(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .



Evolution of phenotypic plasticity leads to tumor heterogeneity with implications for therapy [1]

['Simon Syga', 'Center For Interdisciplinary Digital Sciences', 'Department Information Services', 'High Performance Computing', 'Tud Dresden University Of Technology', 'Dresden', 'Harish P. Jain', 'Njord Centre', 'Department Of Physics', 'University Of Oslo']

Date: 2024-10

Abstract Cancer is a significant global health issue, with treatment challenges arising from intratumor heterogeneity. This heterogeneity stems mainly from somatic evolution, causing genetic diversity within the tumor, and phenotypic plasticity of tumor cells leading to reversible phenotypic changes. However, the interplay of both factors has not been rigorously investigated. Here, we examine the complex relationship between somatic evolution and phenotypic plasticity, explicitly focusing on the interplay between cell migration and proliferation. This type of phenotypic plasticity is essential in glioblastoma, the most aggressive form of brain tumor. We propose that somatic evolution alters the regulation of phenotypic plasticity in tumor cells, specifically the reaction to changes in the microenvironment. We study this hypothesis using a novel, spatially explicit model that tracks individual cells’ phenotypic and genetic states. We assume cells change between migratory and proliferative states controlled by inherited and mutation-driven genotypes and the cells’ microenvironment. We observe that cells at the tumor edge evolve to favor migration over proliferation and vice versa in the tumor bulk. Notably, different genetic configurations can result in this pattern of phenotypic heterogeneity. We analytically predict the outcome of the evolutionary process, showing that it depends on the tumor microenvironment. Synthetic tumors display varying levels of genetic and phenotypic heterogeneity, which we show are predictors of tumor recurrence time after treatment. Interestingly, higher phenotypic heterogeneity predicts poor treatment outcomes, unlike genetic heterogeneity. Our research offers a novel explanation for heterogeneous patterns of tumor recurrence in glioblastoma patients.

Author summary Intratumor heterogeneity presents a significant barrier to effective cancer therapy. This heterogeneity stems from the evolution of cancer cells and their capability for phenotypic plasticity. However, the interplay between these two factors still needs to be fully understood. This study examines the interaction between cancer cell evolution and phenotypic plasticity, focusing on the phenotypic switch between migration and proliferation. Such plasticity is particularly relevant to glioblastoma, the most aggressive form of brain tumor. By employing a novel model, we explore how tumor cell evolution, influenced by both genotype and microenvironment, contributes to tumor heterogeneity. We observe that cells at the tumor periphery tend to migrate, while those within the tumor are more inclined to proliferate. Interestingly, our analysis reveals that distinct genetic configurations of the tumor can lead to this observed pattern. Further, we delve into the implications for cancer treatment and discover that it is phenotypic, rather than genetic, heterogeneity that more accurately predicts tumor recurrence following therapy. Our findings offer insights into the significant variability observed in glioblastoma recurrence times post-treatment.

Citation: Syga S, Jain HP, Krellner M, Hatzikirou H, Deutsch A (2024) Evolution of phenotypic plasticity leads to tumor heterogeneity with implications for therapy. PLoS Comput Biol 20(8): e1012003. https://doi.org/10.1371/journal.pcbi.1012003 Editor: Jean Clairambault, INRIA Rocquencourt, FRANCE Received: March 15, 2024; Accepted: July 23, 2024; Published: August 9, 2024 Copyright: © 2024 Syga et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability: The Python code used to generate the simulations for this study was published as a Zenodo snapshot here: doi.org/10.5281/zenodo.10806326. Funding: SS and AD are funded by the European Union (ERC, subLethal, 101054921, https://erc.europa.eu/). Views and opinions expressed are however those of the authors only and do not necessarily reflect those of the European Union or the European Research Council Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. SS and AD acknowledge support by Worldwide Cancer Research (23-0177, https://www.worldwidecancerresearch.org). HPJ acknowledges support by the European Union’s Horizon 2020 research and innovation programme (CompSci TraCS, 945371, https://cordis.europa.eu). HH thanks Volkswagenstiftung for its support in the "Life?" program (96732, https://www.volkswagenstiftung.de). He has received funding from the Bundesministerium für Bildung und Forschung under grant agreement No. 031L0237C (MiEDGE project/ERACOSYSMED, https://www.bmbf.de). He also acknowledges the support of the RIG-2023-051 grant from Khalifa University (https://www.ku.ac.ae) and the AJF-NIH-25-KU grant from the NIH-UAE collaborative call 2023 (https://www.aljalilafoundation.ae). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors have declared that no competing interests exist.

Introduction Cancer remains a significant challenge in public health despite improvements in clinical treatment. One of the leading causes of treatment failure is the emergence of therapy resistance, enabled by tumor heterogeneity [1]. Tumor heterogeneity can refer to the existence of genetically distinct tumor subclones or the existence of cancer cells with different phenotypic characteristics, including gene expression, metabolism, motility, proliferation, metastatic potential, and resistance to treatment [2, 3]. It results from genome instability and mutations, and non-genetic mechanisms of cancer progression and adaptation [4, 5]. Due to its heterogeneous nature, tumor growth is increasingly viewed as an evolutionary and ecological process in which abnormal cells compete for space and resources with each other and with healthy cells in the surrounding tissue [6, 7]. Lately, the ability of cancer cells to develop phenotypic plasticity has been proposed as a new hallmark of cancer [8]. Important examples include the epithelial to mesenchymal transition [9] and the change of metabolism from oxidative phosphorylation to anaerobic glycolytic metabolism (Warburg effect) [10]. Understanding phenotypic states and transitions, mechanisms driving plasticity, and how to measure and model these behaviors are significant challenges in the study of cancer plasticity [11]. One particularly interesting example of phenotypic plasticity is the migration-proliferation dichotomy, which has been observed for non-neoplastic cells [12] as well as during tumor development [13, 14]. This phenotypic switch is especially relevant in the context of glioblastoma, the most lethal form of brain cancer, with important implications for treatment [15]. The precise molecular mechanisms underlying this dichotomy remain poorly understood. The switch between migrating and proliferating phenotypes has been suggested to depend on the microenvironment of cells, such as growth factor gradients, extracellular matrix properties, or altered nutrient availability [9]. Because tumor cells secrete relevant factors, produce toxic metabolites, and consume oxygen and nutrients, their local cell density correlates with those chemicals’ concentration [16, 17]. Therefore, the local cell density can be considered a proxy for analyzing the dependence of the switch on the tumor microenvironment. Several mathematical models have shown that the migration-proliferation plasticity has a significant impact on tumor spread [18–22]. Böttger et al. [23] investigated the effect of a density-dependent switch between migrating and proliferating phenotypes on tumor persistence. They assumed that migration and proliferation are mutually exclusive and that cells switch between phenotypes according to a switching probability that depends on local cell density. This approach led to the notion of attractive and repulsive go-or-grow strategies, meaning cells that tend to proliferate more or less with increasing cell density, respectively. They also showed that the attractive strategy could lead to a strong Allee effect, a situation in which the effective growth rate of a population can become negative for low cell densities [24]. Gallaher et al. [25] investigated the phenotypic evolution of cancer cells under proliferation-migration trade-offs. The authors assumed that the cellular phenotype is characterized by migration speed and proliferation rate and studied the effect of different trade-off conditions on the evolutionary dynamics. They showed that a higher migration rate is generally favored over proliferation at the tumor’s edge and vice-versa in the tumor bulk. In a similar setting, the spatiotemporal phenotype distribution could be predicted by mean-field theory [26]. However, previous studies investigating the evolutionary dynamics of migration and proliferation traits did not explicitly consider the cell decision-making process governing the phenotypic switch between migration and proliferation. In these studies, all cells had an identical phenotypic switch regulation in response to microenvironmental changes, and the phenotypic selection was based solely on volume exclusion effects. This strongly simplified the biological complexity of cell decision-making. Cancer cells are phenotypically plastic and can adapt in response to their microenvironment according to their genotype, see Fig 1A. However, it is unknown how this phenotypic plasticity influences the evolutionary dynamics in general and in the context of the migration/proliferation switch in particular. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 1. Hierarchical representation of cell decision-making. (A) The cell’s genotype dictates the regulation of the phenotypic switch. This switch determines the cell’s reaction to its microenvironment. Subsequently, the cell interacts and shapes its microenvironment according to its phenotype. (B) In the mathematical model and the context of the go-or-grow dichotomy, the genotype is represented by the parameter κ which controls the phenotypic switch between the migratory and proliferating phenotypes dependent on the local cell density. After assuming either phenotype the cell influences the local microenvironment by either reducing the cell density (by migration) or increasing it (by proliferation). We neglect possible epigenetic changes by a persistent microenvironment on long time scales. https://doi.org/10.1371/journal.pcbi.1012003.g001 It remains to be determined whether there is an optimal regulation of the phenotypic switch between migration and proliferation and how the microenvironment might influence this. It is also uncertain why cells adopt an attractive go-or-grow strategy, particularly as this increases the risk of extinction due to the strong Allee effect [23]. Furthermore, the effects of go-or-grow plasticity on therapy outcomes have yet to be fully understood. To tackle these questions, we propose a novel cellular automaton model for the emergence and subsequent natural selection of cells with different go-or-grow strategies. Cells switch between migratory and proliferative phenotypes as dictated by a phenotypic switch function that depends on their individual (fixed) genotype and the (variable) local microenvironment in the form of the local cell density relative to a cell density threshold (switch threshold), see Fig 1B. Consequently, the probability of either phenotype is not given by its genotype alone but by the combination of its genotype and microenvironment. The cells’ genotypes can be inherited or acquired via mutations. We study the cellular automaton model by extensive computer simulations and a mathematical analysis based on a mean-field approximation. We perform a parameter scan varying the death rate and the cell density threshold while recording the evolutionary dynamics. We quantify the phenotypic and genetic heterogeneity that arises from these dynamics. Furthermore, our study extends to the implications of this heterogeneity for cancer therapy. We simulate cancer treatments and monitor the time until recurrence, which we correlate with the observed phenotypic and genetic tumor heterogeneity. Lastly, we discuss how medical evidence from glioblastoma patients supports the predictions made by our model.

Methods Model definition We define a stochastic, spatiotemporal, cell-based model to study the co-evolution of different go-or-grow strategies. This allows us to describe the cell decision-making of individual cells that each possess a unique go-or-grow strategy, which we associate with its genotype, see Fig 1B. We assume that a cell’s decision-making depends on the local microenvironment. We use a cellular automaton model with a state space accounting for cell velocity (represented by channels). This class of cellular automata is called lattice-gas cellular automaton (LGCA) and originates in fluid dynamics simulations [27]. It was later extended to model biological phenomena such as excitable media, collective migration, and tumor growth [28–31]. In the LGCA, individual cells reside in channels on nodes of a regular lattice . Every node has b nearest-neighbor nodes, where b depends on the lattice geometry, e. g., for a one-dimensional lattice b = 2. Nodes are connected to their nearest neighbors by unit vectors c i , i = 0, …, b − 1, which we call velocity channels. We have c 0 = 1, c 1 = −1 on a one-dimensional lattice. Each node has one rest channel (channels with zero velocity), c b = 0. The state of each node is updated independently and simultaneously in discrete time steps k = 1, 2, 3, …. First, each state is updated in the interaction step using a stochastic update rule , incorporating all relevant biological mechanisms. Subsequently, cells residing in velocity channels are deterministically transported to nearest-neighbor nodes in the direction of the respective velocity channel, i. e., (1) The rules of our model are designed to implement the following assumptions about cellular behavior, see Fig 2. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 2. The lattice-gas cellular automaton (LGCA) model. (A) Left: The LGCA model is implemented on a one-dimensional lattice, where each node comprises two velocity channels for movement to the nearest-neighbor nodes and one rest channel for no movement. Each channel can be occupied by any number of cells. Cells in velocity channels (marked red) have the migratory phenotype, and cells in rest channels have the proliferative phenotype (blue). White channels denote the absence of tumor cells. Right: Schematic illustration of model dynamics: Cells switch between phenotypes depending on their genotype κ and the local tumor cell density. Proliferative cells divide with a constant rate α. Migratory cells randomly choose a new direction of movement. All cells die with a constant rate δ. (B) Cell migration. Each migratory cell (red) moves in the direction of its respective velocity channel. (C and D) Phenotypic plasticity. The phenotypic switch probability r κ depends on the tumor cell density in the microenvironment and the cell’s genotype κ. The sign of κ determines the switch regulation. (C) κ > 0 results in attractive behavior (proliferating phenotype triggered by high cell density), and κ < 0 (D) leads to repulsive behavior (cells switch to migratory phenotype if the local cell density becomes too high). The parameter θ indicates the cell density threshold, where the probability for either phenotype is 1/2. https://doi.org/10.1371/journal.pcbi.1012003.g002 Death. All cells die with a constant probability δ in each time step.

All cells die with a constant probability δ in each time step. Phenotypes. Cells in the proliferating phenotype do not migrate, while cells with the migratory phenotype do not proliferate. Cells with the proliferative phenotype divide with fixed probability α(1 − n( r )/K), where n( r ) is the number of cells on the node and K > 0 is the carrying capacity, a free parameter.

Cells in the proliferating phenotype do not migrate, while cells with the migratory phenotype do not proliferate. Cells with the proliferative phenotype divide with fixed probability α(1 − n( )/K), where n( ) is the number of cells on the node and K > 0 is the carrying capacity, a free parameter. Phenotypic switch. Cells switch between a proliferating and a migratory phenotype depending on the local microenvironment and their unique phenotypic switch regulation

Cells switch between a proliferating and a migratory phenotype depending on the local microenvironment and their unique phenotypic switch regulation Mutations. During proliferation, the regulation of the phenotypic switch of the daughter cell d, κ d , representing the cell’s genotype, is drawn from a Gaussian distribution centered on the respective parameter of the mother cell κ m and has a fixed standard deviation Δκ. This ensures that daughter cells inherit κ values similar to those of the mother cells.

During proliferation, the regulation of the phenotypic switch of the daughter cell d, κ , representing the cell’s genotype, is drawn from a Gaussian distribution centered on the respective parameter of the mother cell κ and has a fixed standard deviation Δκ. This ensures that daughter cells inherit κ values similar to those of the mother cells. Migration. Migratory cells perform an unbiased random walk. Due to the variable proliferation behavior caused by the phenotypic switch between migration and proliferation, a constant cell death rate, and a steady influx of variation caused by the mutations, cells are subjected to Darwinian evolution, potentially leading to a complex spatial distribution of κ values. This is the key novelty in this paper compared to [23], where all cells during one simulation followed the same phenotypic switch function r κ (ρ), i. e., had the same κ value. In the model, we divide the stochastic update rule into several steps: First, a death operator removes dying cells. Next, all cells switch their phenotype according to the phenotypic switch function, which depends on their regulation of the switch κ and the average density in their microenvironment . Then, proliferative cells divide according to the probability mentioned above. Newly born cells are also placed in the rest channel, i. e., start out in the proliferative state. The switch parameter of the daughter cells is sampled from a Gaussian distribution centered on the switch parameters of the mother cells κ m , i.e., . For a detailed definition of the mathematical model including a table of parameters, see S1 Text. We propose that the transition between proliferative and migratory cellular behavior relies on the local cell density. We regard the cell density as a proxy for tumor cell interactions with extracellular matrix components, chemical signals, and stromal cells. These effects are modeled based on their correlation with cell density. Instead of reproducing the phenotypic switch process’ intricacies involving signaling networks, we simplify the intracellular details into stochastic cell-based rules, yielding an analytically tractable model. This increases our understanding of the underlying dynamics. Although the precise dependence of the phenotypic switch on cell density is uncertain, we opt for the simplest form: a monotonous dependence. This choice enables the discrimination of two complementary types of plasticity: attraction to or repulsion from densely populated regions. In the attraction scenario, cell motility decreases with local cell density, promoting proliferation in crowded areas, as shown in Fig 2C. Conversely, in the repulsion scenario, cells avoid densely populated areas by switching to the migratory phenotype and moving away while increasing proliferation in sparsely populated regions, as depicted in Fig 2D. An independent strategy is also possible where the phenotypic switch does not depend on cell density, i. e., κ ≈ 0. Corresponding cells adopt either phenotype with equal probability irrespective of their microenvironment. We assume that transitions between migrating and proliferating phenotypes, i. e., velocity and rest channels, are dictated by the phenotypic switch function (2) where represents the average density of tumor cells in the neighborhood, κ is the unique regulation parameter of the switch and θ is a constant threshold parameter. The intensity of the dependence of the switch on the cell density is determined by the absolute value of κ. The sign of κ indicates whether the cell applies the attractive (κ > 0) or the repulsive strategy (κ < 0) [23]. The cell density threshold at which the switching probabilities between phenotypes are equal is denoted by the parameter θ. The functional form of the switching function was first suggested in a study that examined tumor invasion [20]. Analysis of growth dynamics We simulate the LGCA model on a one-dimensional lattice with L = 1001 nodes, corresponding to a maximum tumor diameter of 5 cm (see S1 Text), which is large enough that no cells reach the lattice boundary. We place K cells in the central node’s rest channel, i. e., these cells are in the proliferative phenotype, as an initial condition and choose their switch parameters κ l from a uniform distribution in the interval [−4, 4]. We investigate the impact of the death probability δ and the phenotypic switch threshold θ on evolutionary dynamics. To this end, we perform a parameter scan, systematically varying these parameters in the intervals 0 ≤ δ ≤ 0.25 and 0 ≤ θ ≤ 1. We fix the proliferation probability and capacity at α = 1, K = 100 and the standard deviation Δκ = 0.2. For an overview of mathematical symbols and parameter values see Tables A and B in S1 Text. We simulate the system for 1000 time steps, corresponding to about four years (see S1 Text), avoiding synthetic tumors reaching the lattice boundary at the end of the simulation. As observable, we record the spatial distributions of resting/migrating cells and the genotype distribution ψ(r, κ). We also perform simulations on two-dimensional hexagonal lattices to ensure that our results do not depend on the one-dimensional lattice. We choose a hexagonal lattice over a square lattice to reduce artificial patterns that can occur in spatially discrete models [32]. To this end, we place K = 50 cells in the middle of a hexagonal lattice of sidelength L = 250 and monitor the temporal dynamics for 300 time steps. Other parameters and observables are unchanged compared to the one-dimensional simulations. Due to the computational costs involved, we do not perform a systematic parameter scan but simulate the system for selected parameter sets to ensure that the observed dynamics in two dimensions match those in one dimension qualitatively. Characterization of therapy response Starting from simulation endpoints of the growth dynamics simulations, we remove 99.9% of cells to simulate cancer therapy. This value was previously used to estimate the effect of the resection of a glioblastoma tumor [19]. We then let the tumor grow again until the number of cells has reached the level before treatment and record the required time as recurrence time. Quantification of heterogeneity To quantify the heterogeneity in our synthetic tumors, we calculate the Shannon entropy of the observed distributions of phenotypes and genotypes. This information-theoretic score is maximal if every tumor cell subpopulation is equally probable and minimal if all cells are identical [33]. For the phenotype distribution, we can directly compute the entropy as (3) where p migr , p prolif are the frequencies of migrating and proliferating cells, respectively. As the distribution of genotypes is continuous in our model, we apply a histogram-based estimator. We first bin our genotype values κ i using the method histogram of the Python package numpy 1.24.2 with automatic bin estimation. Next, we can estimate the genotype entropy as (4) where n i is the number of cells in bin i and Δ i is the width of bin i.

Discussion Describing, reconstructing, or even predicting the evolutionary dynamics in tumors using mathematical models can potentially improve cancer therapy. However, this remains a formidable challenge due to the multiscale interplay of genetic, phenotypic, and microenvironmental heterogeneity. This study investigated the interplay of evolutionary dynamics and the phenotypic plasticity between migration and proliferation, which is especially relevant for glioblastoma cells. Specifically, we asked if an optimal regulation of the phenotypic switch exists and how it depends on environmental parameters. Moreover, we studied the clinical implications of this phenotypic plasticity. To this end, we applied a novel cellular automaton model for cells that switch between a migratory and proliferating phenotype depending on their microenvironment. Here, we used the local cell density as a proxy for other environmental factors. Cell density is compared to the switch threshold θ. We enabled each cell to employ its unique go-or-grow strategy determined by the cell-specific phenotypic switch parameter κ, which we envision as a simplified representation of a cell’s genotype. By doing so, we could distinguish three qualitatively different strategies: The independent strategy (κ ≈ 0), where cells switch between the two phenotypes independent from the local microenvironment and both phenotypes are always equally likely; the attractive strategy (κ > 0), where cells are increasingly likely to be in the proliferative phenotype with increasing cell density; and finally, the repulsive strategy (κ < 0), with which cells are more and more likely to be in the migratory phenotype with increasing cell density. Both phenotypes have equal probability at the cell density threshold θ, irrespective of κ. The switch parameter θ is fixed and assumed to be a function of the tumor microenvironment. Assuming that the phenotypic switch is triggered by low resource availability, we proposed that a large switch threshold κ corresponds to a well-perfused tumor in which nutrients and oxygen are readily available. Conversely, a low switch threshold indicates a low resource availability. We then investigated the evolutionary dynamics in a growing tumor for varying death probability δ and switch threshold θ and observed the emergence of phenotypic and genetic heterogeneity. Notably, we found that for a low cell density threshold (resource scarcity), the attractive strategy is favored in the tumor bulk, while the repulsive strategy dominates at the tumor border. This situation is reversed in the case of a high switch threshold, which we could predict by a mean-field analysis. In both situations, cells are preferably in the proliferating phenotype in the tumor bulk and the migratory phenotype at the tumor border. However, the tumor’s genetic makeup is fundamentally different in both cases and depends on the microenvironment, i. e., the apoptosis rate δ and the switch threshold θ. We also investigated a treatment scenario and recorded the response of synthetic tumors, specifically the recurrence time. We found that the recurrence time varies significantly, like for many cancers in medical practice [35, 41, 42]. Interestingly, high phenotypic heterogeneity was associated with a poor clinical outcome, while moderate genetic heterogeneity is correlated with better therapy success. The evolution of phenotypic heterogeneity with more migratory cells at the tumor front and cells with higher proliferation rate in the tumor bulk had been observed in several previous modeling studies [20, 25, 26, 43]. However, so far, theoretical studies typically either study the effects of phenotypic plasticity or the evolution of genetic traits. They do not consider the cellular phenotype as the consequence of a cell decision-making process involving both the genotype and the cellular microenvironment. Here, we filled this gap by implicitly modeling the cell decision-making process in the context of the so-called go-or-grow dichotomy using a phenotypic switch function depending on cell density. This implies that we do not distinguish cancer cell clones that are inherently more migratory or have a higher proliferation rate. Instead, each cell switches between these two phenotypes as a response to its microenvironment and according to its genotype. Experimentally identifying the phenotypic switch function would require studying the respective tumor cell lines in varying microenvironmental contexts. While we did not focus on specific tumor cell lines here, it has been hypothesized before that the attractive strategy (κ > 0) corresponds to a low-grade tumor, and the repulsive strategy (κ < 0) to a high-grade tumor [23]. This aligns with clinical evidence on the effect of blood perfusion and postoperative infarction in glioblastoma tumors. Our model also predicted that increased phenotypic heterogeneity correlates with a worse prognosis, in contrast to high genetic heterogeneity that can be associated with longer recurrence times. This agrees with results by Sharma et al. [44], who found that in non-small-cell lung cancer, genetic heterogeneity alone was insufficient to capture the heterogeneity observed at the transcriptomic level. They showed that phenotypic heterogeneity results from various intra- and extracellular sources. For example, the proliferation rates of cells of the same subclone varied depending on the tumor microenvironment. They concluded that tumor heterogeneity should be assessed at multiple levels to find the sources of phenotypic heterogeneity affecting clinical outcomes. In our proposed model, we integrate the impact of the microenvironment on the phenotypic transition between motile and proliferative cellular behaviors through a dependency on local cell density. This assumption is reasonable, as other potential environmental factors, such as nutrient and oxygen availability, molecular signaling gradients, or additional cell-cell interactions, are also mediated by and correlated with local cell density. Simplifying the inherent complexity allows a more tractable exploration of fundamental organizational principles. We anticipate that significant population characteristics, such as the emergence of phenotypic and genetic spatial heterogeneity, are still observable even if the reliance of plasticity on local density is refined into a more accurate dependency on specific cell-microenvironment interactions. Future research can build on this foundation by exploring additional aspects of tumor evolution and therapy response. For instance, incorporating explicit dependencies on specific cell-microenvironment interactions and additional environmental factors, such as diffusing nutrients and oxygen, could provide a more comprehensive understanding of tumor growth’s complex dynamics. This would lead to a more complex, nonlinear, non-monotonic, phenotypic switch function with multiple inputs, potentially resulting in even richer dynamics resulting from the dynamic and heterogeneous tumor microenvironment. In conclusion, we showed that the interplay of phenotypic plasticity and Darwinian evolution leads to multi-scale heterogeneity, i. e., heterogeneous spatial distributions of cancer genotypes and phenotypes, affecting cancer treatment outcomes. The therapy response of synthetic tumors showed considerable variability and depended on biological parameters in a complex, non-linear manner, reminiscent of clinical practice. Associating the recurrence time with genetic and phenotypic heterogeneity revealed that phenotypic heterogeneity is the more important prognostic marker associated with worse treatment outcomes.

[END]
---
[1] Url: https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1012003

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.