A computational approach to resolve cell level contributions to early glandular epithelial cancer progression

Abstract

Background: Three-dimensional (3D) embedded cell cultures provide an appropriate physiological environment to reconstruct features of early glandular epithelial cancer. Although these are orders of magnitude simpler than tissues, they too are complex systems that have proven challenging to understand. We used agent-based, discrete event simulation modeling methods to build working hypotheses of mechanisms of epithelial 3D culture phenotype and early cancer progression. Starting with an earlier software analogue, we validated an improved in silico epithelial analogue (ISEA) for cardinal features of a normally developed MDCK cyst. A set of axiomatic operating principles defined simulated cell actions. We explored selective disruption of individual simulated cell actions. New framework features enabled recording detailed measures of ISEA cell activities and morphology.

Results: Enabled by a small set of cell operating principles, ISEA cells multiplied and self-organized into cyst-like structures that mimicked those of MDCK cells in a 3D embedded cell culture. Selective disruption of "anoikis" or directional cell division caused the ISEA to develop phenotypic features resembling those of in vitro tumor reconstruction models and cancerous tissues in vivo. Disrupting either process, or both, altered cell activity patterns that resulted in morphologically similar outcomes. Increased disruption led to a prolonged presence of intraluminal cells.

Conclusions: ISEA mechanisms, behaviors, and morphological properties may have biological counterparts. To the extent that in silico-to-in vitro mappings are valid, the results suggest plausible, additional mechanisms of in vitro cancer reconstruction or reversion, and raise potentially significant implications for early cancer diagnosis based on histology. Further ISEA development and use are expected to provide a viable platform to complement in vitro methods for unraveling the mechanistic basis of epithelial morphogenesis and cancer progression.

Figure 1

The twelve ISEA axiomatic operating principles. Table 1 is a listing and explanation of the operating principles. The 2D space and all objects within are hexagonally discretized. Simulation time advances in steps corresponding to simulation cycles. During a simulation cycle, every CELL, in a pseudo-random order, decides what action to take based on its internal state (POLARIZED or UNPOLARIZED) and the composition of its adjacent neighborhood. A set of axioms determines what action is taken for each possible neighborhood configurations. Objects represented: POLARIZED CELL (red), UNPOLARIZED CELL (gray), MATRIX (white), and LUMINAL SPACE (black). At the top, selected decision-making CELLS at the start of simulation cycle n are numbered to indicate each of the twelve axiomatic preconditions being satisfied (they are listed in Table 1). For purpose of this illustration, the unnumbered CELLS are inactive. At the bottom, the system at the start of simulation cycle n + 1 shows the consequences of applying all twelve axioms.

Figure 2

MDCK and simulated cysts. (A) MDCK cells grown in 3D extracellular matrix form lumen-enclosing cystic structures surrounded by a layer of polarized cells. Cells composing cysts maintain three surfaces: apical (red), basal and lateral (green). Note the roundish contour typical of MDCK cysts. For growth and staining details, see [22]. Bar: ~10 μm. (B) Representative, stable ISEA CYSTS. CELLS in EMBEDDED condition produced stable, cystic structures enclosing LUMINAL SPACE; all CELLS were POLARIZED (red). CYSTS had convex shapes. (C) This illustration shows that convex polygonal CYSTS in discretized 2D hexagonal space map to a roundish structures in continuous 2D space. Because such a mapping provides no added scientific or mechanistic information, subsequent ISEA structures are shown as they appeared at simulation's end in 2D hexagonal space.

Figure 3

Dysregulation of Axiom 5 or 6 has a disruptive effect on ISEA CULTURE morphology in a severity-dependent manner. Axiom 5 dictates ANOIKIS (which maps to a form of cell death) when the CELL in its neighborhood has at least two CELLS and LUMINAL SPACE but no MATRIX. In simulations dysregulating Axiom 5, CELLS evaded ANOIKIS (by doing nothing) with a parameter-controlled probability, p, when Axiom 5's precondition was met. Axiom 6 dictates oriented CELL DIVISION when the CELL has at least one CELL and MATRIX but no FREE SPACE. When Axiom 6 was dysregulated, CELLS carried out disoriented CELL DIVISION: the CELL copy replaced a randomly selected MATRIX neighbor without regard for CELL neighbor number. Shown are CULTURE images and corresponding morphology index values after 50 simulation cycles of growth. One simulation cycle maps to 12 h in vitro. Each object is represented as a hexagon: POLARIZED CELL (red), UNPOLARIZED CELL (gray), MATRIX (white), and LUMINAL SPACE (black). One CELL width maps to 10 μm. (A) Axiom 5 dysregulation caused progressively disorganized CULTURE formations. (B) Axiom 6 dysregulation showed a similarly severity-dependent effect. The changes were less prominent but nevertheless clearly aberrant in both analogues.

Figure 4

Dysregulation of Axiom 5 (ANOIKIS) and its effect on ISEA CULTURE growth and morphology. Axiom 5 dictates CELL DEATH when a CELL has in its neighborhood at least two CELLS and LUMINAL SPACE but no MATRIX. With a parameter-controlled probability, p, CELLS evaded ANOIKIS (by doing nothing) when Axiom 5's precondition was met. Doing so caused distinct changes in growth and structural characteristics of the EMBEDDED CULTURE. (A) CELL CULTURE growth rate increased monotonically with the severity of dysregulation. CULTURE growths at six levels of dysregulation are shown. (B) Disrupting operation of Axiom 5 resulted in the formation of progressively aberrant MULTICELL structures, as indicated by the numeric scale. Higher values indicate a more disorganized morphology. Dysregulation had no observable effect on CULTURE morphology in the early stages (~5 simulation cycles) of growth. One simulation cycle maps to 12 h in vitro. The effect became progressively evident as simulation time advanced. (C-D) Axiom 5 dysregulation altered CELL DIVISION and DEATH event patterns. The changes became more evident at later times. In both simulations, the effect on CELL DEATH and DIVISION was monotonic, except when p = 0. The mean occurrence of CELL DIVISION and DEATH fell when p = 0 (vs p = 0.8). The data are mean values of 100 Monte Carlo runs.

Figure 5

Axiom 6 (oriented CELL DIVISION) dysregulation and its effect on ISEA CULTURE growth and morphology. Axiom 6 dictates CELL DIVISION when a CELL has at least one CELL and MATRIX but no FREE SPACE in its neighborhood. The CELL copy is placed at an adjacent MATRIX position that maximizes its number of CELL neighbors. With a parameter-controlled probability, p, CELLS followed an alternate, dysregulated action (disoriented CELL DIVISION) when the Axiom 6 precondition was met. The CELL copy replaced a randomly selected MATRIX neighbor without regard for CELL neighbor number. Doing so caused changes in growth and structural characteristics of the EMBEDDED CULTURE. (A) CELL CULTURE growth rate increased monotonically with the severity of dysregulation. (B) Shown are changes in growth morphology. Similar to Axiom 5 dysregulation, this analogue showed no observable effects during the early growth stage but obvious differences over time. (C-D) Axiom 6 dysregulation altered CELL DIVISION and DEATH event patterns. Near the maximally dysregulated state (p = 0), the system exhibited a proportionately larger increase in CELL DEATH events at later times. The data are mean values of 100 Monte Carlo runs.

Figure 6

Simultaneous dysregulation of Axioms 5 and 6 and its effect on ISEA CULTURE growth and morphology. Axioms 5 and 6 dictate ANOIKIS (a form of CELL DEATH) and oriented CELL DIVISION; both are essential to normal CYST growth in EMBEDDED CULTURE. With a parameter-controlled probability, p, for each of the two axioms, CELLS followed an alternate, dysregulated action. For Axiom 5, the alternate action was to evade ANOIKIS (i.e., do nothing). For Axiom 6, it was disoriented CELL DIVISION; the CELL copy replaces a randomly selected matrix neighbor. The lower case letters (a-i) in (A) and (B) correspond to the morphologies in (C). (A) ISEA CELL population, (B) morphology index values, and (C) simulation images after 50 simulation cycles of growth. One simulation cycle maps to 12 h in vitro. Each object is represented as a hexagon: POLARIZED CELL (red), UNPOLARIZED CELL (gray), MATRIX (white), and LUMINAL SPACE (black). One CELL width maps to 10 μm. The CELL count and morphology measurements are mean values of 100 Monte Carlo runs.

Figure 7

Microscopic images of in vitro epithelial cysts exhibiting characteristics of early-stage epithelial glandular tumor. (A-B) Loss of PALS1 expression results in the disruption of MDCK cell polarity and impaired development of cyst lumen [33]. PALS1 is involved in the establishment of cell polarity. Cysts composed of PALS1-ablated cells exhibited multiple small lumens (A) or developed a larger but incomplete lumen (arrow in B). (C-D) Overexpression of ErbB2 receptor leads to the formation of multi-acinar structures with filled lumens in 3D Matrigel [34]. (C) The structures consisted of multiple acinar-like units with filled lumens. The size range of at least 200 structures is shown. (D) Optical section of a single structure along the z-axis. (E-F) Inhibition of luminal apoptosis in proliferating structures results in lumen filling [3]. MCF-10A cells were infected with retroviruses encoding expression of proliferative oncoproteins--cyclin D1 or human papilloma virus (HPV) 16 E7--and anti-apoptotic Bcl family proteins (Bcl-2 or Bcl-X_L). (E) Acinar structures formed by cells expressing HPV 16 E7 and Bcl-2. (F) Structures formed by cells expressing cyclin D1 and Bcl-X_L. Panels A-B were reproduced with permission from [33]^© The American Society for Cell Biology. Panels C-D were reproduced with permission from [34]^© Nature Publishing Group. Panels E-F were reproduced with permission from [3]^© Nature Publishing Group.