A computational study of VEGF production by patterned retinal epithelial cell colonies as a model for neovascular macular degeneration
- 486 Downloads
The configuration of necrotic areas within the retinal pigmented epithelium is an important element in the progression of age-related macular degeneration (AMD). In the exudative (wet) and non-exudative (dry) forms of the disease, retinal pigment epithelial (RPE) cells respond to adjacent atrophied regions by secreting vascular endothelial growth factor (VEGF) that in turn recruits new blood vessels which lead to a further reduction in retinal function and vision. In vitro models exist for studying VEGF expression in wet AMD (Vargis et al., Biomaterials 35(13):3999–4004, 2014), but are limited in the patterns of necrotic and intact RPE epithelium they can produce and in their ability to finely resolve VEGF expression dynamics.
In this work, an in silico hybrid agent-based model was developed and validated using the results of this cell culture model of VEGF expression in AMD. The computational model was used to extend the cell culture investigation to explore the dynamics of VEGF expression in different sized patches of RPE cells and the role of negative feedback in VEGF expression. Results of the simulation and the cell culture studies were in excellent qualitative agreement, and close quantitative agreement.
The model indicated that the configuration of necrotic and RPE cell-containing regions have a major impact on VEGF expression dynamics and made precise predictions of VEGF expression dynamics by groups of RPE cells of various sizes and configurations. Coupled with biological studies, this model may give insights into key molecular mechanisms of AMD progression and open routes to more effective treatments.
KeywordsMicropatterning Auto-regulation Vascular endothelial growth factor Age-related macular degeneration Retinal pigment epithelial cells Age-related macular degeneration
Age-related macular degeneration
Partial differential equations
Retinal pigment epithelial cells
Vascular endothelial growth factor
Age-related macular degeneration (AMD) is a leading cause of irreversible blindness, particularly among adults over the age of 50 [2, 3, 4]. In AMD, degeneration of retinal pigment epithelial (RPE) cells, a type of neural cell that provides metabolic support to photoreceptor cells, severely damages vision. There are two forms of AMD, one involving acellular debris (dry AMD), and the other involving neovascularization of the retina from the underlying choriocapillaris (wet AMD) .
Cell culture models that control the spatial organization and growth of RPE cells can provide valuable tools for understanding cell behavior in AMD and its interaction with vascular endothelial growth factor (VEGF). VEGF is the primary signaling molecule that stimulates angiogenesis  and is an important biomarker of AMD . In the retina, VEGF is secreted in the RPE and is the primary driver of retinal vasculature development [8, 9]. Monitoring the expression of VEGF within controlled environments of model systems can lead to new insights that improve our understanding of the initiation and progression of AMD.
Developing an in silico framework for the cell culture micropatterning AMD model provides a beneficial system for evaluating the spatiotemporal effects of VEGF transport and expression within these controlled environments and in replicating the pathology of AMD to gain new insights on disease progression and outcomes. In silico models can also be used to study internal and external regulatory mechanisms influenced by feedback from the evolving cellular environment. Developing these predictive models is essential to identify biological pathways that may be targeted by new pharmaceutical agents.
The goal of this study was to develop an in silico model to replicate and extend the cell microprinting model for AMD reported in . The in silico model employs a two dimensional representation of the cellular culture because in the microprinting model, a monolayer of RPE cells form on the printed disks. While a two dimensional model is sufficient to replicate this bioengineered study, more realistic models that incorporate photoreceptors, and bipolar, amacrine, ganglion cells would require three dimensions.
Using this computational model, we studied the growth of RPE cells in discrete patches of different sizes and configurations to learn how cell arrangements can effect VEGF expression. The level of VEGF in each group of cells was studied as a function of cell number and patch area over time. To explore the hypothesis that VEGF expression is linked to global VEGF concentration, VEGF expression from various sized patches was quantified following VEGF administration. This study complements experiments using cell culture and provides a framework that can be used to investigate the influence of cell patterning on the secretion of VEGF by the RPE and opens a path towards mimicking the effects of tissue damage. This model extended the study of Vargis et al  and made predictions about VEGF regulation and expression in cell configurations that could not have been produced experimentally. The in silico model has the potential to examine the effects of anti-VEGF agents that may aid in the optimization of anti-angiogenic therapeutics and to be extended to other disorders that involve localized cell death within an epithelium.
Hybrid agent-based model framework
The agent-based modeling framework known as iDynoMiCs  was extended to simulate the effect of RPE cell distribution on VEGF expression. This modeling framework consists of discrete and continuous elements, making this a hybrid model. The discrete elements are particles each representing an individual cell. Particles mechanically interact with one another and secrete, consume or react to soluble molecules. They are positioned in space and occupy the volume of a single cell. The continuous elements of the model are a collection of soluble molecules (referred to as solutes) that could include nutrients, oxygen, and signaling molecules such as VEGF. A set of partial differential equations (PDEs) defines the interactions of molecules with cells and each other as they diffuse and participate in a variety of reactions.
Model’s parameter descriptions
As the simulation proceeds, particles move because of growth and division, and solute distributions change due to reactions and diffusion. In a single simulation step, first the biomechanical forces arising from growth are relaxed by moving particles to avoid overlap , then a PDE solver is applied to resolve all the local changes in solute concentrations. The solute fields are kept in steady-state with respect to the particle movement because reactions and diffusion occur much more rapidly (on the order of seconds or minutes) than changes in cell positions (on the order of hours to days).
Cyclic boundary conditions are used for the sides of the 2D simulation domains illustrated in Fig. 1. This boundary type allows the representation of larger domains by assuming that the computation domain is replicated indefinitely. Solute concentrations are kept constant across cyclic boundaries, and any particle crossing one of the cyclic boundaries is instantly moved to the connected boundary.
Modeling different RPE configurations
The distributions of cells studied by Vargis et al. were replicated as shown in Fig. 1. Patch diameters were 100, 200, 300 and 400 μ m. In these configurations, the total area of the domain was 5.76 mm2 and the area occupied by patches that could support the growth of cells was constant across all simulations and equal to 1.13 mm2 or 20% of the simulation domain.
Initial number of cells in each patch. These values match those used in the Vargis et al. study 
PatchSize (μ m)
Initial number of cells.
The VEGF distribution is determined by diffusion through the medium, secretion as a function of particle mass M and local VEGF concentration V as given by Eq. 1. The diffusion coefficient of VEGF (D V ) is set to 5.8×10−11 m 2/s, from experiments described in . The first term of Eq. 1 accounts for VEGF concentration changes due to diffusion, and the second term accounts for the auto-regulation of VEGF through negative feedback.
Table 1 summarizes the parameters and their values used in the equations above.
Exploration of autoregulation of VEGF expression
We hypothesized that there would be higher levels of VEGF per cell expressed in small patches because these cells experienced lower initial levels of VEGF than did the cells in larger patches. Although cells were plated at the same cell density in all patch configurations, cells in the smallest patches have a greater chance of being at the edge of a patch rather than surrounded by neighboring cells. This edge effect would lead to a lower average VEGF concentration around each cell. Given the negative feedback loop that regulates VEGF expression, cells in smaller patches are predicted to produce more VEGF. A prediction of this hypothesis is that adding VEGF early in the model run will suppress VEGF production, particularly from cells in the smallest patches. This prediction was tested and shown to be correct in the cell culture system. Experimentally, this was accomplished by the addition a form of VEGF (represented as V a) that binds with equal affinity to the RPE receptors, but can be distinguished from the VEGF produced by cells at assay . In a simulation, this effect of the added VEGF is represented in Eq. 2.
As for previous results, the simulated and experimental results are in qualitative agreement. Cells in smaller patches on average did respond more strongly than cells in larger patches to the added VEGF. However, on average the model predicted about a 2-fold lower response to added VEGF in the medium relative to the experimental observations. A possible explanation for this discrepancy is that the values for VEGF binding affinity β and half saturation k, two key parameters that control negative feedback for VEGF production (see Eq. 2), are estimated from a computational study . Slight discrepancies between the estimated and actual, but unknown, values of β and k could easily lead to the mismatch between experimental and simulated values.
In both the cell culture investigation and the simulation, the VEGF levels per cell obtained from smaller patch sizes (100, 200, 300 μm) decreased after the VEGF was added. This result supports the hypothesis that cells within these smaller patches reduce VEGF expression levels because of the increased levels of VEGF within their local environment. Cells in patches of larger sizes (400 μm) already encountered higher levels of VEGF. Therefore, they showed smaller changes in VEGF expression levels after VEGF addition.
Using the Model to Extend Experimental Observations - VEGF Distributions
Using the Model to Study the Effect of New Cellular Configurations
The predicted VEGF expression per cell in these inverted patterns of cell-free circles of different sizes are shown in Fig. 6b. The notable findings are that the differences between patch sizes are virtually eliminated and that the amount of VEGF expressed per cell is reduced to roughly half of that seen in the standard pattern of cell-filled circles. Both results are likely due to the larger cell-filled area in the inverted configuration (20% cell-filled area in each standard configuration and 80% cell-filled area in each inverted configuration). This pattern reduces edge effects, is predicted to result in higher local VEGF levels. In turn, this is predicted to reduce VEGF expression through negative feedback.
The model was used to provide insights into molecular events that are not accessible using current experimental techniques. Here, the model predicted that VEGF would be present at lower levels and be more evenly distributed when cells were configured in many small patches than in fewer large patches. These predicted VEGF distributions are consistent with both the experimentally determined and model-based results that VEGF expression per cell is strongest in cells distributed in small patches. These results are significant in understanding how different patterns of retinal necrosis may affect neovascularizationin AMD.
The model was applied to predict how a cellular configuration that cannot be easily designed in the laboratory will influence VEGF expression. In this configuration, a regular grid of open circles without cells is embedded in a surface with full cell coverage. This pattern represents an inversion of the standard tissue-print model of AMD and more closely resembles necrotic lesions within the retinal epithelium. The model predicted that VEGF production is nearly invariant with respect to the size of the open circles. Importantly, this result demonstrates that without significant empty space bordering the fields of cells, negative feedback predominates, leading to low, constant VEGF production independent of the size of cell-free zones.
A next step will be to extend these studies to different tissue configurations, including those that more closely match the diseased retina in AMD, and to consideration of additional parameters, such as oxidative stress  and the effects of inflammatory cytokines , that are important in the development and progression of AMD. Understanding how different patterns of necrosis can disrupt VEGF signaling will be important for developing rational therapies of neovascular AMD. Pairing cell culture studies that use micropatterning and precise measures of VEGF expression with model-based approaches offers a promising route toward accomplishing this goal.
Cell culture provides a model for replicating disease states associated with the deterioration of retinal tissue during AMD, the stimuli leading to enhanced VEGF secretion from RPE cells and the subsequent neovascularization of the choroid are still not fully understood [22, 23], and little is known about how VEGF production is regulated in the eye .
This study presents a hybrid agent-based model to support and extend cell culture models of AMD. The modeling framework was validated using experimentally gathered data on VEGF expression by RPE cells micropatterned in tissue culture dishes . Simulated results were in excellent agreement with the qualitative findings of Vargis et al.  and overall were in good quantitative agreement regarding the amount of VEGF expressed per cell in different patterning configurations.
Thanks to Jan-Ulrich Kreft School of Biosciences, University of Birmingham for providing the original version of iDynoMiCs.
Research reported in this publication was supported by the National Institute Of General Medical Sciences of the National Institutes of Health under Award Number P50GM076547, Luxembourg Centre for Systems Biomedicine, the University of Luxembourg, the Institute for Systems Biology, Seattle, USA, the Knights Templar Eye Foundation and a Ralph E. Powe Junior Faculty Award from the Oak Ridge Associated Universities. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Availability of data and materials
The data sets used and/or analysed during the current study are available from the corresponding author on reasonable request.
QBB wrote the code, ran the experiments and analysed the results. GJP designed the model and experimental approach, and performed data interpretation. EV provided the biological and biochemical data for model validation and participated in data interpretation. NSF designed the model and algorithms, and assisted in code development. All authors contributed to writing the manuscript. All authors read and approved the final manuscript.
The authors declare that they have no competing interests.
Consent for publication
Ethics approval and consent to participate
Not applicable, no human subjects or animal.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
- 3.Ding JD, Johnson LV, Herrmann R, Farsiu S, Smith SG, Groelle M, Mace BE, Sullivan P, Jamison JA, Kelly U, Harrabi O, Bollini SS, Dilley J, Kobayashi D, Kuang B, Li W, Pons J, Lin JC, Rickman CB. Anti-amyloid therapy protects against retinal pigmented epithelium damage and vision loss in a model of age-related macular degeneration. Proc Natl Acad Sci. 2011; 108(28):279–87. doi:10.1073/pnas.1100901108.CrossRefGoogle Scholar
- 4.Bird AC, Bressler NM, Bressler SB, Chisholm IH, Coscas G, Davis MD, de Jong PTVM, Klaver CCW, Klein BEK, Klein R, Mitchell P, Sarks JP, Sarks SH, Soubrane G, Taylor HR, Vingerling JR. An international classification and grading system for age-related maculopathy and age-related macular degeneration. Surv Ophthalmol. 1995; 39(5):367–74. doi:10.1016/s0039-6257(05)80092-x.CrossRefGoogle Scholar
- 8.Stone J, Itin A, Alon T, Pe’er J, Gnessin H, Chan-Ling T, Keshet E. Development of retinal vasculature is mediated by hypoxia-induced vascular endothelial growth factor (VEGF) expression by neuroglia. J Neurosci. 1995; 15(7 Pt 1):4738–47.Google Scholar
- 9.Klettner A, Westhues D, Lassen J, Bartsch S, Roider J. Regulation of constitutive vascular endothelial growth factor secretion in retinal pigment epithelium/choroid organ cultures: p38, nuclear factor ?B, and the vascular endothelial growth factor receptor phosphatidylinositol 3 kinase pathway. Mol Vis. 2013; 19:281–91.Google Scholar
- 22.Pollreisz A, Afonyushkin T, Oskolkova OV, Gruber F, Bochkov VN, Schmidt-Erfurth U. Retinal pigment epithelium cells produce VEGF in response to oxidized phospholipids through mechanisms involving ATF4 and protein kinase CK2. Exp Eye Res. 2013; 116:177–84. doi:10.1016/j.exer.2013.08.021.CrossRefGoogle Scholar
- 30.Baker QB, Podgorski GJ, Johnson CD, Vargis E, Flann NS. Bridging the multiscale gap: Identifying cellular parameters from multicellular data. In: Computational Intelligence in Bioinformatics and Computational Biology (CIBCB), 2015 IEEE Conference On. Piscataway: IEEE: 2015. p. 1–7. doi:10.1109/cibcb.2015.7300323.http://dx.doi.org/10.1109/cibcb.2015.7300323
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.