A Simple Method for Comparing Immunogold Distributions in Two or More Experimental Groups Illustrated Using GLUT1 Labelling of Isolated Trophoblast Cells

Abstract

Colloidal gold-labelling, combined with transmission electron microscopy, is a valuable technique for high-resolution immunolocalization of identified antigens in different subcellular compartments. Whilst the technique has been applied to placental tissues, few quantitative studies have been made. Subcellular compartments exist in three main categories (viz. organelles, membranes, filaments/tubules) and this affects the possibilities for quantification. Generally, gold particles are counted in order to compare either (a) compartments within an experimental group or (b) compartmental labelling distributions between groups. For the former, recent developments make it possible to test whether or not there is differential (nonrandom) labelling of compartments. The methods (relative labelling index and labelling density) are ideally suited to analysing label in one category of compartment (organelle or membrane or filament) but may be adapted to deal with a mixture of categories. They also require information about compartment size (e.g. profile area or trace length). Here, a simple and efficient method for drawing between-group comparisons of labelling distributions is presented. The method does not require information about compartment size or specimen magnification. It relies on multistage random sampling of specimens and unbiased counting of gold particles associated with different compartments. Distributions of observed gold counts in different experimental groups are compared by contingency table analysis with degrees of freedom for chi-squared (χ²) values being determined by the numbers of compartments and experimental groups. Compartmental values of χ²which contribute substantially to total χ²identify the principal subcellular sites of between-group differences. The method is illustrated using datasets from immunolabelling studies on the localization of GLUT1 glucose transporters in cultured human trophoblast cells exposed to different treatments.

Introduction

The use of colloidal gold-labelled protein-A or secondary antibody, combined with transmission electron microscopy, provides a valuable technique for high-resolution immunolocalization of one or more identified antigens in different subcellular compartments [1]. Whilst the technique has been applied to placental tissues [2], [3], [4], [5], [6], [7], [8], there have been few attempts to abstract quantitative data. Subcellular compartments exist in three main categories, viz. organelles (volume structures), membranes (surface structures) or filaments/tubules (linear structures). This fact influences the possibilities for immunogold quantification which is applied in two main ways: (a) to compare compartments within an experimental group, or (b) to compare compartmental labelling distributions between groups. In both cases, robust tools for specimen sampling, stereological estimation and statistical evaluation of subcellular gold-labelling distributions are important design features [9].

For drawing within-group comparisons, recent developments make it possible to test whether or not there is differential (nonrandom) labelling of particular compartments by estimating a labelling density (LD) and/or a relative labelling index (RLI) for each compartment of interest [9], [10]. RLI >1 when compartments are preferentially labelled and the RLI of each compartment can be estimated directly [9]or indirectly via LD values [10]. These new methods have been used to analyse the distribution of surfactant protein A in type II pneumocytes, lysosome-associated membrane protein-1 in macrophages, vault poly(ADP-ribose) polymerase in HeLa cells and prion proteins in neurons [10], [11], [12].

Estimating RLI or LD is attractive because, by relating counts of gold particles to the sizes of compartments, data proportional to antigen concentration is obtainable [1]. Reducing the estimation of compartment size to a simple count of test points or line intersections improves efficiency [10]. However, the fact remains that, to compare compartments within an experimental group, RLI and LD are best suited to dealing with label localized in, say, organelle compartments or membrane compartments rather than in a mixed set of categories [9], [10].

If the purpose of a study is merely to compare compartmental labelling distributions in different experimental groups (control versus treated, different levels of drug exposure, time series, etc), it is not necessary to estimate RLI or LD. A further potential benefit is that such comparisons need not be constrained by the category of compartment but are applicable to quantifying label localized in organelles and membranes and filaments. Here, a simple and efficient method for the characterization and statistical evaluation of between-group differences in labelling distribution is presented. Its utilityis illustrated by reference to the distribution of GLUT1 in cultured trophoblast cells exposed to different glucoseconcentrations [6].