Having detailed knowledge about the shape of the signal expected from a point object (like a molecule) means that we can mathematically derive the very centre of that signal. This is in theory the point from where the light originated - ie the precise location of the molecule. If we can do this repeatedly then we could in theory localise the precise position of every molecule in our sample - building up a map describing the molecular pattern

The ability to  constrain the illumination of a sample to around 100 nm or so using TIRFM means that we can be sure that the signals we see are from a very thin section near the cell membrane. This effectively defines the 'z-section' resolution in this mode of imaging.

A theoretically infinitely small point of light in a microscope will always appear larger than it really is. The apparent size of the object is defined by the wavelength of the light (ie the 'colour') and the optics of the microscope itself. In any case, objects will appear to be around 200-300 nm diameter, even if they are actually much smaller. In the case of single fluorescent molecules, say 5 nm radii, the light they emit will form an image in the microscope of 250 nm diameter. The point is that we know these signals originate from a smaller object than we can see. The image, right, shows the image arising from a 175 nm diameter spherical fluorescent bead - something smaller than the signal it produced. Obviously the image we get is not spherical - the complex shape is due to diffraction in the microscope. This distortion in the image, compared to the real object, can be physically predicted and mathematically described, as something called the point spread function.

The way around this could be to limit the number of labelled molecules in the sample. This will work, but suffers from the severe drawback of examining a tiny number of molecules in each cell. As each molecule is likely to behave in a stochastic manner, we would like to image ideally many 10000s or 100000s of molecules in each cell. A better solution therefore is to turn on and off the fluorescence emitted from each molecule, so that they are effectively spaced apart in time. If we could see a small subset of all the molecules present in each cell, note their positions, then turn them off before examining another group, then by the end of a (long) period, we'd have a note of the position of every molecule in the membrane of a cell (illustrated in the animation, left). If all the spots in this animation (from Colin Rickman) were visible simultaneously, they would all overlap and not be distinguishable as single objects (illustrated in the lower left panel). Spacing them apart in time allows the localisation of each single object co-ordinates - imaging as many of these discrete events as possible allows the visualisation of the underlying structure. Finally, rendering the accumulated XY - coordinates, including information regarding the uncertainty of each fit (encoded as brightness), results in the final 'super-resolution' image (below, right).

To be sure that the signals we see arise from single molecules, as opposed to small groups or aggregates, we need to apply some basic criteria to include the data in our analyses. Signals must appear the shape and size expected (defined by an understanding of the physics of the microscope system).

LSI Molecular imaging - PALM

Images and data copyright LSI Laboratory

Noting the position of every signal, from every molecule (as in the animation) delivers a huge array of precise (to 20 nm or so) XY-coordinates, describing the positions of every molecule in the membrane of the cell. These data can be reconstructed into an image (shown left) - the right hand image is a zoom from the white box in the centre of the left-hand panel. This universe of single molecules and their spatial patterns provides for the first time an accurate map to cell biologists, hopefully to use to dissect what goes wrong in conditions like diabetes. These data are from fixed, static samples - but we can extend this to look at molecules in living cells too.

This is made feasible by using fluorescent molecules we can activate, image, then destroy irreversibly. Which molecules activate is random - we can control how many will 'turn on', but not which ones. In this way we can determine a density of molecules to ensure that their signals don't overlap, repeating it often enough to determine the position of every molecule in the sample. The illustration on the right shows the localised position of each single point from the simulation on the left. Using photoactivatable variants of GFP enables this technique in cells - in this case, the approach is commonly called Photoactivation Localisation Microscopy, or PALM.

Introduction        PALM       Single particle tracking PALM        FCS        FLIM        GSD Microscopy

How can we be certain of the location of single molecules, or large cohorts of single molecules, in the plane parallel to the cover-glass - ie the xy-plane? This requires a basic understanding of how fluorescence appears in a microscope coupled with some mathematical analysis.

in the cell membrane. A problem arises when there are too many molecules, too close together. When the 250 nm signals from each point object overlap, we lose the ability to tell one from another. This is what defines resolution - if 2 molecules come within the radius of each other's signal (around 250 nm apart) - they appear as one.

The number of photons that can be emitted maximally by a single molecule can be estimated (it's in the 1000s), so each signal must not exceed this. Importantly, if the signals do arise from a single molecule, the data must behave in a quantal manner - that is, the light must appear and disappear in 'one step'.