What a COA Does Not Test For: Endotoxin, Water Content and Residual Solvents on a Two-Test Certificate

Most certificates issued with research peptide material report two things: chromatographic purity by reversed-phase HPLC, and identity by mass spectrometry. Everything else measurable on a lyophilised powder is absent — bacterial endotoxin, microbial count, water content, residual solvents from synthesis and purification, counterion content, net peptide content, elemental impurities, sterility. None of those absences is announced anywhere on the page, because a certificate records what was tested and says nothing at all about what was not. That is why the same document can be entirely truthful and still leave most of a material's quality attributes unmeasured.

This is a question of scope, not of honesty. ISO/IEC 17025:2017, the standard for the competence of testing laboratories, requires a report to identify the method used for each result and to state explicitly that the results relate only to the items tested ^[1]. A laboratory that ran two assays reports two assays, correctly. The gap opens on the reader's side: a page headed 'Certificate of Analysis' invites the assumption that the analysis was comprehensive. ICH Q6A draws the same boundary from the specification side — a specification is a defined list of tests, analytical procedures and acceptance criteria, and material conforms to it only in respect of the tests it contains ^[10]. Conformance is never a general statement about quality.

What a two-test certificate actually establishes

The purity line is normally area-percent by reversed-phase HPLC with ultraviolet detection near 214 nm, where the peptide bond absorbs. It answers one question: of everything that eluted and absorbed at that wavelength, what fraction of the integrated peak area belonged to the main peak. The mass-spectrometry line answers a second, independent question: does the observed mass of that main species match the mass calculated from the stated sequence. Both are useful, and they are orthogonal, which is why they are the pair most often run.

Neither measures the contents of the vial. Area-percent is a ratio among detected species, not a mass fraction of the powder — anything that does not absorb at the detection wavelength, does not elute, or co-elutes with the main peak is invisible to it, a set of exclusions worked through in how a purity percentage is calculated and what it leaves out. Mass spectrometry confirms a species of the right mass is present; it does not quantify how much of the weighed solid that species accounts for. A lot can read 99% pure by HPLC and confirmed by MS while a substantial share of what is in the container is water, salt and solvent.

Which tests are missing, and which method would have run them

Each row below names an attribute a two-test certificate leaves blank, the compendial or reference method that measures it, and the specific question that goes unanswered without it.

Endotoxin: the contaminant a purity figure cannot see

Bacterial endotoxin is a lipopolysaccharide fragment of the Gram-negative cell wall. It is not a peptide-related impurity, does not appear in a chromatogram of the peptide, and is not bounded by any purity percentage, which is why endotoxin and bioburden have to be established by their own testing rather than inferred from a chromatogram. The compendial test is USP <85>, which sets out three amoebocyte-lysate procedures: gel-clot, giving pass or fail at a defined sensitivity, and the turbidimetric and chromogenic methods, which are quantitative and report in endotoxin units ^[2]. Since 2024 a parallel chapter, USP <86>, covers the same determination using recombinant reagents rather than lysate from horseshoe crab haemolymph, published for early adoption with an official date of 1 May 2025 ^[3]. A certificate saying only 'endotoxin: pass' has not named a method, and so has not told you the sensitivity the result was read at.

The familiar numeric limit — 5 endotoxin units per kilogram of body mass per hour, scaled by the maximum administered amount to give a concentration limit — is written for licensed injectable pharmaceuticals and enforced as a release criterion for those products ^[4]. It is not a specification for research material and does not become one by being quoted. The laboratory argument for measuring endotoxin is separate and empirical: residual endotoxin within the levels many protein suppliers routinely guarantee — on the order of one endotoxin unit per microgram of protein, roughly 100 picograms of lipopolysaccharide — was sufficient to activate primary human CD1c+ dendritic cells in culture, producing activation that would otherwise be attributed to the protein under study ^[8]. That result is in vitro, in human cells, and concerns recombinant proteins rather than synthetic peptides; the mechanism, though, belongs to the contaminant, not to the molecule it travels with.

Two physical facts follow. Sterilising filtration through a 0.22 µm membrane removes organisms but not endotoxin, which passes freely. Autoclaving kills organisms but does not destroy endotoxin either — depyrogenation of glassware conventionally requires dry heat, typically 250 °C for at least 30 minutes. A solution can therefore be sterile and pyrogenic at once, which is why endotoxin sits in its own chapter, separate from the sterility test.

Water content: the number that moves every mass calculation

Karl Fischer titration is the compendial determination of water, USP <921> Method I. It rests on the quantitative reaction of water with sulfur dioxide and iodine in the presence of a buffering base, run as either a direct or a residual titration with the specimen normally dissolved in anhydrous methanol ^[6]. Its distinguishing property is specificity: it measures water and only water. The commonly substituted alternative, loss on drying, is gravimetric and measures everything volatile that leaves the sample — water plus residual solvent plus anything else — so it systematically over-reports water in material that still holds solvent from purification.

This matters because lyophilised peptide solids are hygroscopic, and the water they hold is counted in every weighing. Without a water figure the label mass is nominal, and the gap between stated and actual solid content is not derivable from any other line on the certificate. The same property governs handling: a container opened before it has equilibrated from cold storage to ambient temperature condenses atmospheric moisture onto the solid, so a manufacturer's water result describes the material as packed, not as it stands after repeated cold-to-warm cycles.

Residual solvents: what ICH Q3C covers, and what it leaves out

Solid-phase synthesis and reversed-phase purification leave volatile organics behind. The reference framework is ICH Q3C, maintained by the International Council for Harmonisation and revised most recently as Q3C(R9), which sorts solvents by toxicological class and assigns each a permitted daily exposure from which an Option 1 concentration limit is derived ^[5]. Class 1 solvents are to be avoided and carry limits in the low parts per million. Class 2 solvents are limited, and several are exactly the reagents a peptide chemist reaches for. Class 3 solvents are treated as of low toxic potential and share a default limit of 5000 ppm ^[5].

• Acetonitrile, Class 2 — permitted daily exposure 4.1 milligrams, Option 1 concentration limit 410 ppm ^[5].
• Dichloromethane, Class 2 — Option 1 concentration limit 600 ppm ^[5].
• Methanol, Class 2 — Option 1 concentration limit 3000 ppm ^[5].
• N,N-dimethylformamide, Class 2 — Option 1 concentration limit 880 ppm ^[5].
• Benzene, Class 1 — Option 1 concentration limit 2 ppm, and to be avoided rather than controlled ^[5].

Those are the limits a headspace GC result is read against, and they are why a bare 'complies with ICH Q3C' line is weaker than it looks: compliance is a statement about a specific list tested by a specific method, and it does not extend to reagents outside that list. Trifluoroacetic acid is the instructive case. It appears in Q3C not in Class 1, 2 or 3 but in the table of solvents for which no adequate toxicological data was found — a category with no permitted daily exposure and therefore no Option 1 concentration limit at all, the manufacturer being asked to justify residual levels instead ^[5]. A certificate can be fully Q3C-compliant and say nothing about the trifluoroacetate the material carries.

The counterion is mass that is not peptide

Peptides purified by reversed-phase chromatography with trifluoroacetic acid in the mobile phase are isolated as trifluoroacetate salts, with counterion associating at basic residues. That salt is real mass in the vial and it is not peptide, which is why net peptide content parts company with gross weight: the figure is determined by amino acid analysis or nitrogen determination, not by chromatographic purity, and it sits routinely well below the purity percentage printed on the certificate. Neither figure is on a two-test document.

The counterion is also not inert in cell culture. Trifluoroacetate at 10⁻⁸ to 10⁻⁷ M reduced cell number and thymidine incorporation in fetal rat osteoblast cultures, and when trifluoroacetate and hydrochloride salts of the same peptides were compared in that system, proliferation differed by salt form — enough, the authors noted, for a proliferative effect to be missed or an antiproliferative one wrongly attributed to the peptide ^[9]. That is a 1999 in vitro study in rodent primary cells and is not a general toxicity claim; the counterion literature is not consistent across peptides, and for some sequences the acetate salt has proved the more cytotoxic. The defensible generalisation is narrower: counterion identity is an experimental variable that differs between suppliers and between lots, and a two-test certificate does not tell you which salt you have or how much of it.

Bioburden is not sterility, and neither is on the page

Two compendial chapters cover the microbiological quality of non-sterile material, and they answer different questions. USP <61> is quantitative: total aerobic microbial count and total combined yeast and mould count by plate count, with a suitability step demonstrating that the material does not itself inhibit recovery of the organisms being counted. USP <62> is qualitative and looks for named species — among them Staphylococcus aureus, Pseudomonas aeruginosa, Escherichia coli, Salmonella species, Candida albicans, and bile-tolerant Gram-negative bacteria ^[7]. Neither is a sterility test; sterility is USP <71>, with its own sampling and incubation requirements. A powder manufactured under ordinary chemical-plant conditions is non-sterile, and no purity result changes that.

How to ask a supplier for what is missing

Requests that name a chapter get answered; requests that ask whether something is 'clean' do not. Work down this list and record the reply verbatim against the lot.

1. Ask which tests were performed on this lot, by name and method, rather than asking for 'full testing'.
2. For endotoxin, ask which chapter and procedure — USP <85> gel-clot, turbidimetric or chromogenic, or USP <86> with recombinant reagents — and at what sensitivity ^{[2] [3]}.
3. For water, ask for Karl Fischer by USP <921> Method I rather than loss on drying, and note which was actually run ^[6].
4. For residual solvents, ask for the headspace GC result solvent by solvent against Q3C limits, and ask separately for trifluoroacetate, which Q3C does not limit ^[5].
5. Ask whether the purity figure is area-percent, at what detection wavelength, and whether net peptide content was determined at all.
6. Treat 'sterile' as a claim requiring USP <71>, and ask whether <61> and <62> were run at all ^[7].
7. File every 'not tested' answer with the certificate, against the lot.

The honest limit on all of this: no supplier in this category runs the full panel, and expecting one to is not a realistic standard. Endotoxin, bioburden and elemental impurity determinations carry real per-lot cost, and material sold for laboratory use is not manufactured to the release requirements of an injectable pharmaceutical. The useful posture is not to demand a certificate that will not exist, but to know exactly where the document's boundary falls — and to stop reading a purity figure as a summary of quality when it is one ratio, at one wavelength, on one sample, on one day. A certificate is evidence in proportion to the tests it names, and a document that draws its own boundary plainly is worth more than one that implies it has none.