Viability Assays Measure Metabolism, Not Life

Here is a report you can find without much searching: “viability” quantified by a tetrazolium assay, read at one time point, with no interference control anywhere in the design. What that experiment measured is the rate at which something in the well reduced a dye. What it announced is that cells survived or did not. The announcement may even be true. Nothing in the experiment establishes it, and the gap between those two sentences is where a great deal of this literature gets into difficulty.

The chemistry the plate reader sees

Tetrazolium assays and their resazurin-based cousins share a mechanism. Dye goes into the well, something inside the cells reduces it, and the reduced form is quantified by color or by fluorescence. NAD(P)H-dependent oxidoreductases do most of that reducing. The quantity being measured, therefore, is reductive metabolic capacity present in the well.

Cell number enters the picture only through an assumption that is seldom stated and often wrong: that each cell contributes a constant amount of metabolic activity. Hold that assumption and the assay serves as a fair stand-in for how many living cells are present. Break it, by testing a compound that alters metabolism, and the assay ceases to stand in for anything.

Three questions, one dye

A metabolic dye gets pressed into service as a proxy for all three of the questions below and directly measures none of them.

Watch what a falling signal is compatible with: fewer cells, an unchanged number of less active cells, or a population that simply stopped dividing. Those are three distinct findings and the dye reports the same drop for each.

When the assay reads out the mechanism under study

For work in this catalog the concern is concrete rather than philosophical. Anything acting on mitochondrial function alters the rate of dye reduction directly, with no cell needing to die or divide for the number to move.

Apply a tetrazolium readout to a mitochondrially-targeted compound and part of what you have generated is a measurement of the very mechanism under investigation, rather than an independent verdict on whether the cells came through it. Compounds affecting NAD availability raise the identical problem, NAD(P)H being the species that performs the reduction; see the coenzyme role laid out in what a coenzyme is and is not along with the targeting discussed in SS-31 and cardiolipin.

Under those conditions a rising signal might mean more cells, or healthier cells, or an unchanged population reducing dye more quickly. No quantity of replication will resolve which, because the ambiguity is structural rather than statistical.

Signals with no biology behind them

Four artifacts move the reading without any cellular event occurring. The first is direct chemical reduction: a reducing agent sitting in the well can reduce the dye entirely on its own, with ascorbate and thiol-bearing material the familiar culprits. A cell-free well containing compound plus dye catches it, filling the role that the solvent condition plays in vehicle controls.

Optical interference is the second, and it is immediate rather than theoretical for a visibly colored copper complex of the sort noted in GHK and GHK-Cu; anything colored adds absorbance of its own. Third, material dropping out of solution scatters light and shifts the reading, which is among the reasons for pinning down the solution behavior covered in peptide aggregation before the plate is ever seeded. Fourth and most mundane: over a long incubation, evaporation concentrates everything in the perimeter wells.

One reading is a snapshot of a process

Single-timepoint data forces a process into a still image. Suppose one compound delays division and another kills a fraction of the population. At 24 hours both can produce an identical reduction in signal. By 72 hours the two have separated entirely.

Which means a lone reading cannot distinguish cytostatic from cytotoxic. Where that distinction carries weight, the fix is additional time points or a different assay. Reading the same assay more precisely does not help.

Exposure is not what you pipetted

These experiments typically run somewhere between 24 and 72 hours, ample time for the actual exposure to drift away from the nominal one. Peptide adsorbs onto the plate in the manner described under adsorptive loss to surfaces, and peptidases in serum-containing medium degrade what remains.

So an apparent absence of effect at a stated concentration may record an absence of exposure rather than an absence of activity. When a compound does nothing across an entire concentration range, the hypothesis to eliminate before drawing any conclusion is that very little of it was still present and free when the plate was read.

What would make the number mean something

Agreement between two assays that work by different mechanisms is the central requirement. Pair a metabolic readout with a membrane-integrity readout and each answers its own question while jointly constraining the interpretation, exactly the orthogonality principle that governs analytical work and that is described in confirming with two independent methods.

Around that, four additions: a cell-free control ruling out direct reduction, an optical control containing compound alone, readings at more than one time, and a genuine cell count somewhere in the study to anchor what the signal actually corresponds to. When assessing someone else’s result, those are the same five things to look for, together with the question of whether the compound could plausibly interfere with the assay chemistry in the first place.