Both normal and neoplastic mammary gland biology are the result of complex cellular interactions. These interactions are spatially heterogeneous, both locally and gland-wise, and need to be studied in a cell-by-cell basis. Given that organs are inherently three-dimensional, we rely on imaging methods that provide volumetric imaging at microscopic resolution. Since visually quantifying three-dimensional images is practically impossible, we couple our imaging methods with 3D image analysis tools that give us quantitative information without requiring strenuous human interaction.

A good example of this approach is the quantification of genetic instability in breast cancer. Genetic instability is characteristic of advanced stages of breast cancer (DCIS, invasive carcinoma), and of most solid tumors. It provides the disease with the genetic pool that supports the clonal expansion of the disease. Although it has been studied in the rough context of the evolution of the disease, an accurate model paralleling genetic instability and progression from preneoplasia (usual hyperplasia, atypical hyperplasia) to neoplasia (DCIS, invasive and metastatic carcinoma) remains to be drawn.

Quantifying genetic instability requires being able to enumerate chromosome copy number in most, if not all the cells of a tissue biopsy. Since an accurate cell-by-cell enumeration of all chromosome's number would be impractical, our approach is to approximate it using FISH probes attached to two different chromosomes and calculating the relative number of probes in every cell of thick (up to 40microns) tissue blocks of normal, preneoplastic and neoplastic tissue. This requires software for automatically delineating (segmenting) every cell in confocal image sets taken from selected volumes of the tissue and then to peer inside the nuclei to automatically detect the signals from all hybridized loci. We have successfully developed and used these tools (Ortiz de Solórzano et al 99, Ortiz de Solórzano et al 2001, Lieb at al 2000, Sarti et al 2000).

When analysis and integration of molecular and morphological information at high resolution is aimed on a bigger scale (an entire tissue biopsy or mammary gland), the only existing approach requires sectioning the tissue, followed by both histological and molecular staining of consecutive tissue sections. The analysis is normally performed by visual inspection of the sections under the microscope. This approach greatly limits the extent and accuracy of the analysis, due to the difficulty that our visual system faces when composing (extrapolating) meaningful 3D information from a series of 2D sections. Since only a few colors (2-4) can be discriminated, both in bright field and fluorescence microscopy, along with other practical problems related to multicolor IHQ or (F)ISH, we do the histological and the molecular staining on different, alternative, sections. This further complicates the visual analysis and integration of molecular and morphological information.

To overcome these problems, we have developed a three-dimensional microscopy system that integrates computer analysis and visualization tools (Fernández-González et al 2002). These tools automate or greatly reduce the amount of interaction required for the acquisition, reconstruction and morphologically directed analysis of thick tissue samples. Our system can be used to reconstruct tissue structures and to quantitatively measure the presence and spatial distribution of different molecular elements (e.g. genes, RNA's, proteins) in their intact cellular environment. This tool is currently being used to study breast cancer, where heterogeneity and three-dimensionality are at the very base of both disease initiation and clonal progression. In particular we are looking at the distribution of the Her2 amplification along in situ carcinomas to determine if there is a pattern indicative of the history and hopefully predictive of the progression of the disease.

We are also using our system to model mammary gland development in the context of the action of hormone receptors (Estrogen and Progesterone receptors) to compare it with the development of transgenic models (c-neu), which are known to have impaired gland development and long latency mammary tumors.

Finally, our interesting in understanding normal and neoplastic mammary gland development is leading us towards the use GFP transgenic animals to study the role of GFP labeled mammary epithelial cells upon transplantation into fat-cleared mammary glands of wild type animals.

All the above-described problems share in common a whole-istic approach that requires the use of three-dimensional microscopy and quantitative image analysis.