Molecular predictors of 3D morphogenesis by breast cancer cell lines in 3D culture

Abstract

Correlative analysis of molecular markers with phenotypic signatures is the simplest model for hypothesis generation. In this paper, a panel of 24 breast cell lines was grown in 3D culture, their morphology was imaged through phase contrast microscopy, and computational methods were developed to segment and represent each colony at multiple dimensions. Subsequently, subpopulations from these morphological responses were identified through consensus clustering to reveal three clusters of round, grape-like, and stellate phenotypes. In some cases, cell lines with particular pathobiological phenotypes clustered together (e.g., ERBB2 amplified cell lines sharing the same morphometric properties as the grape-like phenotype). Next, associations with molecular features were realized through (i) differential analysis within each morphological cluster, and (ii) regression analysis across the entire panel of cell lines. In both cases, the dominant genes that are predictive of the morphological signatures were identified. Specifically, PPARgamma has been associated with the invasive stellate morphological phenotype, which corresponds to triple-negative pathobiology. PPARgamma has been validated through two supporting biological assays.

Figure 1. Computational pipeline for differential and association studies between colony morphologies and gene profiles for the panel of breast cancer cell lines cultured in 3D.

Figure 2. The consensus matrices for different numbers of clusters based on morphological representations are shown.

A darker block indicates higher morphological similarity between two cell lines. One can hypothesize that the larger block for has been partitioned into two blocks for ; however, the order is not preserved.

Figure 3. The CDF and its derivative, computed from the consensus matrix, is used to identify the number of clusters.

(A) CDF for each cluster, and (B) change in CDF as a function of cluster size, indicates that three is the optimum number of sub-populations.

Figure 4. Three cell lines displaying aggressive phenotypes are discovered with .

All other cell lines are grouped in a different subpopulation.

Figure 5. Three subpopulations of the 24 breast cancer cell lines grown in three-dimensional cell culture assay are revealed through consensus clustering.

Figure 6. Treatment of a MDA-MB-231 with a PPARG-inhibitor indicates reduction in the proliferation rate.

(A) untreated line, (B) treatment with Gw-9662, and (C) proliferation index. The proliferation index was determined by incubating cultures with cell proliferation analysis reagent, WST1, on Day 5.

Figure 7. PPARG is expressed in triple negative human breast cancer tissue.

(A–B) Localization of PPAR in normal and triple negative of human mammary tissue sections indicates that (i) in normal tissue, localization is apical and unbound to the nuclear regions, and (ii) in triple negative tissue, localization is nuclear-bound and heterogeneous. (C–D) Quantitative analysis on a cell-by-cell basis indicates that PPAR (i) is upregulated in triple negative patients, and (ii) has a heterogeneous distribution.

Figure 8. The elliptical contours indicate the half-peak magnitude iso-curves of the Gabor filters, in the frequency domain, at 6 orientations and 4 scales.

At each scale, mean filter response is invariant to rotation.

Figure 9. Regions associated with the multicellular colonies are differentiated through proposed computational method.

(a)(c) original images of two types of colonies with contrast reversal (e.g., dark regions in the bottom row versus bright regions in the top row), and (b)(d) the corresponding segmented results. Segmentation is feasible as a result of the Gabor filter bank that encodes oriented texture features.