The Machine That Learned to Weigh a Molecule

Picture a chemist in about 1980 holding a vial of a purified hormone and wanting to know one thing: is this actually the molecule I think it is? There is a machine down the corridor built for exactly that question. It weighs molecules, and it weighs them with a precision no balance can approach — it will tell you a mass to four decimal places. The chemist cannot use it. To get a sample into that machine, the sample has to be turned into a gas of charged particles, and the only reliable way to do that is to heat it until it evaporates and then knock electrons off it. A protein put through that treatment does not evaporate. It chars. It stays in the inlet as a brown film, and the machine, which is capable of extraordinary things, reports nothing at all.

That was the situation for something like sixty years, and it is easy to underrate how much it shaped what biochemistry could ask. The most powerful analytical instrument of the century had a size limit somewhere around a thousand daltons, and almost everything interesting about proteins and peptides lived on the far side of it.

Weighing by flight

The principle underneath every mass spectrometer ever built is almost embarrassingly simple. Give a particle an electric charge, then push it with an electric or magnetic field. A heavy particle responds sluggishly; a light one responds briskly. Measure how it responds — how far it bends in a magnetic field, or how long it takes to fly down a tube — and you have measured its mass, or rather its mass divided by its charge, which is the number these instruments actually report.

This was worked out in the first quarter of the twentieth century and it was immediately magnificent. It resolved the isotopes of the elements. It underpinned the chemistry of petroleum. By the middle of the century a mass spectrometer was a standard tool for identifying organic compounds, and it was better at the job than anything else, because mass is not an approximate property. Two compounds with different formulas have genuinely different masses, and a good instrument can tell them apart at the level of a single hydrogen atom.

The catch was in the first word of the principle: charge. Everything downstream assumes the molecule is already an ion, already flying free in a vacuum. Getting it there was called ionisation, and every method available was violent. Electron impact fired a beam of electrons at a vapour. Chemical ionisation was gentler but still needed the sample in the gas phase to begin with. And nothing gets a large, polar, hydrogen-bonded molecule like a peptide into the gas phase by heating, because its intermolecular attractions are stronger than its own covalent bonds. Push in enough energy to lift it into vapour and you have already put in enough to take it apart.

So the field developed a folk understanding that biological macromolecules were simply outside the instrument's remit. Chemists weighed small molecules. Biochemists worked out sequences by chemistry and estimated sizes by how far a band ran down a gel. The two worlds barely touched.

The idea that waited twenty years

In 1968 Malcolm Dole and colleagues published a paper with the arresting title Molecular Beams of Macroions, describing an attempt to get large charged molecules into the gas phase by spraying a solution through a fine charged nozzle and letting the resulting droplets evaporate 1. The reasoning was that if a droplet carrying a charge shrinks, the charge has nowhere to go but onto whatever solute is left behind. Do it carefully enough and the molecule ends up airborne and charged without ever being heated.

The paper is now treated as a founding document, but at the time it went more or less nowhere. The results were ambiguous, the instrumentation was not up to it, and the community had other priorities. The idea sat in the literature for the better part of two decades.

What revived it was John Fenn, a chemical engineer at Yale who had spent his career on molecular beams and gas dynamics rather than on biochemistry. He was in his sixties when he began working seriously on electrospray, which is worth pausing on: this was not a young researcher's speculative side project but a late-career pursuit of an unfashionable technique by someone whose expertise was in how gases and droplets behave, which turned out to be exactly the expertise the problem required.

That multiple charging is the quiet genius of the method, and it was not obvious in advance. A protein sprayed from solution does not emerge carrying one charge. It emerges carrying a whole family of charge states at once, producing not a single peak but a characteristic ladder of them. Each peak is the same molecule wearing a different number of protons, and because the spacing between them is arithmetically constrained, the ladder can be solved backwards to give one mass — usually a more accurate one than any single measurement, because it is derived from many.

Fenn's group published the result that changed things in 1989, reporting electrospray mass spectra of proteins well past the old ceiling and demonstrating the charge-state ladder in a form nobody could argue with 2. He later gave his Nobel lecture the title Electrospray Wings for Molecular Elephants, which is both a fair description of what the technique does and an unusually good joke for the genre 4.

A second answer, arriving at the same time

Techniques rarely arrive alone, and the late 1980s produced a second solution to the same problem from an entirely different direction. If the difficulty is that a laser or a heat source deposits too much energy directly into a fragile molecule, then the answer might be to give the energy to something else — to mix the molecule into a large excess of a small compound chosen because it absorbs the laser wavelength efficiently, crystallise the mixture, and fire at the crystal. The matrix takes the hit, vaporises explosively, and carries the intact analyte along with it.

Michael Karas and Franz Hillenkamp described this in 1988, reporting laser desorption of proteins with masses well over ten thousand daltons — a figure that reads unremarkably now and was startling then 3. Koichi Tanaka, working in Japan, reported a related approach using a suspension of fine metal particles in glycerol at almost the same moment. The technique became known as matrix-assisted laser desorption ionisation, universally shortened to MALDI, and it has a different temperament from electrospray: it tends to produce singly charged ions, it pairs naturally with time-of-flight analysers, and it tolerates messy samples better.

The 2002 Nobel Prize in Chemistry went to Fenn and Tanaka for soft desorption ionisation methods, alongside Kurt Wüthrich for a quite separate achievement in nuclear magnetic resonance. The omission of Karas and Hillenkamp was noticed at the time and is still discussed; the history of MALDI has more hands in it than a prize with a three-person limit can accommodate. What is not in dispute is that between them, the two techniques removed the size ceiling in the space of about five years.

What identity used to cost

It is worth being concrete about what confirming a peptide's identity involved before this, because the contrast is the whole point of the story.

You could hydrolyse the sample down to its constituent amino acids and count them — amino acid analysis, which tells you the composition but not the order, and destroys the material to do it. You could run Edman degradation, which strips residues from one end of the chain one at a time and identifies each in turn: elegant, genuinely sequence-determining, and slow, consuming a cycle of chemistry per residue and stalling if the chain has a blocked end. You could estimate size by how far the material migrated through a gel, which is not a measurement of mass so much as a comparison against other things you have already assumed something about.

All of these were real techniques that produced real knowledge. All of them were slow, sample-hungry, and answered a slightly different question from the one being asked. None of them could tell you, in an afternoon, whether the white powder in the vial had the mass it was supposed to have.

After electrospray, that question became answerable in minutes, on micrograms, without destroying the rest of the sample. And because the ions could then be selected and deliberately fragmented inside the instrument, the mass differences between the fragments could be read off as the sequence itself. Within about a decade this had generated a whole discipline, in which proteins are identified in complex mixtures at a rate no chemical degradation method could approach 5.

Why a document can claim a mass at all

Which is how the story ends, in the least dramatic place imaginable: on a page of numbers stapled to a batch of material. Any analytical document accompanying a synthetic peptide today will state a measured mass alongside the theoretical one calculated from the sequence, and the two will be expected to agree to within a fraction of a dalton.

That single line is a compressed history. It presumes an instrument that can get an intact peptide into the gas phase without cooking it, a charge-state ladder that can be solved into one number, and a calibration good enough that agreement to a decimal place means something. Every one of those presumptions was a research problem within living memory.

It is also worth knowing precisely what such a line does and does not establish. A matching mass says the material has the elemental composition the sequence predicts. It does not, by itself, say the residues are in the right order, since rearranging them changes nothing about the total. It says nothing about how much of the sample is that molecule rather than something else — that is a separation question, answered by a different instrument. And it says nothing at all about whether the compound does anything.

But it settles the first question, the one the chemist in 1980 could not get answered, and it settles it so cheaply and so routinely that the settling has become invisible. That is usually the sign that an instrument has finished changing a field: not when it produces a famous result, but when its output becomes a line on a form that nobody thinks to be impressed by.