Immunoelectron microscopy using colloidal gold particles maps target components in the complex array of cell compartments and structures displayed in ultrathin sections [1]. Sectioning opens up compartments and presents them to the labeling system, while the gold signal represents a digital readout that is easily quantifiable. Over the past few decades, an array of methods have been developed for quantifying gold particles on sections [2–4]. These methods can be combined with appropriate sampling to allow unbiased and efficient readouts of gold particle concentration (density) over cellular structures or of the gold particle distribution over a series of compartments/structures.

Gold labeling in immuno-EM is composed of two key quantities: (a) specific labeling of target components to which the labeling is directed in the first place and (b) nonspecific labeling which represents interactions of the labeling system with nontarget components and brings noise to the system [5]. Despite the widespread use of immunogold EM, the labeling is often quantified without further analysis of the specificity. However, there is now a wide range of established methods for modifying the expression of proteins in situ, and these provide opportunities for estimating the fraction of the labeling that represents the component of interest (e.g., see [6–8]).

The principle described here is best explained using the knockdown/knockout approach in which two samples are prepared [5]: one to which no treatment has been applied and the other in which there has been a reduction in the expression of the target component. The two samples are then fixed, embedded, and sectioned before labeling with the affinity probe of interest and the colloidal gold reagent. Gold labeling is carried out under standardized conditions and as far as possible using the same solutions.

Using established methods for quantification, the first step is to estimate labeling density over compartments of interest under the two conditions. The densities may be expressed as gold particles per unit profile area or membrane profile length, each of which can be estimated stereologically. The first key indicator of specificity is the difference in the density of labeling between the untreated and the treated (knockdown/knockout) sample. This density drop provides an estimate of the specific labeling density. The next step is to calculate the fraction of labeling that is specific by comparing the change in labeling density with the gold signal obtained under conditions of initial expression. If the fraction specific is, for example, 0.8, then each gold particle has an 80 % chance that it labels the target component under the initial condition. Finally, it is possible to correct the initial labeling distribution by removing nonspecific background signal. This is done by multiplying the fraction specific and the number of gold particles counted for each compartment in the initial condition. In the case of complete knockdown, the specific labeling distribution will reflect the distribution of available molecules in the sample. If a knockdown is partial, then the specific labeling will represent that fraction of the component.

This same principle can be applied to labeling reagent controls such as peptide inhibition (see example below), a, albeit, less rigorous way of controlling specificity. An additional approach to establishing specificity is when expression in the sample has been enhanced (by expressing tagged protein, novel targets, or uptake of exogenous molecules). Here the strategy is to calculate the fraction specific by comparing the change in labeling density with the signal density after enhancement (again see below).