Dynamic Light Scattering

Very little. The low-volume quartz cuvette takes 12 µl, the small disposable cuvettes 45 µl, and a standard measurement in a general-purpose cell around 70 µl. The sample comes back afterwards, since nothing is consumed.

Concentration scales inversely with size, because scattering rises steeply with particle volume. A working estimate for proteins is 1.44/MW(kDa) mg/ml as a lower limit: about 0.6 mg/ml for a 10 kDa protein, about 0.06 mg/ml for a 100 kDa one. Most people work between 0.5 and 2 mg/ml, and data quality improves with concentration until you reach the point where the particles start to interact and the apparent size begins to fall. For a new sample a two- or three-point concentration series is the safe way to check that you are not in that regime.

Filter or spin it, then handle it as if it were an optics sample rather than a biochemistry sample.

Dust is the main enemy. A single 1 µm particle drifting through the beam scatters as much as a million 10 nm proteins and produces the spikes and jumping baselines that ruin a correlogram. Filter the buffer at 0.02-0.1 µm, filter or centrifuge the sample — 10 minutes at 15 000 g is often enough, and gentler than forcing a fragile complex through a membrane — and pipette slowly to avoid bubbles.

Cuvettes must be clean and handled by the frosted faces. Detergents left from washing show up as a micelle peak around 5-10 nm and look convincingly like a protein.

The primary measurement is the intensity autocorrelation function. From it a cumulant fit gives the z-average hydrodynamic diameter and the polydispersity index PdI, and a distribution analysis gives peaks in intensity, which can be converted to volume and number distributions using the refractive index of the material.

For proteins the rules of thumb are: PdI below 0.1 means a monodisperse sample; 0.1-0.4 means moderately polydisperse and usually still interpretable; above 0.4 the distribution analysis is not meaningful and only the presence of large material can be reported. A percent polydispersity below 20% is the corresponding criterion in the protein-specific analysis.

The hydrodynamic diameter is the diameter of a sphere with the same diffusion coefficient as your particle, including its hydration shell. For an elongated or flexible protein this is systematically larger than the crystallographic dimensions, which is a property of the quantity and not an artefact.

Not properly, and this is the most common misunderstanding about the technique.

DLS measures a diffusion coefficient. Converting it to a mass requires assuming a shape, and the standard conversion assumes a compact globular particle. For a well-behaved globular protein the estimate lands within a factor of two or so; for anything elongated, disordered, glycosylated or multi-domain it can be off by a factor of several. A monomer-dimer distinction needs a size ratio the method cannot deliver reliably, because two species have to differ by roughly a factor of 3-5 in diameter before DLS resolves them as separate peaks at all.

If oligomeric state is the question, SEC-MALS or AUC answers it. DLS is the right tool for a different question: is this sample aggregated, and how badly.

Because the two techniques weight the sample completely differently. Scattered intensity goes roughly as the sixth power of diameter, so a 100 nm aggregate scatters about a million times more than a 10 nm monomer. A contamination of 0.01% by mass in the form of large aggregates can dominate the intensity distribution and hide the monomer entirely.

This sensitivity is the point of the measurement, not a flaw. DLS is the cheapest way to find out, before you commit protein to a crystallisation screen, a cryo-EM grid or an SPR chip, that a fraction of your preparation is aggregated. A gel does not see it, and a SEC column may retain it on the frit.

Interpreting the numbers with this in mind: the intensity distribution tells you what is there, the volume distribution tells you how much of it there is, and the two often look very different for a sample with a small aggregate population.

The viscosity and refractive index of the solvent, because the diffusion coefficient is converted to a size through the Stokes-Einstein relation and both enter directly. For standard aqueous buffers below 200 mM the water values are close enough.

Beyond that they are not. Ten percent glycerol raises viscosity by roughly 30%, and a size calculated with the water value would be 30% too large. Sucrose, high salt, urea and guanidinium all need their own values, so send the exact composition and we use tabulated or measured viscosities.

Detergents deserve a separate mention: above the CMC you will see the micelle population, typically 3-10 nm, alongside your protein. That is a genuine measurement of what is in the tube, but it can be mistaken for a protein species if nobody says the detergent is there.

A temperature ramp is one of the standard uses. Following size and count rate from 20 to 80 °C gives the aggregation onset temperature, and comparing that across buffers, pH values and additives is a fast way to rank formulations with milligram quantities of protein.

The same ramp distinguishes two failure modes that get confused: a protein that unfolds and then aggregates, and a protein that stays folded but aggregates colloidally at high concentration. Combining a DLS ramp with a nanoDSF melting curve on the same conditions separates conformational from colloidal stability, and the two do not always move in the same direction.

Isothermal measurements over hours also work for slow aggregation — a sample held at 37 °C and measured every ten minutes gives a growth curve.

Liposomes, nanodiscs, extracellular vesicles, virus-like particles, polymer and inorganic nanoparticles, micelles, nucleic acids and their complexes. The working range runs from roughly 0.3 nm to a few micrometres, with the upper end limited by sedimentation of the particles out of the beam during the measurement.

Zeta potential is available on the same instrument for colloidal samples. It needs more material — around 750 µl in a folded capillary cell — and low ionic strength, since a physiological salt buffer both screens the surface charge and heats up under the applied field. Samples are usually measured in 1-10 mM salt for that reason.

Membrane proteins in detergent are measurable, though the micelle signal has to be accounted for; a matched empty-micelle control is worth sending.

A single measurement is three runs of ten to fifteen seconds each, so under a minute of acquisition; with equilibration at temperature and cuvette handling, a few minutes per sample. A temperature ramp of 20-80 °C in 2 °C steps takes one to two hours.

The sample is untouched by the measurement and can be recovered, minus what stays in the cuvette. If it is precious, say so and we will use the quartz low-volume cell and return it.

Because the measurement is fast and cheap in material, it is worth running DLS as a first check on any sample that will then go into a longer experiment, and we often do it as a matter of course before ITC or SPR.

DLS is the easiest instrument here to learn to operate, and the hardest to interpret carelessly. A session or two covers cuvette handling, the standard operating procedures and the acquisition settings.

The part that takes longer is knowing when to distrust a result: number fluctuations from too few particles, a correlogram that does not decay to baseline, an intensity peak at 1000 nm that is one dust particle rather than a population, a second peak that the software reports at 45% because the volume conversion used the wrong refractive index. We go through those cases with the data rather than in the abstract.

Sending samples as a service works equally well, and the report includes the correlograms so that you can see for yourself whether the fit deserves belief.