This protocol was optimized using an inverted microscope equipped with a 1.45 NA 100× oil immersion objective. Highly fluorescent organic dyes (e.g., ATTO647N or Cy3) are detected using a sensitive and rapid EM-CCD camera. Fluorescence excitation sources are CW lasers (e.g., frequency doubled Nd:Yag laser, solid state laser or He:Ne laser). Excitation laser beams enter through the fluorescence epiport of the microscope and illuminate a wide field area of 10–20 μm in diameter of the sample by focusing the beam in the back aperture of the objective. For uPAINT, oblique illumination is required to avoid exciting the out of focus fluorescent ligands in the solution. This is achieved by translating the focused beam with respect to the axis of the objective, on the periphery of the back aperture of the objective (Fig. 2), akin to total internal fluorescence microscopy. Illumination intensities are of a few kW/cm². For each type of dye used, an appropriate set of fluorescent filters is required. The total detection efficiency of a typical experimental single molecule setup is in the range of 5–10%.

With this detection efficiency, 1,000 counts are commonly detected in 50 ms from a typical single fluorophore (11). This translates in a typical pointing accuracy (or localization precision) of 40–50 nm. We emphasize that different definitions of the pointing accuracy are often used. Here, we use the Full Width at Half Maximum (FWHM) of the distribution of the localizations obtained from a fixed molecule which is repetitively imaged. With this definition and using the Rayleigh criterion to define the resolution at which two molecules can be discriminated, the resolution is equal to the pointing accuracy. Some authors alternatively use the standard deviation of fixed molecule localizations for defining the pointing accuracy or localization precision, which provides values 2.3 times lower than the definition used here. We generally prefer our definition, since, in the latter case, the pointing accuracy gives lower values than the resolution (factor of 2.3), which might induce confusing comparisons of different sets of data.

If longer integration times are used (at the price of fewer data points in a trajectory), improved pointing accuracies are obtained. For a shot noise limited detection, the pointing accuracy will improve with the inverse of the integration time (10) yielding ∼20–30 nm resolution for 150 ms integration time. It is worth mentioning that when mobile molecules are imaged, the movements of the molecules might affect the pointing accuracy.

The number of points of a single molecule trajectory depends on the photo-physical properties of the imaged dye. In practice, and for a given dye, it depends on the excitation intensity and integration time used. Typically, one uses close to video rate imaging (50 ms integration times) at saturation intensities (∼1 kW/cm² for the most common dyes). With these parameters, one can obtain typically 10–15% of the trajectories lasting more than 1 s, in the case of ATTO647N dyes (7) (Fig. 3).

With uPAINT, and unlike standard single molecule methods, single point source emitter detection is ensured by the real time imaging of each individual ligand binding to its target membrane molecule regardless of the number of fluorophores it carries. Thus, one can take advantage of this feature to improve the pointing accuracy by coupling multiple dyes per ligand. For instance, with four fluorophores per ligand instead of one, the average pointing accuracy is improved by a factor of 2 under identical excitation intensity. Optionally, in order to lengthen the obtained trajectories, one can reduce the excitation intensity while keeping the same pointing accuracy as when a single dye molecule is used (7).

By creating a thin sheet of light above the coverslip surface, oblique illumination allows selecting individual fluorescent ligands which have bound to the cell surface while rejecting the great majority of molecules in solution. This improves signal to noise ratios and avoids photobleaching of unbound fluorescent ligands. The choice of the angle depends on the thickness of the cells to be imaged. A typical choice is ∼5° such that the resulting sheet of light has a thickness of Δx ∼2 μm in the center of the field of view. Assuming 3-dimensional isotropic diffusion of the fluorescent ligands freely diffusing in the solution, the diffusion coefficient D is well known and relates the ligand hydrodynamic diameter d and the fluid viscosity η by the Stokes-Einstein relation: , where k _B is the Boltzmann constant and T is the temperature. For ligand hydrodynamic sizes of d ∼5–10 nm, D is of the order of ∼40–80 μm²/s and the average time they spend in the 2D sheet of light is 2Δx ²/2D which lies below ∼100 ms. This indicates that such unbound fluorescent molecules should not appear statistically in more than one or two consecutive images if 50 ms integration times are used. Such unwanted events can be rejected in the analysis by omitting the first two molecular detections in each trajectory (see below).