Studies of epithelial cell cultures are providing knowledge about how individual cell activities are mediated by intrinsic and environmental factors to create the diverse phenotypes of normal epithelial morphogenesis and epithelial cancers. There is a need for additional methods to facilitate achieving a deeper, integrated understanding of the growing body of experimental observations. Past efforts have demonstrated how combined experimental and computational approaches contribute to that process [36
]. Our goal is to broaden and strengthen that effort by developing software analogues that are useful 1) as instantiated, working hypotheses of epithelial morphogenesis and tumorigenic phenotype in vitro, and 2) as an extensible, interactive resource of available biological knowledge about the mechanisms implicated in those processes. Progress described herein represents an early step towards achieving those goals.
We revised and extended the axiomatic operating principles of an earlier model [12
] to those shown in Fig. . The revised ISEA consistently produced roundish, convex CYSTS
with smooth margins, a cardinal feature of normal in vitro MDCK phenotype. We enabled mechanistic tracing during simulations of all processes essential for normal ISEA development. Two critical axioms were targeted for dysregulation: Axiom 5, which controlled ANOIKIS
, and Axiom 6 that dictated an abstract form of oriented cell division. The causal chains of events responsible for ISEA phenotype were explored in detail following dysregulation, a process which is infeasible using current state-of-the-art in vitro methods.
Dysregulated ISEA morphology exhibited features reminiscent of those associated with in vitro cancer reconstruction models and early cancer progression in vivo (see selected in vitro images in the Appendix). By increasing dysregulation of the two axioms, we altered ISEA morphology progressively to mimic features of epigenetic change that accompany early precursor lesions like atypical ductal hyperplasia [1
]. ISEAs using dysregulated ANOIKIS
(Axiom 5) developed MULTICELLULAR
structures having ill-formed LUMINAL SPACES
containing disorganized nests of CELLS
. With increased dysregulation, LUMINAL CELLS
sometimes broke out through the enclosing monolayer to PROLIFERATE
into the surrounding MATRIX
, as illustrated in Fig. . Although such behavior has not been observed in studies of apoptosis inhibition in 3D culture, the activation of certain growth factor receptors able to promote luminal space survival, such as ErbB2, do exhibit similar expansive phenotypes in 3D [3
]. If a mapping does exist between those ISEA behaviors and phenomena of epithelial systems, it suggests that epigenetic changes may be capable of inducing invasive behaviors in otherwise apparently normal cells in vitro or in vivo [1
]. The phenomena merits further in silico exploration.
Similar, but less dramatic changes were observed when we dysregulated oriented CELL DIVISION
(Axiom 6). Simultaneous dysregulation of the two axioms produced nonadditive effects but no new morphological features emerged: the structures were virtually indistinguishable from those obtained by dysregulating only Axiom 5 or 6. Consequently, without a priori dysregulation knowledge, one would be unable to reliably deduce the operational cause of a change in CULTURE
phenotype based solely on morphology images. A similar conclusion has been reached based on in vitro findings that phenotypic changes such as lumen filling in 3D cultures can be induced by deregulation of different molecular mechanisms [13
]. To the extent that the in silico-to-in vitro and in vitro-to-in vivo mappings are valid, the results support the idea that morphologically similar dysplasia can have different causes, and that may have implications for early diagnosis of cancer based on morphology alone, as very aggressive, early stage cancers may appear morphologically similar to potentially less aggressive, abnormal, non-cancerous growths.
Dysregulation of either axiom enabled some CELLS
to survive in the LUMINAL SPACE
. That ISEA behavior maps to in vitro observations [13
]. How the latter occurs has not been determined. How it occurs within ISEA may provide insight. A subset of INTRALUMINAL CELLS
contact by producing MATRIX
de novo (via Axiom 4 use). So doing enabled them and some other CELLS
to survive in aggregates inside the LUMINAL SPACE
, where they underwent cycles of PROLIFERATION
. Blocking the CELLS'
ability to produce matrix (Axiom 4) reduced INTRALUMINAL CELL
survival dramatically, and facilitated clearing of residual INTRALUMINAL CELLS
development (data not shown). In vitro, similar phenomena have been observed in MCF-10A epithelial cell cultures: cells accumulated inside cyst lumens when an anti-apoptotic protein, Bcl-2 was overexpressed [3
]. However, unlike in ISEA simulation, the cells eventually died and disappeared. Mechanisms underlying the latter process are unknown. Interestingly, some evidence suggests that Bcl-2 activates matrix metalloproteinase (MMP), which degrades ECM surrounding cells [39
]. Do the above INTRALUMINAL CELL
survival observations have an in vitro counterpart, or are these ISEA behaviors outside phenotype overlap? If there is an in vitro counterpart, then intraluminal epithelial cells in 3D embedded culture may evade apoptosis and further insure their survival by secreting matrix de novo for anchorage. In such a scenario, MMP activation could have an opposing effect by degrading the cell-secreted matrix, rendering the cells vulnerable to anchorage-dependent anoikis.
Dysregulation of Axiom 6 demonstrated the importance of proper DIVISION
direction during CULTURE
growth. Evidence supports a mapping to in vitro counterparts. Similar structures form when cell polarity is disrupted in MDCK cell cultures by ablating the mammalian ortholog of PALS1, a gene involved in epithelial polarity in Drosophila [33
]. Similar to ISEA behaviors, the structures contain multiple intraluminal cell clusters and resemble certain patterns observed in breast ductal carcinomas in situ and prostate hyperplasia [35
]. Because cell polarity is critical to cell division orientation, one could speculate that a disruption in oriented cell division by PALS1 ablation may have contributed to the observed phenomenon. We also note that several groups have discovered that Ric-8 protein plays a key part in the positioning of the division axis in Drosophila morphogenesis [40
]. It is not yet known if Ric-8 plays a similar role in oriented mammalian cell division in cultures. Nevertheless, ISEA behaviors indicate that compromising one or more of the mechanisms managing oriented cell division can contribute to features that mimic early stage, cancer-like structures in 3D cultures.
The ISEA methods used to mimic attributes of cancer reconstruction can be compared to those used to model tumor growth in vitro and in vivo. Recent models [42
] have represented cancerous cells as permanently transformed cell line. We explored incremental dysregulation of specific ISEA mechanisms. Galle et al. [47
] used a similar, creative, individual cell-based approach to simulate and study epithelial cell monolayer growth. They used selective "knockouts" of cell level growth regulation and control mechanisms to investigate how those different mechanisms collectively acted together to influence population morphology. More recently, Rejniak and Anderson [48
] introduced single cell-based, immersed boundary simulation models of epithelial acini development in vitro, and applied the models to investigate different conditions of growth that contribute to normal and abnormal acinar development. Other studies have used single cell-based cellular Potts models and extensions to simulate various aspects of development including embryonic cell patterning and tumor invasion [16
axioms are high level, low-resolution placeholders for more detailed representations of the actual complex mechanisms driving epithelial cell behavior. Use of axioms precludes explicit representations of the abundant, detailed subcellular information that is available. However, starting with the current more abstract set of axioms provided the simplest method and approach for building a useful, working model, positing principles of operation, and testing hypotheses as discussed above. On the other hand, a key advantage of the approach built into ISEA and its framework are their adaptability for inclusion of additional attributes and details through an iterative model refinement process [8
]. The current analogue and its components, including CELL
axioms, can be further developed to reflect new biological information (e.g., cell positioning mechanisms). We can elaborate ISEA to include higher granularity components and mechanisms that map to subcellular details such as cell lifecycle pathways and intercellular signaling networks when validation against an expanded set of targeted attributes requires doing so. From an engineering perspective, doing so is straightforward and can be accomplished by swapping the current component (e.g., CELL
) for a more detailed composite agent as described in additional file 1
: Supplementary Material. Replacement could also occur at the intra-component level, for example by replacing CELL
axioms with more detailed logic based on interacting components. A challenging task will be to insure cross-model validation between the different analogue variants, and to develop appropriate automated validation measures.