ELISA for Peptide Quantification and Its Limits

A sample has degraded substantially. The assay reports full concentration anyway, and reports it confidently. Nothing malfunctioned. An immunoassay was asked how much peptide was present and answered a different question: how much material two antibodies were willing to recognize. Those two quantities diverge the moment anything in the sample resembles the target closely enough to pass for it.

Immunoreactive material is not intact peptide

Raise an antibody against a peptide and what it learns is a stretch of that sequence. Anything else carrying the same stretch becomes a candidate for detection, and three consequences follow that are routine rather than exotic in this material.

Breakdown products that kept the epitope get counted as though nothing had happened to them, which is exactly the scenario in the opening paragraph. Analogs separated from the target by one residue or by a terminal modification can register just as strongly, which is precisely the predicament with closely related sequences of the sort set side by side in three compounds one dalton apart. And a fragment considerably shorter than the target can react fully.

So the reported quantity is immunoreactive material. Wherever that distinction carries weight, and for any stability question it always does, the work belongs to a separation-based method instead, because a column resolves what an antibody cannot.

How the antibody came to exist, and why that limits it

Short peptides make poor immunogens by themselves, so raising antibodies against one requires conjugating it to a carrier protein first. The conjugation chemistry anchors the peptide by one end, meaning the antibodies that emerge have learned the exposed stretch and may be entirely blind to whatever region served as the attachment point.

There is a direct consequence for what any resulting assay can perceive. Couple through the C-terminus and the antibodies tend to key on N-terminal features, leaving a C-terminal change, an amidation say, or a residue lost from that end, potentially invisible. Where that very modification is what separates two compounds, as with the terminal chemistry described in acetylation and amidation, the assay has no means of telling them apart whatsoever.

What the plate is actually doing

Under the familiar sandwich arrangement, an antibody fixed to the plate captures the analyte while a second labeled antibody attaches to a different region of it, and the label throws off signal in proportion to how much got trapped between the two. Converting that signal into a concentration requires a standard curve built from known quantities.

Demanding two independent binding sites is what lends the format its specificity, and it is equally what makes the format awkward for short peptides, which may not display two epitopes separated widely enough for both antibodies to occupy at once. Small analytes are therefore handled competitively instead: sample and a labeled counterpart compete for a restricted supply of antibody, so rising concentration drives signal downward.

The curve is an assumption in disguise

Every concentration reported is read off a curve constructed from a reference preparation, and the whole number rests on that preparation being what its label claims.

Where the standard is the same material under measurement, what the assay delivers is a ratio, carrying forward whatever content uncertainty the standard itself has, the problem examined in why a 10mg vial is not 10mg of peptide. Weigh out a standard without allowing for water and counter-ion and it runs systematically high, and every sample measured against it runs systematically low by the very same proportion.

Range is the other constraint. Anything read beyond the topmost standard is extrapolation rather than measurement, and a competitive curve is steep through its middle and close to flat at both extremes, so samples landing at either end arrive with enormous uncertainty attached.

Whatever else is in the tube

Samples are not buffer. Serum, medium and everything else present can obstruct antibody binding, coat the plate, or add background of their own.

Spike-and-recovery is the standard guard against this: put a known amount of analyte into the genuine sample matrix and see whether the assay hands it back. Recovery well away from complete indicates interference, and the results then need correcting or the approach needs changing. Dilutional linearity runs in parallel: halve a sample and the reading should halve. Where it does not, something in the matrix is acting on the result in a concentration-dependent way.

Two ways to get a wrong number without noticing

First, the high-dose hook. Push analyte concentration far enough in a sandwich format and both antibodies become independently saturated, so no sandwich forms and the signal drops. A wildly concentrated sample then reads low. Two dilutions expose it; one dilution does not.

Second, losses that occurred before anything reached the plate. Preparation, dilution and transfer all happened upstream, and peptide disappears onto surfaces along that route, as adsorptive loss describes. The assay then measures, accurately, whatever survived the journey.

What it is genuinely good at

These assays reach concentrations chromatography cannot, cope with complicated samples, and run many at once. Nothing else is as practical for finding a known analyte at low concentration in a messy matrix.

They are the wrong instrument for settling identity, for judging purity, and for any question requiring degradation products to be told apart from intact material. All of those sit with the techniques discussed in HPLC and mass spectrometry.

Interrogating a reported concentration

Which standard was used, and how its content was pinned down. Whether cross-reactivity was characterized against the degradation products and related analogs that were plausible, and if so at what percentage. Whether spike-recovery and dilutional linearity were run in the real matrix rather than in buffer. Whether the samples themselves landed inside the calibrated range.

Absent those, a concentration is nothing more than a signal pushed through an uncharacterized curve, and the sensible way to treat it is as an order of magnitude and not as a figure.