Conceived and designed the experiments: AS JAG JSG FF YJ HEG. Performed the experiments: AS. Analyzed the data: AS JAG. Contributed reagents/materials/analysis tools: AS MS. Wrote the paper: AS JAG.
The authors have declared that no competing interests exist.
Choroidal neovascularization (CNV) of the macular area of the retina is the major cause of severe vision loss in adults. In CNV, after choriocapillaries initially penetrate Bruch's membrane (BrM), invading vessels may regress or expand (CNV initiation). Next, during Early and Late CNV, the expanding vasculature usually spreads in one of three distinct patterns: in a layer between BrM and the retinal pigment epithelium (sub-RPE or Type 1 CNV), in a layer between the RPE and the photoreceptors (sub-retinal or Type 2 CNV) or in both loci simultaneously (combined pattern or Type 3 CNV). While most studies hypothesize that CNV primarily results from growth-factor effects or holes in BrM, our three-dimensional simulations of multi-cell model of the normal and pathological maculae recapitulate the three growth patterns, under the hypothesis that CNV results from combinations of impairment of: 1) RPE-RPE epithelial junctional adhesion, 2) Adhesion of the RPE basement membrane complex to BrM (RPE-BrM adhesion), and 3) Adhesion of the RPE to the photoreceptor outer segments (RPE-POS adhesion). Our key findings are that when an endothelial tip cell penetrates BrM: 1) RPE with normal epithelial junctions, basal attachment to BrM and apical attachment to POS resists CNV. 2) Small holes in BrM do not, by themselves, initiate CNV. 3) RPE with normal epithelial junctions and normal apical RPE-POS adhesion, but weak adhesion to BrM (e.g. due to lipid accumulation in BrM) results in Early sub-RPE CNV. 4) Normal adhesion of RBaM to BrM, but reduced apical RPE-POS or epithelial RPE-RPE adhesion (e.g. due to inflammation) results in Early sub-retinal CNV. 5) Simultaneous reduction in RPE-RPE epithelial binding and RPE-BrM adhesion results in either sub-RPE or sub-retinal CNV which often progresses to combined pattern CNV. These findings suggest that defects in adhesion dominate CNV initiation and progression.
This paper tests hypotheses for the mechanisms of choroidal neovascularization (
We first review the key components of the retina and the processes commonly hypothesized to underlie CNV. We then discuss our main hypotheses for CNV mechanisms and why we believe adhesion may play an important role in both initiation and progression of CNV. We then use a multi-cell computer simulation of a mechanistic computational model of the choriocapillaris, BrM and photoreceptors to investigate the effects of adhesion variations on CNV initiation and progression. Finally, we focus on how adhesion in the BrM-RPE-POS complex changes due to aging and inflammation both in human retina and in animal models of CNV and discuss the biomedical implications of our results.
In pathological angiogenesis,
The hallmark of
The diverse CNV scenarios are categorized based on histological
Developing more effective targeted intervention strategies will depend on understanding CNV mechanisms. However, because of the structural complexity of the normal and diseased retina and the numerous homeostatic and developmental mechanisms operating concurrently, experiments have yet to identify clearly the mechanisms responsible for either CNV initiation or progression. As a novel approach to developing such understanding, this paper applies quantitative models and computer simulations to test hypotheses for the mechanisms leading to CNV initiation and controlling early and late Type 1, 2 and 3 CNV.
Multiple hypotheses compete to explain CNV initiation, growth and patterning (for comprehensive reviews, see
Excessive expression of VEGF, mainly in response to injuries and hypoxia, without balancing expression of angiogenesis inhibitors is a major stimulator of neoangiogenesis in most tissues. Excess VEGF has been considered a primary cause of CNV
Both age-related changes of the retina and pathological conditions can increase VEGF expression. Life-long accumulation of lipids in BrM and their gradual oxidation (producing reactive oxygen species and recruiting immune cells) correlate with increased production of VEGF by the RPE and greater likelihood of developing CNV
Irregularities of BrM include focal breaks and thinning in BrM, abnormal production of ECM by the RPE, and formation of soft drusen. All of these BrM defects correlate with CNV
However, while BrM does not form a mechanical barrier to persistent EC penetration of the retina, BrM and the RPE attached to it clearly do form an effective barrier to choroidal penetration, even in the presence of small holes in BrM. The nature of this barrier is not clear. Haptotaxis may play a role. ECs exhibit strong haptotactic preference for their own basement membrane. The basement membrane of the RPE (RBaM) differs in structure and components from the CC basement membrane (CC BaM) (reviewed in
Since overexpression of VEGF and reduction in BrM's barrier function may not fully explain CNV initiation, multiple types, loci and progression
While not usually considered crucial to CNV, a great deal of experimental evidence suggests that failures of adhesion are essential for the development of CNV (
Context | Adhesion | Effects | ||||||
Condition | Subject | RPE-RPE | RBaM-BrM | RPE-POS | RPE Viability | CNV Loci | Simulation Results | |
1 | Normal Aging (No Drusen) | Human | + | + | + | + | - | No initiation even in presence of small holes in |
2 | Hard Drusen |
Human | − | − | + | − | - | See the |
3 | Soft Drusen |
Human | −/+ | −/− | −/+ | −/+ | Sub-RPE | |
4 | Sub-retinal Drusenoid (reticular pseudodrusen) |
Human | −/− | −/−/+ | −/− | −/+ | Sub-Retinal and/or Sub-RPE | |
5 | BrM Calcification |
Human | −/+ | −/+ | + | −/+ | * | * |
6 | Active Inflammation |
Young Human | − | + | −/−/+ | −/+ | Sub-Retinal | |
7 | Retinal Detachment |
Cat | −/− | + | − | + | * | |
8 | High Fat Diet+Aging+Blue Light |
Mouse | −/+ | −/− | + | +/− | Early Sub-RPE | |
9 | Chemotoxicity |
Rabbit | − | −/+ | − | − | Sub-Retinal | |
10 | Sub-Retinal Injection |
Rat, Rabbit | − | −/+ | − | −/+ | Sub-Retinal | |
11 | Sub-retinal Injection and VEGF Overexpression |
Rat | − | −/+ | − | −/+ | Sub-Retinal |
Since the relative importance, roles and interactions among the different types of adhesion impairment during CNV initiation and progression are unclear, this body of experimental evidence motivated us to study the role of adhesion failures in the BrM-RPE-POS complex in CNV.
While a detailed experimental analysis of adhesion effects in CNV is desirable, it is currently impractical. No animal model exhibits the full range of AMD-related CNV pathologies
CNV involves the interaction of two complex components, the retina, with its supporting structures, and the choriocapillaris. We briefly review the functional and structural properties of these components in the context of CNV in supplementary
To allow unbiased study of CNV mechanisms, our model of the retina-RPE-CC includes objects and processes (
Object Types | Processes | ||
|
|
Adhere via |
|
Adhere via |
|||
Take up |
|||
Secrete |
|||
Secrete |
|||
Die when |
|||
Have intrinsic random motility | |||
Migrate via chemotaxis up gradients of |
|||
All Processes of |
|||
Migrate via chemotaxis up gradients of |
|||
Grow in response to |
|||
|
All Processes of |
||
Migrate via chemotaxis up gradients of |
|||
Secrete |
|||
|
Adhere via |
||
Adhere via |
|||
Secrete |
|||
Have intrinsic random motility | |||
|
|
Adhere via |
|
Adhere via |
|||
Have intrinsic random motility | |||
|
All Processes of |
||
|
|||
|
|
Non-diffusing solid material (implemented as non-motile generalized |
|
Adheres via |
|||
Degraded by |
|||
|
Adheres via |
||
Fills space unoccupied by |
|||
|
|
Diffuses | |
Decays | |||
|
Diffuses | ||
Decays | |||
|
Diffuses | ||
Decays | |||
|
Diffuses | ||
Decays | |||
Degrades |
Since CNV is usually limited to the outer retina, we model the choriocapillaris, BrM, RPE and parts of the outer retina in detail, and represent the inner retina implicitly through appropriate boundary conditions at the
The
Our model has two types of
Our model represents the in-plane epithelial junctions between healthy RPE cells by
Our model includes two types of
In an endothelium, ECs mainly adhere to other ECs via
Adherent cells suspended in liquid assume a spherical shape, meaning that non-specific cell-liquid adhesion is weak. We represent this weak cell-liquid adhesion by weak
We do not model explicitly the differences in adhesion between the apical, lateral, and basal surfaces of biological RPE cells and photoreceptors. In our model
Since we can vary independently the strength of the
To aggregate the effects of RPE-derived diffusible growth factors on the
Computer simulations can help us analyze the role of multiple mechanisms during angiogenesis, both in pathological conditions like tumor-induced angiogenesis
Since experiments suggest that both ECs and macrophages can function as tip cells in CNV, we can interpret a
We represent the adhesion-reducing effects of inflammation due to inflammatory factors and immune cells implicitly by weakening
Typically, adherent cells like RPEs need to adhere to other cells (of the same or different types) or to an appropriate substrate to remain viable. Otherwise they die. Modeled
In this section, we discuss how adhesion in the
All simulations begin with either no
We simulated 108 different adhesion
The results of our simulations are 3D time-varying
Morphometric Weights ( |
CNV Type |
|
|
|
|
0.25< |
|
We define the type of
CNV Classification | Relevant Adhesion Scenarios |
|
|
- | |
|
|
- | |
|
|
- |
To classify
Dynamics Classification | CNV Dynamics | Relevant Adhesion Scenarios |
Stable Type 1 ( |
|
|
Sub-RPE to Sub-Retinal Translocation ( |
|
|
Sub-RPE to Sub-Retinal Progression ( |
|
|
Sub-Retinal to Sub-RPE Translocation ( |
|
Not Observed |
Stable Type 2 ( |
|
|
Sub-Retinal to Sub-RPE Progression ( |
|
|
Type 3 to Sub-RPE Translocation ( |
|
Not Observed |
Type 3 to Sub-Retinal Translocation ( |
|
Not Observed |
Stable Type 3 ( |
|
|
We classify
We use multiple-regression analysis to relate the
3D plot of the regression-inferred
3D plot of the regression-inferred average
3D plot of the regression-inferred average (1−
3D plot of the regression-inferred probability of occurrence of
The strong adhesion of
We performed multiple-regression analysis (see the
We believe that our simulated initiation probabilities are higher than those observed in experiments due to two simplifying assumptions of our model: 1) We assumed that all
To classify
Because our simulations are stochastic, different replicas of the same adhesion scenario can lead to different combinations of
Typical Adhesion Scenario | ||||||
Early CNV | Sub-classes |
|
|
|
|
|
Type 1 | 1 | 3 |
|
1,2 | 3 | |
2 | 1 | 3 | 1 | 1,2 | 3 | |
Type 2 | 1 | 3 |
|
2,3 |
|
1 |
2 | 1 |
|
2,3 |
|
|
|
3 | 1 | 1 | 1 | |||
Type 3 | 1 | 1 |
|
1,2 |
|
3 |
Key:
: all strength levels.
Multiple-regression analysis of the five adhesivities accounted for 93% of the observed variance in the average
In most adhesion scenarios that develop
A) Total number of
3D visualization of a simulation replica exhibiting
Typical Adhesion Scenarios | ||||||
CNV Progression Dynamics | Sub-classes |
|
|
|
|
|
S11 | 1 | 3 | 2, 3 | 1, 2 | 3≤ |
3 |
T12 | 1 | 3 | 1 | 1 | ||
P13 | 1 | 1 |
|
1 | 1, 2 | 3 |
S22 | 1 | 3 |
|
2, 3 |
|
1 |
2 | 1 |
|
2, 3 |
|
|
|
3 | 1 | 1 | 1 | |||
P23 | 1 | 1 | 1, 2 | 1 | 1 | 1 |
S33 | 1 | 1 |
|
2 | 3 | |
2 | 1 |
|
1 | 3 | 3 |
To simplify, we list only the adhesion scenarios most prone to each type of
Adhesion scenarios in which some replicas exhibit
A) Total number of
3D visualization of a simulation replica exhibiting
A) Total number of
3D and 2D visualizations of a simulation replica exhibiting
In
3D plot of the regression-inferred probability of occurrence of
A) Total number of
3D visualization of a simulation replica showing
A) Total number of
3D and 2D visualization of a simulation replica forming
In
Generally,
A) Total number of
3D and 2D visualization of a simulation replica developing
Our simulations show that variations in five key adhesion strengths suffice to explain many of the experimentally and clinically observed dependencies of CNV initiation on drusen, inflammation, retinal detachment and iatrogenic to reproduce the main observed types and progression dynamics of CNV associated with those defects. Since pathological conditions can cause multiple adhesion failures in the BrM-RPE-POS complex, we simulated factorial combinations of graded impairments of the five adhesion types to explore the effects of biologically-coupled adhesion failures. In this section, we discuss the effects of both individual adhesion failures and their combinations on
As we discussed earlier (see the section
Patients with multiple large soft drusen or
The locus of
While experimental and clinical results are incomplete, as we discussed above, they do suggest that patients with stable sub-RPE CNV do not suffer high rates of severe RPE detachment, indicating that their RBaM-BrM adhesion is not severely impaired. Among simulations with
Since we hypothesize that more severe impairment of RBaM-BrM adhesion facilitates CNV spreading and progression, we expect higher variability of outcomes in patients with moderately impaired adhesion than in patients with severely impaired adhesion. In older patients, CNV progression timing and the size and growth rate of the CNV-affected area vary significantly patient-to-patient
Overall, the
In older patients, sub-RPE CNV may later also invade the sub-retinal space (sub-RPE CNV to CNV sub-retinal progression is a common CNV progression scenario). The factors involved in this transition are not well understood. Gradual degradation of the RPE due to sub-RPE hemorrhaging, formation of sub-RPE fibrosis and inflammation triggered by initial sub-RPE CNV are associated with this transition. The death of RPE cells during this degradation indicates that RPE-RPE adhesion is impaired. The rapid vision loss associated with the transition from sub-RPE CNV to sub-retinal CNV, indicates impaired RPE-POS adhesion, though, clinically, we do not know whether RPE-POS adhesion impairment is a cause or result of the transition. Thus these conditions imply impairment of
Clinically, adhesion strengths may change as CNV progresses. However our simulations show that P13 CNV progression can occur in patients even for time-independent adhesion. We can thus use our simulations to develop a prognosis for patients with sub-RPE CNV and in whom RPE-RPE and/or RPE-POS adhesion are impaired in addition to preexisting impairment of RBaM-BrM. Simulations that exhibit
Clinically, ET2 and ET3 CNV are not common in drusen-induced CNV in patients. However, the adhesion scenarios that exhibit
Vascular RPE detachment caused by growth of CNV under the RPE is a common complication of sub-RPE CNV in AMD. We observe a corresponding pathology when
Sub-retinal CNV without prior diagnosed sub-RPE CNV occurs, but is not as common in older patients as drusen-induced CNV
In our simulations, when
Our understanding of CNV dynamics in young patients is incomplete. Neither sub-retinal CNV to sub-RPE CNV progression (P23 CNV) nor sub-retinal CNV to sub-RPE CNV translocation (T21 CNV) has been observed clinically or histologically (since CNV is not fatal, histological data for young patients with CNV is rare). We currently do not know whether this absence of observation is due to major retinal damage due to sub-retinal CNV, which precludes the later transition (and is not included in our model), or whether Late Type 1 CNV is simply overlooked clinically because sub-retinal CNV causes much more severe vision loss. Our simulations make three predictions relevant to inflammation-induced CNV in young patients: 1) If
Chemotoxicity (10% solution of naphthalene force-fed by gavage for 5 weeks) in a rabbit model
Iatrogenic sub-retinal CNV may develop after laser photocoagulation treatment of diabetic macular edema, central serous retinopathy, proliferative diabetic retinopathy, choroidal vascular and neoplastic lesions, vascular occlusive disease and degenerative retinal-pigment-epithelium disorders (reviewed in
Subretinal injections in most animal models (
Subretinal injections can cause acute physical retinal detachment, instantaneously destroying RPE-POS contact at the site of injection. However, RPE active pumping and passive flow gradually remove the excess sub-retinal fluid, allowing the retina to reattach within a few days. Sub-retinal injection also almost always induces significant inflammation, which gradually reduces RPE-RPE epithelial adhesion over a period lasting a few days to a few weeks. Such condition of transient detachment and followed by long-lasting inflammation induces RPE migration
Based on these experimental observations, sub-retinal injections appear mainly to impair RPE-RPE and RPE-POS adhesion comparable to the adhesion scenarios prone to
Our simulations predict that for secondary sub-RPE CNV to develop near a pre-existing sub-retinal CNV RPE-BrM (RBaM-BrM) adhesion must be severely impaired, to develop, independent of the degree of impairment of other types of adhesion in the BrM-RPE-POS complex. To validate our prediction, experiments would need to examine in detail the interface between BrM and the RPE basement membrane in retinal regions far from site of injection and before any initiation of sub-RPE CNV.
In these animal models, sub-retinal injection often causes rapid (less than a month) CNV initiation. We can infer that injection impairs RPE-RPE adhesion because of observed inflammation and RPE proliferation and RPE-POS adhesion because of photoreceptor degradation
We now consider how local cytokines and growth factors that can increase the chemotactic activity of endothelial cells could affect CNV initiation and progression. Our simulations did not explore the role of these factors on the chemotactic activity of ECs directly. However, our simulations using different VEGF-A levels suggest that increased chemotactic activity of
In transgenic mice with inducible expression of VEGF in their RPE cells, induction of excess VEGF only induces CNV if combined with sub-retinal injections which disrupt the RPE
In experiments, ocular hypoxia caused by systemic hypoxia usually promotes retinal angiogenesis, but has no observed effect on the RPE and does not induce choroidal angiogenesis (reviewed in
As we discussed in the
Our current model does not include several mechanisms which may also be important to CNV. In future refinements, we will include multiple types of basal deposits and fibrosis (synthesis of new ECM) explicitly to clarify their role in the initiation and progression of CNV. We particularly are interested in how differences in the size and structure of soft and hard drusen affect the initiation and progression of CNV and the frequency of occurrence of RPE detachment and RPE tear formation after therapeutic intervention to treat CNV.
Many hypothesized mechanisms for CNV initiation and progression involve irregularities in transport. We plan more realistic models including capillary maturation, blood flow and retinal-CC fluid flow to study how oxygen, nutrient and waste transport promote or inhibit CNV.
Since cell adhesion is essential to multicellularity and is important in embryonic development, homeostatic maintenance of adult tissues and diseases like cancer, its importance in CNV is, perhaps, not surprising, given CNV's many parallels with tumor-induced angiogenesis. However,
Beyond retinal CNV, our results on CNV initiation and epithelial breakdown apply to any tissue in which a basement membrane separates a capillary network from a nearby epithelium,
Ultimately, a database of verified simulations for different adhesion scenarios might allow systematic CNV prediction based on clinically-measurable properties of the eye, especially if the adhesion properties can be inferred noninvasively,
One crucial aspect of a model-based approach to CNV diagnosis, prognosis and treatment is that both the simulation database and the statistical predictors could be continuously refined using feedback from both clinical and histopathological sources, so they would improve with use, providing a platform to integrate clinical and histopathological data for even more accurate diagnosis and prognosis.
Our simulations use the Glazier-Graner-Hogeweg model (
Geometrical Parameters | ||
Name | Description | Values |
|
∼30 µm (compare to |
|
|
∼24 µm (compare to |
|
|
Location of OLM measured from |
∼67 µm (compare to |
|
6 µm (compare to |
|
|
12 µm (compare to |
Oxygen Transport Parameters | ||
Name | Description | Values |
|
Diffusion coefficient of oxygen in retinal tissue | 2.0×10−5 cm2 s−1 |
|
Light-adapted (dark-adapted) |
13 (26) ml O2 min−1 |
|
80 mmHg |
|
|
60 mmHg (extrapolated from rat |
|
|
18 mmHg |
|
|
Average oxygen partial pressure in the |
65 mmHg during light-adaptation, 61 mmHg during dark-adaptation (see |
|
Average oxygen partial pressure in the |
49 mmHg during light-adaptation, 45 mmHg during dark-adaptation (see |
|
Oxygen flux from 100 g of |
3.42 (CC3D) or 102 ml O2 (100 g tissue min)−1 during dark-adaptation 2.67 (CC3D) or 80 ml O2 (100 g tissue min)−1 during light-adaptation |
|
Oxygen flux from 100 g of |
2.81 (CC3D) or 84 ml O2 (100 g tissue min)−1 during dark-adaptation 2.05 (CC3D) or 61 ml O2 (100 g tissue min)−1 during light-adaptation |
VEGF Transport Parameters | ||
Name | Description | Values |
|
Decay rate of |
1 h−1 |
|
Diffusion length of |
13.4 µm |
|
Diffusion coefficient of |
0.25×10−10 cm2 s−1 |
|
Diffusion length of |
3 µm |
|
Secretion rate of |
∼50 pg (cell h)−1 (compare to |
|
Basal secretion rate of |
∼25 pg (cell h)−1 (compare to |
|
Basal secretion rate of |
∼25 pg (cell h)−1 (0.2 (voxel MCS)−1) |
|
Basal uptake rate of |
∼300 ligated molecules per EC×2.8×10−4 (internalization rate) |
MMP Transport and BrM Degradation Parameters | ||
Name | Description | Values |
|
Secretion rate of |
0.148 molecule (cell sec)−1 (1 molecule (voxel MCS)−1) |
|
Diffusion length of |
0.2 µm |
|
0.0094 µm3 (sec molecule)−1 (0.075 voxel (MCS molecule)−1) |
Fields | ||
Name | Description | Units |
|
mmHg | |
|
molecule/voxel | |
|
molecule/voxel | |
|
molecule/voxel |
Cell Types | Stalk | BrM | RPE | POS | PIS | Medium |
|
−20 | −10 | −10 | −10 | −10 | 3 |
|
−12 | −38/−28/−18 | 0 | 0 | −1 | |
|
−40/−18 | −16/−1 | −16/−1 | 3 | ||
|
−16 | −16 | 3 | |||
|
−16 | 3 | ||||
|
0 |
Negative contact energies represent adhesive interactions; positive contact energies represent repulsive interactions. More negative contact energies indicate stronger adhesive interactions. (/) separates the reference, moderately impaired and severely impaired levels of
Labile Adhesion Strength | ||||
Cell-Type Pairs | Name | Normal: 3 | Moderately Impaired: 2 | Severely Impaired: 1 |
|
|
−40 | - | −18 |
|
|
−38 | −28 | −18 |
|
|
−16 | - | −1 |
|
|
−16 | - | - |
|
|
−16 | - | - |
More negative contact energies indicate stronger adhesive interactions. (-) denotes
Plastic Coupling Strength | ||||
Cell-Type Pairs | Name | Normal: 3 | Impaired: 2 | Severely Impaired: 1 |
|
|
300 | 60 | 30 |
|
|
300 | 60 | 30 |
|
- | 30 | - | - |
|
- | 30 | - | - |
|
- | 30 | - | - |
|
- | 200 | - | - |
|
- | 150 | - | - |
|
- | 50 | - | - |
|
- | 50 | - | - |
|
- | 50 | - | - |
|
- | 200 | - | - |
|
- | 25 | - | - |
|
- | 25 | - | - |
Larger plastic coupling strengths represent stiffer linear springs. (-) denotes values of
All our simulations use the open-source CompuCell3D simulation environment (
(PDF)
(PDF)
(PDF)
(PDF)
(PDF)
(PDF)
(PDF)
(PDF)
(PDF)
(PDF)
(PDF)
(PDF)
(PDF)
(PDF)
(PDF)
(ZIP)
(AVI)
(AVI)
(AVI)
(AVI)
(AVI)
(AVI)
We would like to thank Dr. Thomas J. Gast for his biological insights and guidance and Randy Heiland for providing invaluable CompuCell3D support and instruction.