Cell Culture Practice for Peptide Experiments: Serum, Passage, Mycoplasma and Identity

A cell-based number is a statement about a system, not about a molecule. The system includes the medium, the serum lot, the plastic, the passage history of the cells, and whatever organisms are sharing the flask. Each can move a measured value by more than the effect being studied, and none appear in the figure. This guide is about the parts of that system that most often make a peptide result unrepeatable, and about what has to be recorded for a reader to judge whether a figure means anything.

The Good Cell and Tissue Culture Practice 2.0 guidance document sets out six principles for in vitro work — characterization and maintenance of essential characteristics, quality management, documentation and reporting, safety, education and training, and ethics — and states explicitly that it does not intend to impose detailed procedures but to describe potential quality issues ^[1]. That is the right posture here. There is no single correct protocol; there are failure modes, and the question of whether an experiment looked for them.

The medium is a reagent, and it is not fully defined

Serum is the least characterized thing in most cell culture experiments. A workshop report on replacing fetal bovine serum describes animal serum as a supplement with an undefined and variable composition, and notes that commercially available defined supplements often come with limited formulation information and therefore cannot be regarded as completely defined ^[2]. A protocol line reading "10% FBS" specifies a volume fraction and nothing else — not the growth factor content, not the protease content, not the binding protein content, not the lot.

Medium components are also chemically reactive, and they can consume a test compound outright. In a study of the phospholipase C inhibitor U73122 — a small molecule with an electrophilic maleimide group, not a peptide — the compound's half-life on incubation was reported as approximately 150 minutes in phosphate-buffered saline, 60 minutes in Hanks' buffered saline solution with 2 mM glutamine, 32 minutes in MCDB131 basal medium without BSA, 30 minutes in complete medium, and 18 minutes in DMEM with 2 mM L-glutamine; glutamine conjugates were identified in cell culture medium, and the compound was not recoverable at all from medium supplemented with 0.5% BSA ^[3]. The authors further found that inhibition of EGF-stimulated calcium release in the A431 human epidermoid carcinoma line was substantially reduced by the presence of BSA in a time-dependent manner, and concluded that chemical reactivity may limit availability in complex cellular assays and lead to misinterpretation of results ^[3].

That result concerns one maleimide-bearing small molecule and does not transfer to peptides as a class. The design point does: the buffer in which a compound is stable is frequently not the medium in which it is assayed, and the difference was visible only because recovery was measured in each formulation rather than assumed.

Serum binding, adsorption, and the gap between nominal and free

Two distinct processes remove a peptide from solution before it reaches a receptor: binding to serum proteins, and adsorption to labware. Both reduce recovery, and separating them is genuinely hard. A study of oritavancin — a lipoglycopeptide antibiotic — states that biophysical methods for studying its serum protein binding are confounded by nonspecific drug adsorption to labware surfaces, which is why the authors instead used a microbiological growth-based method under conditions allowing near-quantitative drug recovery; binding measured that way ranged from 81.9% in human serum to 87.1% in dog serum ^[4]. The methodological admission is the more generally useful part.

Work quantifying the interaction of lipids with polymer surfaces used radiolabeled linoleic acid and cholesterol, and also measured loss across a sterilizing membrane filter; the authors found potential for significant loss of hydrophobic components through non-specific binding at timescales relevant to a typical cell culture process, with the extent depending on both the nature of the component and the type of surface, and noted that the level of loss is very dependent on the specific system ^[5]. The species studied were lipids, not peptides. What a peptide workflow shares is the mechanism: component-specific and surface-specific partitioning onto plastic and onto filters.

Low-binding polypropylene, carrier protein in dilution buffers, and pre-rinsed filters are standard bench responses. We are not aware of a primary study establishing a general recovery benefit of any of these for research peptides as a class; treat them as conventions with a plausible mechanism rather than measured effects. The measurable version is a recovery check — quantify the peptide in a cell-free well, same plate, same incubation, and see what fraction survives.

A worked free-fraction calculation

Suppose, hypothetically, a peptide is 90% bound under the serum conditions of an assay. The free fraction is 1 − 0.90 = 0.10, so a nominal 100 nM addition leaves 100 nM × 0.10 = 10 nM free. If only the free species is active, and the same peptide shows an apparent half-maximal concentration of 12 nM under serum-free conditions where nominal and free coincide, the serum-containing condition needs roughly 12 nM ÷ 0.10 = 120 nM nominal for the same free concentration — a ten-fold rightward shift produced entirely by binding, carrying no information about intrinsic activity. The numbers are invented to make the arithmetic legible; the durable point is that the shift factor is the reciprocal of the free fraction, so a binding estimate has to exist before potencies measured under different serum conditions can be compared.

Depletion: when the cells consume the peptide

At low concentrations the receptors themselves can remove a meaningful share of the added peptide, flattening the low end of a concentration-response curve and biasing the fitted midpoint. Take a well containing 100 µL of medium, which is 0.0001 L, and 100,000 cells.

• At a nominal 10 nM, the well holds 10 nmol/L × 0.0001 L = 0.001 nmol, which is 1 pmol, or 1,000 fmol.
• At 10,000 receptors per cell, the well contains 100,000 × 10,000 = 1 billion binding sites. Divided by Avogadro's number, 1 billion ÷ (6.022 × 10 to the 23) = 1.66 femtomoles.
• Receptor occupancy could therefore sequester 1.66 fmol ÷ 1,000 fmol = 0.17% of the added peptide. Negligible.
• Drop the nominal concentration to 10 pM and the well holds 1 fmol — less than the 1.66 fmol of available sites. Here the receptor population alone can substantially deplete the free peptide.

The denominator in the third step is the total peptide added, not the receptor number; reversing it inverts the conclusion. The common rule of thumb that bound ligand should stay under about 5% of added ligand is a convention rather than a measured threshold, and the receptor density above is an assumption. Both should be labeled as such wherever the calculation appears.

Degradation over the incubation, and what a nominal concentration averages to

A peptide that degrades during the assay is not at its nominal concentration for most of the incubation. If degradation is approximately first-order with half-life t½, the rate constant is k = ln2 ÷ t½. For a hypothetical 6-hour half-life, k = 0.693 ÷ 6 = 0.1155 per hour. Over a 24-hour incubation, kt = 0.1155 × 24 = 2.772, and the surviving fraction is e to the −2.772 = 0.0625, or 6.25%.

The time-averaged concentration over that window is C0 × (1 − e to the −kt) ÷ (kt) = C0 × 0.9375 ÷ 2.772 = 0.338 × C0. The denominator is kt, the dimensionless product, not the elapsed time in hours; using 24 there instead of 2.772 understates the average by about nine-fold. So a well labeled 100 nM delivered a mean exposure near 34 nM and finished at about 6 nM. Whether that matters depends on the readout, but only a stability measurement in the actual assay medium tells you which regime you are in.

Passage number and the drift of what you are measuring

Cell lines change with time in culture, and the change is not slow. A genomic analysis of 106 human cell lines grown in two laboratories showed extensive clonal diversity, and comprehensive characterization of 27 strains of the breast cancer line MCF7 uncovered rapid genetic diversification, with similar results in multiple strains of 13 additional lines ^[6]. The functional consequence was measured directly: when the 27 MCF7 strains were tested against 321 anti-cancer compounds, at least 75% of the compounds that strongly inhibited some strains were completely inactive in others ^[6]. Barcoding experiments in the same study indicated that this evolution results from positive clonal selection that is highly sensitive to culture conditions ^[6]. The compounds tested were anti-cancer agents, not peptides, and the authors' claim concerns within-line variation rather than any particular target.

Functional properties can drift in different directions in different lines. A study of permeability protocols in Caco-2, MDCK and MDCK-MDR1 monolayers reported that the paracellular route became more restricted with higher passage numbers in both protocols tested, while P-glycoprotein functionality assessed with Rhodamine-123 permeability did not change with passage number in Caco-2 cells under either protocol but increased with passage under both protocols in MDCK and MDCK-MDR1 cells; for compounds absorbed by passive diffusion the authors saw no clear trend in either permeability values or their variability ^[7]. "High passage" is therefore not a single direction of effect. It is a reason to fix a passage window and show that the property you rely on is stable within it.

Passage count is also an ambiguous unit, because a passage is not a fixed amount of growth. A 1:8 split corresponds to log base 2 of 8 = 3 population doublings, so 20 passages at that ratio is roughly 60 doublings, whereas 20 passages at 1:2 is about 20. Reporting population doublings alongside passage number removes that ambiguity; this is a documentation convention, not a finding.

Mycoplasma

Mycoplasma contamination is invisible under a microscope, does not cloud the medium, and changes the biology. An unbiased survey of sequence data from 9,395 rodent and primate samples across 884 series in NCBI's Sequence Read Archive found 11% of the series contaminated, using a threshold of at least 100 reads per million mapping to mycoplasma in one or more samples; the authors examined a single-cell RNA-seq dataset and found 61 host genes significantly associated with mycoplasma-mapped read counts at P < 0.001, and concluded that the study suggests contamination is still prevalent and poses substantial risk to research quality ^[8]. Note the operational definition: 11% is the fraction of series exceeding a stated read threshold, not a direct measurement of infection rate.

The effect on a pharmacological readout is not hypothetical. In human vocal fold fibroblasts, basal mRNA expression differed significantly with mycoplasma, and differential responses to dexamethasone were observed across multiple pathways as a function of mycoplasma presence, with apoptosis, DNA damage repair and G1-to-S cell cycle signaling among the pathways significantly enriched in mycoplasma-positive cells ^[9]. The authors state that it remains unclear whether their findings represent contamination or the baseline state of that specialized tissue ^[9]. A stimulus-response experiment can therefore return a different answer depending on a variable nobody recorded.

Contamination can also be silent in one assay format and visible in another. Comparing 2D and 3D cultures of pig kidney epithelial LLC-PK1 cells, one study observed no effect of mycoplasma contamination on proliferation in 2D monolayer culture, but found cyst formation in 3D tissues altered in number, size and structure ^[10]. A clean growth curve is not evidence of a clean culture.

Testing on receipt, at intervals during use, and before banking, with a documented method and a recorded result, is standard practice. We cite no primary source for that cadence because it is convention rather than an evidence-derived optimum.

Cell line authentication

A peptide result attributed to the wrong cell line is wrong in a way no amount of replication will fix. Of the genomic technologies applied to cell line authentication, only short tandem repeat profiling has been the subject of a comprehensive and definitive standard, ASN-0002, and the same authors discuss how a combination of methods could be used in a holistic fashion to assess various critical aspects of cell line quality ^[11]. The practical obstacle has historically been analysis rather than generation of profiles: the Cellosaurus knowledge resource was built to compile publicly available STR profiles, and at its release of 31 July 2019 held 6,474 distinct cell lines with an associated STR profile, queryable through the CLASTR similarity search tool ^[12].

Sequencing-based profiling is one alternative to the standard capillary electrophoresis workflow. An evaluation of next-generation sequencing STR profiling at 18 loci for human and 15 loci for mouse cell lines, analyzed with the STRight program, reported it as superior to standard STR-CE in reporting the sequence context of repeat motifs, in sensitivity, and in flexible multiplexing capability, and describes it as a valuable alternative ^[13]. That is the authors' assessment of their own method, and "alternative" rather than "replacement" is their word; ASN-0002 remains the standard ^[11].

What to record

• Cell line name, source and catalog identifier, date and result of the most recent STR authentication, and passage number plus cumulative population doublings at the time of the experiment.
• Mycoplasma test method, date and result.
• Medium formulation: base medium, glutamine source and concentration, serum type, volume percent, and lot number.
• Whether the assay ran in serum-containing, serum-reduced or serum-free medium, and for how long.
• Peptide stock solvent, storage, freeze-thaw count, plate and tube material, and any measured recovery or stability in the assay medium.
• Seeding density, well volume, and confluence at the time of treatment.

None of this is burdensome, and its absence is what makes a discrepancy between two laboratories unresolvable.

What is not known

Peptide-specific quantitative data on binding and adsorption losses in routine culture formats is thin. The binding figures cited above are for a lipoglycopeptide antibiotic measured in serum rather than in complete medium ^[4], and the surface-loss data are for lipids ^[5]; neither licenses a numerical expectation for any particular research peptide. Whether the passage-dependent changes observed in permeability models ^[7] extend to receptor expression in other lines is not established by those data. The mycoplasma prevalence figure describes sequencing series deposited in a public archive, which is not a random sample of laboratories ^[8]. And the relationship between a nominal concentration and a free, intact concentration at the cell surface goes unmeasured in most published peptide cell work — not because it is unimportant, but because measuring it requires an analytical method in the assay matrix that most cell-biology workflows never build.