Designing In-Vitro Concentration–Response Experiments for Peptides

A concentration–response experiment estimates three things: potency (EC50 or IC50), the size of the response between its floor and ceiling, and the steepness of the transition. Each estimate is only as good as the concentrations, controls and replication behind it, and most of those decisions are made before the first well is filled. This guide covers in-vitro designs only, in cells, tissues and receptor or enzyme assays; it does not deal with dosing animals or people.

What the curve estimates

The terms used to describe these experiments, including EC50, IC50, pEC50 and Emax, are defined in the International Union of Pharmacology's recommendations on quantitative pharmacology ^[1]. Most sigmoidal concentration–response data are fitted with the four-parameter logistic model, which is the same function as the Hill equation ^[2]:

Response = Bottom + (Top − Bottom) / (1 + 10^((logEC50 − log[A]) × nH))

There are two definitions of EC50 in use. The relative EC50 is the fitted midpoint between the two plateaus of the curve itself; the absolute EC50 is the concentration giving a response halfway between the 0% and 100% assay controls, and is only appropriate when a stable 100% control exists and the 50% level can be estimated to within about 5% ^[3].

Choosing the concentration range

The range must reach both plateaus. Guidelines for reportable estimates require at least two concentrations beyond the lower and upper bend points for a relative EC50, and at least two predicted responses on each side of 50% for an absolute EC50 ^[3]. A curve that never flattens at the top yields an EC50 that is an extrapolation rather than a measurement.

A literature value is a starting point, not a centre to cluster around. Ipamorelin, for example, released growth hormone from primary rat pituitary cells with an EC50 of 1.3 nM ^[4]; a first experiment in a different system would still span at least three log units either side of that figure, because cell type, receptor density and readout all move potency. Run a wide pilot, then refine.

Two practical limits bound the range:

• The top concentration. High peptide concentrations bring high counter-ion and vehicle loads. Trifluoroacetate, a common counter-ion in HPLC-purified peptides, reduced osteoblast and chondrocyte proliferation at 10⁻⁸ to 10⁻⁷ M, and TFA salts of amylin and calcitonin produced misleading anti-proliferative results; the authors flagged the issue for any work above 10⁻⁹ M peptide ^[5]. Solubility caps the top end too; see peptide solubility troubleshooting.
• The bottom concentration. Peptides stick to surfaces. Recovery of eight radiolabelled endocrine peptides from glass and plastic tubes varied widely with surface, siliconising reduced it, and 1% BSA improved it ^[6]. Serial dilution through pipette tips multiplies that exposure, which is one reason non-contact acoustic dispensing has been used to prepare peptide dilution series for potency assays ^[7].

Log spacing and the number of points

Concentrations are spaced evenly on a log scale because the response is sigmoidal in log concentration. For a Hill slope of 1, the response rises from 10% to 90% over an 81-fold range (1.9 log units); the window narrows to 81^(1/nH), so a slope of 2 compresses it to ninefold. The spacing has to put several points inside that window.

Half-log or threefold spacing is the usual compromise for a first curve. Twofold spacing over a narrower range suits refinement once the EC50 is roughly known, or steep curves.

Worked example: a half-log series

A 5 mg vial of ipamorelin (711.9 g/mol) reconstituted in 2.0 mL gives 2.5 mg/mL, or 3.51 mM if the fill is taken as pure peptide; correct it with the batch's net peptide content for exact work. To run a curve with a 10 µM top concentration, where peptide is added as one-tenth of the final well volume, the top working solution is 100 µM: 28.5 µL of stock made up to 1.0 mL in vehicle-matched medium. Each half-log step then transfers 100 µL into 216 µL of the same medium.

Because every dilution is made in the same medium, each well carries the same vehicle load. The molarity calculator and dilution calculator reproduce these numbers, and peptide molarity calculations explains the conversions from mg/mL.

Controls

The vehicle deserves the most thought. DMSO at 0.3125% caused minimal cytotoxicity across six cancer cell lines, but the safe limit depended on cell type and exposure time, and ethanol at the same concentration reduced viability by more than 30% after 24 hours ^[9]. Bacteriostatic water contains 0.9% benzyl alcohol, or 9 mg/mL. In human retinal pigment epithelial cells, benzyl alcohol at 9 mg/mL was toxic within five minutes and 0.225 mg/mL impaired function at two hours, while 0.0225 mg/mL did not ^[10]. A stock made in bacteriostatic water and diluted 1:100 into medium carries 0.09 mg/mL, between those two levels, so for cell work sterile water or buffer is the cleaner vehicle and any benzyl alcohol must appear in the vehicle control.

Assay quality can be summarised with the Z-factor, which combines the separation between positive and negative controls with their variability ^[11]: Z = 1 − 3(σp + σn) / |μp − μn|. Values closer to 1 mean a wider window relative to noise; tracking it plate by plate shows when an assay is drifting.

Replication: what counts as n

The unit of replication is whatever receives the treatment independently. A survey of published animal experiments in which parents were treated and offspring measured found pseudoreplication, where non-independent measurements were counted as separate samples, in 46% of studies, and only 22% replicated the correct unit; the authors apply the same reasoning to in-vitro designs, where wells treated from one dilution series on one day are not independent experiments ^[12].

A common design is triplicate wells within each run and at least three independent runs, with the curve fitted separately for each run and the pEC50 values then summarised.

Plate layout and edge effects

Wells at the edge of a microplate behave differently from those in the middle. In one study of 96-well plates, outer wells showed 35% lower metabolic activity than central wells in one brand and 16% lower in another, with smaller deficits extending to the second and third rows; loose wrapping during incubation and a buffer between wells reduced the effect ^[13]. An earlier screening study traced part of the problem to uneven cell settling when freshly seeded plates went straight into a CO₂ incubator, and found that a short rest at room temperature before incubation reduced it ^[14].

Layout choices that follow:

• Fill the outer ring with medium or buffer rather than samples where the plate format allows.
• Do not run a dilution series along one edge with controls at the other; interleave or randomise positions across runs.
• Put vehicle and reference controls on every plate, in both edge-adjacent and interior positions.

Peptide-specific pitfalls

• Proteolysis in the assay. Human plasma degraded GLP-1 to a truncated metabolite with a half-life of about 20 minutes at 37 °C, a reaction blocked by a DPP-IV inhibitor or by cooling to 4 °C ^[15]. Stability in plasma and cell-culture supernatants varied widely between model peptides, and precipitating samples with strong acid lost analyte compared with organic solvents ^[8]. In long incubations, measured potency reflects whatever intact peptide remains.
• Adsorption. Low concentrations lose a larger fraction to tubes, plates and tips ^{[6, 7]}. Carrier protein in the diluent, low-binding plastics and fewer transfer steps all help.
• Counter-ions. At high concentrations the salt can contribute its own effect ^[5]; TFA vs acetate peptide salts covers the choice.
• Stale dilutions. A working dilution kept for a week is a different reagent. Prepare dilutions fresh for each run from frozen single-use aliquots.

Fitting and reporting

Fit the four-parameter model by nonlinear regression on log concentration. Free tools give parameter estimates essentially identical to commercial software and to R's nls function ^[2], so the choice of package matters less than the choices made within it. Constrain the top or bottom to control values only when those controls are reliable, and say so.

IC50 values for antagonists need particular care. An IC50 depends on the agonist concentration it competes against; the Cheng–Prusoff relationship converts it to an equilibrium dissociation constant, and a modified power equation is needed when the agonist curve's slope differs from 1 ^[16].

A complete report states:

• pEC50 or pIC50 with its spread across independent runs, the Hill slope and both plateaus;
• the number of independent runs and wells per concentration;
• whether the EC50 is relative or absolute ^[3], and any constrained parameters;
• the vehicle and its final concentration;
• the peptide's molecular weight, salt form and whether concentrations were corrected for net peptide content;
• incubation time, temperature and medium, including serum.

For the bench

• Convert every stock to molar units before planning a curve; the molarity calculator and reconstitution calculator do the arithmetic.
• Prepare stocks under sterile technique and follow the reconstitution guide for the first dissolution.
• Confirm identity and purity before a large experiment; what HPLC testing measures explains what a purity figure covers.
• For long-acting analogues such as semaglutide, read peptide half-life research before choosing incubation times.
• Combination designs need their own planning; peptide blends explained covers fixed-ratio material and when separate vials are required.