Cell Viability and Cytotoxicity Assays in Peptide Work: What MTT, ATP, LDH and Live/Dead Actually Measure

A cell viability assay does not measure viability. Each of the common formats measures one physical quantity — a dye reduction rate, a luminescence proportional to an analyte concentration, an enzyme activity in a supernatant, a fluorescence ratio — and "viability" is an interpretation laid over that quantity. The interpretation holds only while the relationship between the measured quantity and the number of living cells stays stable across the wells being compared. A test article is added precisely to change conditions, so that assumption is under attack in every experiment worth running.

This guide covers four formats used in peptide bench work: tetrazolium reduction (MTT, MTS, XTT, WST), ATP luminescence, lactate dehydrogenase release, and live/dead fluorescence staining. The subject is what each is physically sensitive to, where each is confounded, and why a percentage labeled "viability" carries far less mechanistic information than its name suggests.

Tetrazolium assays measure a reduction rate distributed across the cell

The MTT assay is often described as reporting mitochondrial succinate dehydrogenase activity. The primary characterization does not support that description as an exclusive account. Working in the bone marrow-derived 32D line, Berridge and Tan found that MTT reduction by whole cell homogenates was greatest with NADH as substrate and least with succinate, and that succinate-driven activity accounted for less than 10% of the combined activities; with succinate, 96% of recoverable MTT-reducing activity was in particulate fractions and 77% in the mitochondrial and light mitochondrial/lysosomal fractions, whereas with NADH and NADPH more activity was associated with soluble fractions ^[1].

The practical consequence is that formazan production integrates several reductive processes, and in a later study of the assay's pitfalls, Stepanenko and Dmitrenko describe MTT reduction as occurring throughout a cell and as something that can be significantly affected by a number of factors, including metabolic and energy perturbations, changes in the activity of oxidoreductases, and endocytosis, exocytosis and intracellular trafficking ^[2]. They argue that over- and underestimation of viability can arise both from adaptive metabolic and mitochondrial reprogramming under drug-treatment stress and from off-target effects of the inhibitors themselves ^[2]. Neither of those is cell death, and neither is cell number.

Two distinct classes of tetrazolium artifact

The first class is direct chemical reduction, which happens with or without cells. Bruggisser and four colleagues obtained conflicting readouts when testing kaempferol on breast cancer cells: cell number fell by crystal violet staining while MTT formazan formation increased, and kaempferol, resveratrol and extracts of Hypericum perforatum and Cimicifuga racemosa all produced instantaneous formazan in the absence of cells, while the antioxidants ascorbic acid, vitamin E and N-acetylcysteine are reported only as interfering with the assay; washing kaempferol-treated MCF7 and HS578 cells before adding MTT significantly reduced the direct dye reduction, and the authors conclude the assay may lead to false positive results when testing natural compounds with intrinsic reductive potential ^[3]. Medium components do the same. Funk, Schrenk and Frei report that bovine and human serum albumin preparations produced a concentration-dependent increase in XTT signal, and the same effect with human albumin in MTT, which they describe as leading to overestimation of cell numbers and underestimation of cytotoxicity; they attribute it to the free cysteine residue, note the same behavior for free cysteine and glutathione, and show that blocking the thiol with N-ethylmaleimide decreased the interference in XTT ^[4].

This class is detectable. A cell-free plate — medium plus test article plus dye, no cells — will show it.

The second class will not appear in that control, and this is where the assay is most dangerous. Maioli and five coauthors found that rottlerin caused overestimation in the MTT test on MCF-7 cells while showing no reactivity toward tetrazolium salts in vitro and no modulation of lactate dehydrogenase activity; the interference occurred only in cultured cells, alongside a decrease in energy charge, and because uncoupling agents produced the same overestimation the authors conclude the artifact is linked to rottlerin's uncoupling action, which by accelerating the oxidative chain incidentally enhances MTT reduction ^[5]. A cell-free blank is necessary. It is not sufficient.

The peptide-specific case: formazan trafficking

There is a confound in this second class that peptide researchers meet directly. Liu and Schubert reported that amyloid β peptide and human amylin dramatically enhance MTT formazan exocytosis, resulting in inhibition of cellular MTT reduction; notably, only the amyloid peptides known to be cytotoxic enhanced exocytosis, and the authors suggest that formazan exocytosis is a multistep process and that cytotoxic amyloid peptides enhance it through an intracellular signal transduction pathway ^[6].

Isobe, Michikawa and Yanagisawa examined the same phenomenon in cultured rat astrocytes with two different tetrazolium salts and found that MTT reduction was inhibited by amyloid β treatment while WST-8 reduction was unaffected, alongside early extracellular appearance of formazan and rapid disappearance of intracellular formazan granules; chloroquine, a perturbant of membrane trafficking, gave similar results, and the authors suggest exocytosis is enhanced in a similar manner by both ^[7].

Two salts, one treatment, and only one of them registered an effect in the same cells. Any peptide that alters membrane trafficking, aggregates at the cell surface, or associates with membranes has a plausible route to this artifact, and the reported "viability" would then be a measurement of where the formazan went.

When the readout inverts

Interference does not only blunt an effect; it can reverse the sign. Ganapathy-Kanniappan and seven coauthors compared MTS against trypan blue staining, ATP depletion and annexin V staining in Vx-2 and HepG2 cells and found that MTS indicated increased cell viability at 1 mM and 2 mM 3-bromopyruvate while the other three methods showed the cells were not viable under the same conditions; cell-free medium confirmed a dose-dependent absorbance increase from direct interaction with the reagent, and the authors recommend employing multiple biochemical and morphological methods ^[8].

A researcher holding only the MTS plate would have written down a protective effect.

ATP luminescence: more sensitive, differently confounded

ATP-based luminescence measures the ATP concentration in a lysate, using it as a proxy for the number of metabolically intact cells. Its principal advantage is dynamic range at low cell numbers. Petty, Sutherland, Hunter and Cree compared the two formats across 1,500–100,000 cells per well in Daudi and CCRF-CEM lines and found the ATP assay detected the lower limit of 1,563 cells per well at luminescence at least 100 times background, whereas the MTT assay could not detect fewer than 25,000 cells per well above background ^[9] — a difference of 25,000 / 1,563 = 16.0-fold between those two figures. Because 1,563 cells per well is the lowest cell number the authors report the ATP format detecting rather than a measured limit of detection, that ratio is a lower bound on the gap, and it holds for those two lines.

The confound is that the detection chemistry is an enzyme, and enzymes can be inhibited by the thing being tested. Thorne, Auld and Inglese note that compound interference is especially hard to identify when it is reproducible and concentration-dependent — characteristics normally taken as evidence of genuine activity — that compounds inhibiting firefly luciferase are a recognized class in screening libraries, and that while reactive chemical groups were once thought to be the primary source of such interference, recent work suggests compound aggregation may play a more significant role in many assay formats ^[10]. That work has also led, in their account, to practices with the potential to reduce interference, for example adding non-ionic detergent to assay buffer to reduce aggregation-based inhibition ^[10]. The effect is not hypothetical: Heitman and five coauthors traced a hit from a receptor reporter-gene screen to direct inhibition of firefly luciferase with an IC50 of 1.7 ± 0.1 µM, competitive with respect to luciferin, and prepared an analogue reaching 0.069 ± 0.01 µM ^[11].

An inhibited luciferase and a dead cell both give low light. A cell-free luminescence check against a defined ATP standard in the presence of the test article separates them. That practice is convention rather than a finding; it is stated here uncited.

LDH release measures membrane integrity and nothing else

Lactate dehydrogenase release is the most mechanistically narrow of the four. Chan, Moriwaki and De Rosa describe permeabilization of the plasma membrane as a key signature of necrotic cells, quantifiable in tissue culture by measuring release of intracellular LDH, and state that measuring LDH release is a useful method for detecting necrosis when combined with other methods ^[12]. That framing is the point. LDH release does not report growth arrest, does not report apoptosis before membrane failure, and does not report metabolic suppression. Cells can lose most of their proliferative capacity with no LDH signal at all.

It also has its own arithmetic trap, because the readout is normalized against two controls rather than one. The conventional expression is (experimental − spontaneous) / (maximum − spontaneous) × 100, where spontaneous is vehicle-treated cells and maximum is a lysis control. The formula, and the practice of running both controls on the same plate, are bench convention and are stated here uncited.

A worked example, with the denominator made explicit

The following absorbances are hypothetical; the arithmetic is worked, not measured. Suppose vehicle wells read 0.180, the lysis control reads 1.480, and the treated wells read 0.505.

• Numerator: 0.505 − 0.180 = 0.325
• Denominator: 1.480 − 0.180 = 1.300
• Result: 0.325 / 1.300 = 0.250, or 25.0% cytotoxicity

Two common mistakes change the number without changing the plate. Dividing by the raw maximum instead of the corrected maximum gives 0.325 / 1.480 = 0.220, or 22.0%. Dividing raw treated by raw maximum gives 0.505 / 1.480 = 0.341, or 34.1%. One dataset, three publishable-looking figures.

The lysis control is the more consequential of the two. Suppose the detergent step in fact liberates only part of the cellular LDH, and complete lysis would have read 1.850. Then the correct denominator is 1.850 − 0.180 = 1.670, and the result is 0.325 / 1.670 = 0.195, or 19.5%. The incomplete lysis control released 1.300 / 1.670 = 77.8% of the total releasable LDH, and the reported cytotoxicity was inflated from 19.5% to 25.0%. Because the numerator is unchanged, the inflation is exactly the ratio of the two denominators, 1.670 / 1.300 = 1.285. So the reported figure overstates the corrected one by 28.5% when the corrected value is the denominator, and the same gap is 0.370 / 1.670 = 22.2% when the reported value is the denominator. Which denominator is in use has to be stated, because 28.5% and 22.2% are not the same claim. Note also that subtracting the rounded percentages and dividing, 5.5 / 19.5, returns 28.2% rather than 28.5%; the relative figure has to be computed from the unrounded ratios.

Live/dead staining: dyes with their own biology

Two-color live/dead staining pairs an esterase-activated dye retained by intact cells with a nucleic acid stain excluded by intact membranes. It is the only format of the four that reports on individual cells rather than a well average, which makes it the right tool for distinguishing a uniform partial suppression from a mixed population of healthy and dead cells — a distinction no absorbance or luminescence value can make.

Its main confound is that the esterase-activated dye is a transporter substrate. Essodaigui, Broxterman and Garnier-Suillerot performed a kinetic analysis in P-glycoprotein-positive K562/ADR and multidrug resistance protein-positive GLC4/ADR cells and characterized calcein acetoxymethyl ester as a common substrate for both proteins, with free calcein effluxed only by the multidrug resistance protein; efflux Vmax for both species by that protein was similar at 1.0–1.2 × 10⁵ molecules per cell per second, while the Km for the ester was much lower, at 0.05 µM, than the 268 µM they report for free calcein ^[13]. In a line expressing these transporters, retained fluorescence is a function of pump activity as well as of esterase activity and membrane integrity, and an inhibitor of the pump would, on this reasoning, read as improved staining — an inference from the transport kinetics rather than a result those authors report.

Why a viability drop is not a mechanism

Every finding above shares a structure: a treatment changed one measured quantity, and the quantity was not the number of living cells. Even when the reading is real, the inference is weak, because a reduced tetrazolium signal is consistent with fewer cells, with the same number of cells reducing more slowly, with faster export of the product, or with the same cells arrested in cycle. Distinguishing those requires a second measurement whose failure modes do not overlap the first: a direct enumeration, a membrane-integrity readout, or a second dye with different chemistry, as in the MTT-versus-WST-8 comparison above ^[7].

Three reporting habits do most of the work, and all three are convention rather than evidence; they are stated here uncited. Report the endpoint time and the seeding density, since both change what a fixed absorbance means. Report the vehicle-control value that served as the denominator, not only the ratio. Keep absorbance readings inside the instrument's linear range, since a plate read at the top of its scale compresses differences that the arithmetic then treats as real.

Correcting a viability ratio for a compound's own signal

These absorbances are also hypothetical. Suppose vehicle wells read 0.920, treated wells read 0.640, cell-free medium with the test article reads 0.075, and cell-free medium alone reads 0.045.

The uncorrected ratio is 0.640 / 0.920 = 0.696, reported as 69.6% viability, a 30.4% loss. Subtracting the matching blank from each arm gives 0.640 − 0.075 = 0.565 for treated and 0.920 − 0.045 = 0.875 for vehicle, so the corrected ratio is 0.565 / 0.875 = 0.646, or 64.6% viability and a 35.4% loss. The 5.0 percentage points between those two figures split into two parts. Subtracting only the common medium background from both arms gives 0.595 / 0.875 = 0.680, or 68.0%, so the medium accounts for 1.6 points; the compound's own reduction, the 0.075 − 0.045 = 0.030 absorbance units present in the treated arm alone, accounts for the remaining 3.4. Relative to the corrected value, the uncorrected figure overstates survival by 5.0 / 64.6 = 7.7%.

What is not established

Several things a reader might want are not settled by the sources here. There is no general threshold below which a viability change can be called insignificant; the sources characterize specific artifacts in specific systems, not a universal noise floor. The prevalence of tetrazolium interference across peptide chemistries is unknown — the exocytosis findings come from amyloid β and human amylin in neural and glial cultures ^{[6, 7]}, and whether other peptide classes behave the same way has not been established by these studies. Whether a compound that interferes in a cell-free tetrazolium check will also interfere in ATP or LDH formats is not predictable from the interference itself; it has to be tested. And where the sources report an artifact in one or two cell lines ^{[1, 5, 8]}, the finding is a demonstration that the artifact exists, not a measurement of how often it occurs.

The defensible position is narrow and worth holding: a viability assay tells you that a number moved. Which number moved, and why, is a separate experiment.