since binding is driven by hydrophobic and electrostatic interaction with the polymer rather than by size alone. Published quantitative recovery data specific to research peptides on named membrane chemistries remain sparse, which is a genuine gap rather than a settled matter.

The direction of the evidence is nonetheless consistent. A 0.2 micrometre rating is a hydraulic descriptor from which biological retention is inferred, that inference holds under validated conditions and degrades outside them, and the batch specific evidence that it held for any particular lot is an integrity test result that no certificate of analysis reports.

For research purposes only. Not for human consumption. Not for diagnostic or therapeutic use.