Cell viability and cytotoxicity assays — MTT, CellTiter-Glo and beyond

Cell viability is measured in almost every drug-discovery and toxicology lab, yet the IC50 values for the same compound on the same cell line reported across publications routinely span an order of magnitude. The reagent matters less than most researchers think; the protocol, normalization, and curve-fitting decisions matter far more. This piece covers the assay landscape and the analytical pipeline that makes viability IC50s reproducible and comparable.

Assay selection — what each reagent actually measures

The most consequential decision is choosing the right endpoint for the question being asked. These are not interchangeable:

• MTT / MTS / XTT / WST-1 (tetrazolium reduction): mitochondrial dehydrogenase activity — a metabolic proxy for live cell number. Reduced by metabolically active cells; does not distinguish cytostatic from cytotoxic effects at early time points. Interference: colored compounds, ascorbic acid, phenazine methosulfate, direct reduction by thiols.
• CellTiter-Glo / ATP bioluminescence: intracellular ATP — tightly correlated with viable cell number across proliferating cells. Linear range typically 100–100,000 cells/well; evaporation from edge wells inflates signal in 384-well format. Most sensitive of the common viability assays.
• CyQuant / Hoechst 33258 / DAPI nuclear count: total DNA content — measures total cell number, alive and dead. Useful for proliferation assays where you need to separate growth inhibition from death. Dead cells in suspension are washed away before staining — this is both a feature (counting attached live cells) and a confounder (suspension toxicity is missed).
• SYTOX Green / PI (propidium iodide): membrane impermeant dyes that stain dead cells with disrupted membranes. Cannot distinguish early apoptosis from necrosis; combined with a mitochondrial membrane potential dye gives more mechanistic information.
• Annexin V / PI dual staining by flow cytometry: mechanistic — distinguishes live (Ann−/PI−), early apoptosis (Ann+/PI−), late apoptosis/necrosis (Ann+/PI+). The only assay that quantifies the mechanism of death. Required when the compound or hypothesis concerns apoptosis specifically.

Cell seeding and growth curve calibration

The most common source of IC50 variability is inconsistent seeding density. Cell number at the time of treatment, not at seeding, determines the IC50 — and cells that are sub-confluent respond differently to cytotoxic stress than over-confluent cells. Required before any compound screen:

1. Generate a growth curve (seed day 0, read viability signal every 24 h for 5 days) at three seeding densities.
2. Identify the density where cells are in exponential growth at the intended treatment endpoint — this is the correct seeding density.
3. Validate the linear range of the viability assay: signal must be proportional to cell number with R² ≥ 0.99 across the cell number range that spans the assay window.

Normalization strategies

The choice of normalization determines what the IC50 means biologically:

• T=0 normalization (GI50): normalize to the signal at the time of compound addition, not to the untreated control at endpoint. GI50 measures the concentration that produces 50% growth inhibition — cytostatic and cytotoxic compounds both reach GI50 but for different reasons. The NCI uses GI50 for their drug sensitivity data.
• Vehicle control normalization (CC50 / IC50): normalize to endpoint vehicle control. IC50 here conflates growth inhibition with cell death. For a proliferating cell line treated for 72 h, a compound that completely halts proliferation without killing any cells will read as ~50% viability if the vehicle control has doubled — this is a cytostatic compound at its IC50 by this definition.
• Absolute cell number (true cytotoxicity): a cell count ≤ T=0 means net cell loss — cytotoxicity. Only possible with a nuclear count assay or an absolute cell count (hemocytometer, Coulter, flow). Prefer this normalization when the claim is "compound kills cells."

State explicitly in every paper and report which normalization was used. An IC50 from GI50 normalization is not directly comparable to an IC50 from vehicle-control normalization.

IC50 curve fitting for viability data

The three-parameter logistic is often preferred over the four-parameter for viability data because the top asymptote is fixed at 100% (untreated control) and only bottom, IC50, and Hill slope are free:

y = Bottom + (100 − Bottom) / (1 + (x / IC50)^HillSlope)

Practical requirements:

• The bottom asymptote should be allowed to be negative (cells can be below T=0) — constrain to ≥ −20% to prevent over-fitting.
• Hill slopes outside 0.5–3 indicate non-sigmoidal behavior: aggregation, solubility limits, or multi-mechanism kinetics. Flag these curves for visual inspection.
• Partial curves (< 20% effect at max concentration): report as IC50 > max concentration; do not extrapolate.
• Report R² and residual distribution; a systematically bow-shaped residual pattern indicates the wrong model (consider biphasic logistic for dual-mechanism compounds).

Selectivity — the number that matters for drug development

An IC50 in isolation is meaningless for lead selection. The selectivity index (SI) = IC50_normal_cell_line / IC50_target_cell_line is the minimal selectivity metric. For oncology compounds, the NCI 60-cell-line panel or a matched normal-to-cancer pair (e.g., MCF10A vs MCF7 for breast) provides the denominator. An SI < 5 is typically insufficient for a viable lead; SI > 20 warrants further investigation.

Apoptosis confirmation — the mechanistic layer

A compound that reaches IC50 = 100 nM by MTT could be preventing proliferation, disrupting mitochondrial function, or actively killing cells. At IC50 and 2× IC50:

• Run Annexin V/PI flow cytometry at 24 h and 48 h to distinguish apoptosis from necrosis.
• Assay caspase-3/7 activation (CaspaseGlo) — confirms intrinsic or extrinsic apoptosis pathway engagement.
• Cell cycle analysis (propidium iodide in 70% ethanol) to confirm whether growth inhibition occurs in G1, S, or G2/M.

Common sources of IC50 irreproducibility

• DMSO concentration: > 0.1% DMSO in culture medium is cytotoxic in many cell lines. Normalize DMSO across all wells to the highest concentration needed at the highest compound dilution.
• Compound solubility: precipitation at high concentrations causes a characteristic flat bottom on the dose-response curve with high variability. Check with nephelometry or a brief plate spin before reading.
• Time of day effect: cells in exponential growth have different metabolic rates at different times of day. Treat all conditions within a 2-hour window.
• Passage number drift: IC50 values shift with passage number for many cancer cell lines. Document passage number per experiment and restrict to a validated passage range.

How AiLabrix fits

Drop the plate reader export plus the compound concentration map. The pipeline imports raw viability signals, applies the chosen normalization strategy (vehicle control, T=0, or absolute count), fits the 3PL or 4PL model per compound per cell line, computes IC50 with 95% CI, flags partial curves and solubility artifacts, calculates selectivity indices against a reference cell line, and generates a compound ranking table with curve class and reproducibility score. Flow cytometry apoptosis data can be co-analyzed with viability in a combined mechanistic report. Signed PDF with curve gallery, heatmaps, and power analysis for the current experimental design. [email protected].