The mechanics of retinal detachment

Tom Chou and Michael Siegel

Depts. of Biomathematics and Mathematics, UCLA, Los Angeles, CA 90095-1766, and Dept. of Mathematical Sciences and Center for Applied Mathematics and Statistics, New Jersey Institute of Technology, Newark, NJ, 07102-1982

Submitted to Proceedings of the National Academy of Sciences of the United States of America

We present a model of the mechanical and uid forces associated with exudative retinal detachments where the retinal photoreceptor cells separate typically from the underlying retinal pigment epithelium (RPE). By computing the total uid volume ow arising from transretinal, vascular, and retinal pigment epithelium (RPE) pump currents, we determine the conditions under which the subretinal uid pressure exceeds the maximum yield stress holding the retina and RPE together, giving rise to an irreversible, extended retinal delamination. We also investigate localized, blister-like retinal detachments by balancing mechanical tension in the retina with both the retina-RPE adhesion energy and the hydraulic pressure jump across the retina. For detachments induced by traction forces, we nd a critical radius beyond which the blister is unstable to growth. Growth of a detached blister can also be driven by inamed tissue within which e.g., the hydraulic conductivities of the retina or choroid increase, the RPE pumps fail, or the adhesion properties change. We determine the parameter regimes in which the blister either becomes unstable to growth, remains stable and nite-sized, or shrinks, allowing possible healing. The corresponding stable blister radius and shape are calculated. Our analysis provides a quantitative description of the physical mechanisms involved in exudative retinal detachments and can help guide the development of retinal reattachment protocols or preventative procedures.
retinal detachment

(a)

(b)

Fig. 1.

Optical Coherence Tomography (OCT) image of exudative retinal detachments. The transverse dimension of each image is approximately 3mm. (a) Exudative retinal detachment in which the retinal photoreceptor layer has separated from the RPE cell layer. The small delamination on the left has lifted the RPE from the choroid as in (b). (b) Detachment in which the RPE layer (red) has separated from the choroid substrate. Used with permission from Dr. T. Bennett, CRA, FOPS, and the Ophthalmic Photographers Society.

Despite the physical forces implicated in retinal detachment and many existing protocols for reattachment [5], a quantitative description of the underlying mechanics is lacking. In this paper, we develop a mathematical model of exudative retinal detachments which incorporates the relevant uid ows, mechanical tensions, and cellular adhesion forces.

mathematical model

blisters

Retinal Structure, Forces, and Flows

The basic forces associated with retinal detachment include pressure-driven ows, active RPE pump ows, cell-cell adhesion forces, and possibly retinal tension forces. Retinal photoreceptor cells normally adhere to the RPE. A schematic of a typical macromolecular adhesion interaction between adjacent cells is shown in Fig. 2. In the absence of hydraulic forces, the potential exhibits a minimum at z = 0, corresponding to the equilibrium adhesion bond length between apposed retinal and RPE cells. When external separation forces exceed a critical value, the cellular adhesion bonds rupture irreversibly. In addition to cellular adhesion, forces arise from pressuredriven uid ows across the dierent tissue layers. Assuming that the uid pressure in the choriocapillaris is Pc and the intraocular pressure of the vitreous humor is PIOP , the total pressure dierence across the entire system is Pc PIOP . Both Pc and PIOP are measurable and somewhat controllable. The capillary pressure is related to the venous blood pressure, while the intraocular pressure PIOP is controlled by the total ow of vitreous humor, which is produced by the ciliary body at the base of the iris and drains predominantly through the trabecullar meshwork at the perimeter of the cornea [6, 7, 8]. Since P Pc PIOP 5 10mm Hg [9], the higher vascular

Introduction
he retina is composed of multiple layers of photoreceptor and nerve cells that sit atop the retinal pigment epithelium (RPE). The RPE is a monolayer of cells attached to the Bruchs membrane, a 2 4m elastic collagen layer that covers the choriocapillaris, the inner, vascular part of the choroid substrate. The choriocapillaris is the part of the choroid that feeds the metabolically-active RPE and retina. Separations between cell layers give rise to retinal detachments that are a major contributor to retinal tissue death and permanent vision loss [1]. The two common types of detachments are rhegmatogenous, in which a hole or tear forms in the thin retinal layer that has separated from the choroid, and exudative, in which a blister forms without a retinal hole (shown in Fig. 1) [2]. This latter type of detachment is typically associated with vitreous forces that pull the retina from the RPE, or with inammation or vascular abnormalities that result in pockets of accumulated subretinal uid, without tearing the retina. In Fig. 1(a), a large blister has formed between the retina and the RPE layer. While this is the typical location of blister formation, separation of the RPE cell layer (red curves) from the Bruchs membrane that covers the choriocapillaris can also occur, as shown by the small blister in Fig. 1(a) and in Fig. 1(b). There exist numerous clinical protocols for preventing and treating detached retinas, including vitrectomy, laser photocoagulation, scleral buckling [3], and pneumatic retinopexy [1, 2]. These methods exploit dierent mechanical forces and are used in dierent circumstances. For example, vitrectomies sever the attachment of the vitreous humor to the detached retina, laser photocoagulation induces injury, allowing a stronger re-adhesion of the scar tissue, and pneumatic retinopexy (injection of a gas bubble to envelope a torn retinal ap) [4].
www.pnas.org

Reserved for Publication Footnotes

PNAS

Issue Date

Volume

Issue Number

19

pressure drives uid ow from the choriocapillaris to the inner eye in the absence of other active processes.

(a) RPEretina interaction

U(z)
posttear interaction pretear interaction

(b) pretear adhering force

max

Pb P IOP > max z

0

U*

max

z * Pb P IOP < max

z
0

z *

max U* /z*

Fig. 2. (a) The intrinsic adhesion energy per area U (z ) between retinal layer cells and the RPE, uniformly separated by distance z . The interaction is dened at a reference pressure dierence where Pb = PIOP (see Fig. 3). At this reference pressure dierence, the RPE-retina binding energy per area is dened by U . Dierences between Pb and PIOP will tilt the potential U (z ) U (z ) z (Pb PIOP ), changing the eective detachment energy per area to U z (Pb PIOP ). The pre-tear interaction is shown by the solid blue curve. When the displacement extends beyond z , bonds are irreversibly broken, and the attractive interaction potential is destroyed (black dashed curve). (b) The adhesive force per area (stress) between the two layers (z ) dU (z )/dz for separations z < z . The maximum yield stress that can be sustained by the tissue is approximately max U /z . We also dene a uid pressure Pb in the narrow extracellular space between the RPE and retina that contains the adhesion bonds (see Fig. 3). Given any specic state of the eye, we will assume that PIOP and Pc are known parameters that can be determined from measurements (e.g., tonometry and blood pressure), but that the subretinal pressure Pb is determined by uid ow through the choroid and retinal layer.

where Lc and Lr are the eective hydraulic conductivities of the choriocapillaris (including that of the RPE and Bruchs membrane) and the retina, respectively. These conductivities will depend on the physiological state of the tissue. For example, loss of integrity of the tight junctions between the RPE cells or inammation of the choriocapillaris may increase substrate leakiness and Lc . The permeability of Bruchs membrane has also been shown to increase with age [10], decreasing Lc . Nonetheless, if Pc > PIOP , passive ows alone will tend to separate the retina from the substrate. In addition to passive ows, there is a component of the volume ux that is actively pumped by the RPE layer. The RPE cell layer is special in that it actively pumps uid from the retinal space through the Bruchs membrane and back into the choroid[2]. This volume ux derives from active cellular processes and thus depends on the physiological state of the RPE, and possibly on the trans-RPE hydrostatic pressure difference Pb Pc > 0. However, since RPE pumps are highly metabolic and regulated, we will assume that Jp is independent of Pb Pc for physiologically representative pressures [11]. Since active RPE pumping is thought to provide an important force keeping the retina attached to the choroid [12, 13], the position and physiological state of the RPE cell layer is an important component of our model. In our analysis, we assume Jp is a pressure-independent parameter that depends on the state of the RPE cells and the regulation mechanisms aecting the pumps. Table S1 in the Supporting Information (SI) lists the parameters used in our modeling and discusses their ranges of values. Whenever possible, our results will be evaluated and discussed within a reasonable range of experimentally measured parameter values.

(z ) = dU(z ) dz

Uniform retinal detachments

First, we consider the stability of attached retina against large, uniform delamination from the RPE. In the presence of uid ow, static equilibrium requires that the volume ow into the subretinal space balances the ow out of it: Jc = Jr + Jp . [2]

z > z*

transretinal flow J r
z retina

PIOP
Lr

transretinal flow
retina

PIOP

11111111111111111111111111111111111111111111111111111 00000000000000000000000000000000000000000000000000000 00000000000000000000000000000000000000000000000000000 11111111111111111111111111111111111111111111111111111 00000000000000000000000000000000000000000000000000000 11111111111111111111111111111111111111111111111111111

P b

Lc

11111111111111111111111111111111111111111111111111111 00000000000000000000000000000000000000000000000000000 00000000000000000000000000000000000000000000000000000 11111111111111111111111111111111111111111111111111111 00000000000000000000000000000000000000000000000000000 11111111111111111111111111111111111111111111111111111 00000000000000000000000000000000000000000000000000000 11111111111111111111111111111111111111111111111111111 00000000000000000000000000000000000000000000000000000 11111111111111111111111111111111111111111111111111111
choriocapillaris Pc

(a)

choriocapillaris Pc

Bruchs RPE (b) membrane

Upon using Eqs. 1 in Eq. 2, we solve for the hydrodynamic pressure dierence Pb PIOP tending to separate the retina from the RPE, and balance this stress with the membrane adhesion force dU (z )/dz : Pb PIOP = = Lc P Jp Lr + Lc Lr + Lc dU (z ) (z ), dz

exudative fluid flow J c

RPE pump flow J p

Schematic of a uniform exudative retinal detachment. The retina, RPE, Bruchs membrane, and choriocapillaris are colored green, red, gold, and pink, respectively. The tissue hydraulic conductivities, intraocular and choroidal pressures, PIOP and Pc , and expected ow directions are labelled. In (a), the RPE pumps are active and the indicated passive uid ow direction results from Pc > Pb > PIOP . In this case, the retinal is stable against delamination at the retina-RPE interface. (b) If Pc is increased, and/or the RPE pump ux Jp is decreased, uid accumulates at the retina-RPE interface, irreversibly separating the cell layers. The resulting delamination can only be reattached by changing conditions to those that allow drainage of the interstitial uid.

Fig. 3.

[3]

Pressure-driven uid ows cross between the choroid space and the subretinal space, and between the subretinal space and the vitreous space. The corresponding volume uxes, Jc and Jr , are expressed as

Jc = Lc (Pc Pb ),
2 www.pnas.org

and

Jr = Lr (Pb PIOP ),

[1]

where P Pc PIOP . For a given stress function (z ) dU (z )/dz , Eq. 3 can be solved to nd the cellular layer separation z as a function of P , the RPE pumping ux Jp , and the tissue permeabilities Lr , Lc . When P < 0, Eq. 3 can be satised only if (z ) is negative. In this regime, the outward passive ow compresses the retina onto the choroid and detachments can occur only through external traction forces from the vitreous humor. On the other hand, when P > 0 is suciently large, and/or the RPE pump ux Jp is suciently small, the uid pressure could push the layers to a new equilibrium separation z > 0. In cases where Pb PIOP > max{dU (z )/dz } max , the maximum yield stress is exceeded, there is no value of z that can satisfy Eq. 3, and the retina irreversibly rips from the RPE (Fig. 2(b)). Thus, when
Footline Author

Jp Lc P > max , Lr + Lc Lr + Lc

[4]

an exudative retinal detachment of innite extent spontaneously forms. Fig. 3(b) shows the ows and pressures at which the retina and RPE have irreversibly separated. Condition (4) is one of our main results and shows that if Lc (Pc PIOP )/(Lr + Lc ) < max , an exudative detachment cannot arise solely from RPE pump failure (Jp 0). Using experimentally-determined parameter values listed in Table S1, we nd that for a reasonable eective bond length z 107 m, a loss of RPE pump function will typically not give rise to spontaneous uniform delamination unless U 104 J/m2 , which is much smaller than typically measured values [14]. Therefore, loss of RPE function alone is insucient to induce large-scale delamination, which requires external traction forces. However, we will show that loss of RPE pump function can dramatically aect the size and stability of a localized pre-existing retinal blister.

Stability at the contact perimeter of the blister requires balancing of the tension T and the binding energy per unit area U (z ) holding the retina to the RPE in the undetached region. If the cellular bonds are ruptured by the total magnitude of the tension force, the tension in the detached retina will obey T U z (Pb PIOP ) U [6]

Retinal blisters
We now consider retinal blisters that result from local traction forces from the vitreous, or from local heterogeneities in the physiological state of the tissue. The basic geometry of a localized retinal blister is approximated by a spherical section as depicted in Fig. 4. The circular footprint of the blister is denoted Ad , while the cap area of the detached retina is Ar . A lesion of inamed tissue, if it exists, will have area A Ad .

Equation 6 represents the external stress required to pull o unit area of retina that is attached to the RPE in the presence of the pressures Pb and PIOP . Henceforth, we assume that the region outside the localized blister is far from uid-driven spontaneous delamination and that Pb PIOP max U /z . Therefore, we can safely accept the last approximation U z (Pb PIOP ) U . In the Supporting Information, we describe a more general rupturing relation that takes into account the angle at which the tension is applied. With suitable constitutive relations between the pressure Pb and the tension T and the blister geometry, Eqs. 5 and 6 can be used to determine stable blister geometries and sizes. In contrast to, say, a liquid-gas interface, the tension T in our problem is not constant but depends on the amount of stretching the detached membrane suers, relative to the at, undetached retina. If we assume the detached retina resists stretching with a force linearly proportional to the amount of excess area, the tension T will depend explicitly on the dimensionless shape ratio f according to T (f ) = Ed (A r A d ) = Ed Ad 1 1 , f [8]
2 Ad R = , Ar Ar

[7]

PIOP T Pb P b R
lesion Al

where E is the Youngs modulus of the detached retina and d is its thickness. The combination Ed represents the stretching elasticity of the detached retina. Geometrically, the radius of curvature R can be expressed as

Pc

R (f ) =

Fig. 4. Schematic of a retinal blister. The intraocular pressure, the pressure in the blister, and the eective pressure underneath the attached retina are denoted , and P , respectively. The blister geometry is dened by a spherical PIOP , Pb b section with radius of curvature R and footprint radius R . A localized lesion with area A may exist within the blister. The tissue within this lesion may be inamed or increased tissue permeability. p with diminished RPE pump activity J We assume that there is negligible lateral uid ow between the blister interior and the innitesimally thin extracellular space between the retina and the RPE, and that the two pressures Pb and Pb are independently approximated by the uid ow through each of the spaces, respectively [15]. Moreover, we neglect the tamponading forces that might arise from the gel-like vitreous inside the eye. This assumption is appropriate for older patients where the vitreous humor has liqueed, and for patients who have undergone vitrectomy. For a blister with the geometry shown in Fig. 4, static equilibrium is achieved only when the appropriate pressure and retinal tension forces are balanced and there is no net volume ow into the blister. According to Laplaces law, the pressure dierence Pb PIOP must be balanced by the retinal tension T times its curvature:
Pb PIOP =

Finally, it will prove convenient to nondimensionalize all hydraulic conductivities by Lc , all pressures by P , all uxes by Lc P , and the tension and adhesion energy by Ed. Our new dimensionless variables are dened by J p = Jp = U . r = Lr , and U , L Lc P Lc Ed [ 10 ]

Ar = 4 (Ar Ad ) 2

R . f (1 f )

[9]

Upon using expressions 8 and 9 for T (f ) and R(f ) in Eqs. 5 and 6, the dimensionless equilibrium conditions can be written as
b IOP = P P

2T (f ) R(f )P

and

[ 11 ]

1 f , [ 12 ] + 1 U respectively. Note that the adhesion energy alone xes possible values of the shape parameter f which determines the normalized blister cross-sectional prole h(y x/R )/R . For h(0) R, basic geometry gives the blister prole as a function of f 1/2: f h(y ) = R 1 y2 4f (1 f )
PNAS Issue Date

2T . R

[5]

1 1. 4f (1 f )
Volume

[ 13 ]
3

Footline Author

Issue Number

In cases where the maximum height h(0) > R (f < 1/2), similar equations for blisters with overhangs can be derived.

. Fig. 5. The blister prole plotted for various normalized adhesion energies U , Note that the blister shape depends only on the normalized adhesion energy U while the radius R is a function of all other parameters in the problem.

Figure 5 plots the normalized blister height h(y )/R as a function of the normalized radial coordinate y = x/R for . The blister becomes various values of the adhesion energy U , with the tranmore bulbous for larger adhesion energies U sition from a blister with h(0) < R to one with an overhang occuring when f falls below 1/2. From Eq. 17, this occurs for > 1, and is illustrated by the U = 3 example in Fig. 5. U For the sake of simplicity, we have neglected the bending rigidity of the detached retinal layer which can be estimated by 1 B 12 Ed3 /(1 2 ), where is the Poissons ratio of the retinal tissue. Incorporation of bending forces would smooth the kink of the retinal surface across the point of contact. The [16] width of this transition region scales as B /T d/ U which, using typical parameters from Table S1 is about 0.5mm and fairly at blisters. Therefore, bending rigidfor small U ity will not qualitatively aect the shape of the blister and we henceforth neglect it. Equation 12 only provides a condition for the blister shape. To determine the blister size, we must now solve Laplaces pressure jump equation (Eq. 11) while preserving + 1)1 found above from balancing the condition f (U the tension with the retinal adhesion energy. Eq. 11 imposes p , L r, an additional constraint on the physical parameters J and U , which can only be satised for certain blister ra dius R . We now specify two models for Pb PIOP . We will rst assume that a circular region of some radius R has been previously mechanically detached and that within this footprint, the molecular bonds have been irreversibly broken, and the physiological state under the blister is uni PIOP can form. A relationship for the pressure dierence Pb be found by imposing volume conservation under the blister ) Ad Jp . The dimensionless Ar Lr (Pb PIOP ) = Ad Lc (Pc Pb form of Eq. 11 becomes: p ) (1 J = r + f L 1 1 f
3/2

above which no stable solution to R exists. Pre-formed bliscr ters with R R are stable while those with footprint racr dius R > R are unstable. To maintain the same stable blister pressure Pb (so that net volume ow into the blister remains zero) as the footprint radius R increases requires that the curvature of the detached retina remain constant. Therefore, stability requires that f decrease and that the blister become more bulbous. Beyond a certain critical radius cr R , the small-f , bulbous blister can no longer be supported + 1), and the blister sponby the force condition f 1/(U taneously grows without limit. Note that for suciently large pump uxes such that p < 0, there is no formal solution to Eq. 11 since any 1J blister will be sucked at onto the RPE, and eventually possibly heal itself. These results highlight the importance of the and of properly functioning retina-choroid binding energy U impart RPE pumps. Suciently large binding energies U stability to large areas of detachment, while RPE pumping that is eective enough to render 1 J p < 0 spontaneously . Using the reabsorbs blisters regardless of the value of U typical parameter values listed in the SI, it is typical that un der normal physiological conditions J 1. Finally, note that p Eqs. 14 and 15 also apply to RPE detachments (as depicted in Fig. 1(b)), where the RPE is lifted o the Bruchs membrane overlying the choroid and uid is actively pumped into the blister. An RPE detachment would thus be described by our representing the existing model, but with Jp Jp , and U RPE-Bruchs membrane adhesion energy. The corresponding solution to R (Eq. 15) indicates that an RPE detachment cr can have a smaller critical radius R due to uid pumping cr into the blister, or a larger R if the RPE-Bruchs membrane adhesion is stronger than the retina-RPE adhesion. For a second class of physiologically-motivated detachment scenarios, we calculate the equilibrium size and shape of an exudative detachment that covers a circular choroidal/RPE lesion of known area A (as shown in Fig. 4). In this case, the equilibrium blister size R is not given by some past trauma, but is to be determined by stability conditions. Tissues within the lesion is in a dierent physiological state and have associ ated parameters L c , Jp , and/or U that can be dierent from Lc , Jp , and/or U of the nonlesioned tissue. For example, the RPE layer, Bruchs membrane, or choriocapillaris may be inamed within the lesion, leading to a higher permeability, L c > Lc . The lesion may also represent tissue with lower RPE pump expression or function, rendering Jp < Jp . We now assume that retina over the lesion has lost adhe sion (U = 0). Upon balancing the total volume ux into the blister by summing the weighted volume ows associated with each tissue type (lesion or non-lesion): Ar Jr
= A Jc + (Ad A )Jc A Jp (Ad A )Jp = Ar Lr (Pb PIOP ).

4Ed R P

[ 14 ]

[ 16 ] Upon using the form Jc = L c (Pc Pb ), dening the relative lesion area A f 1, [ 17 ] g Ar and solving Eq. 16 for the dimensionless transretinal pressure b IOP , we re-express Eq. 11 in dimensionless form: jump P P (f g ) + g L 2T (f ) c (f g )Jp g Jp = . r + (f g ) + g L R P L c [ 18 ]

Solving Eq. 14 with f f yields a nite blister footprint r + f )(1/f 1)3/2 > (1 J p )/ . This radius R only if (L constraint denes a critical blister radius
cr R =

4Ed P

3/2 r (U + 1) + 1 U L

+ 1)(1 J (U p)

[ 15 ]

www.pnas.org

Footline Author

Since the lesion radius R = A / is constant, we express the radius of curvature as R(f, g ) R / 4g (1 f ). Upon using Eq. 8 for T (f ), Eq. 18 can be rewritten in the form of a cubic equation for the ratio g A /Ar : (1 L c ) 3/2 f (f ) g + L c 1 + Jp Jp g r + f ) (L p ) = 0, f (f ) g + f (1 J

p lesion tissue J 0. In this case, Eq. 19 reduces to Jp g f g + f (1 J p ) = 0. Upon nding the physical root g , we construct a normalized blister footprint radius

= R +

2J p 2 4Jp (1 Jp )

[ 22 ]

[ 19 ]

4Ed r + f ) and where (f ) (1/f 1)3/2 (L R P are convenient dimensionless combinations of parameters. By solving Eq. 19 for g as a function of f , , L c , Jp , and Jp , we can nd the parameter range that supports steady-state blisters, and the precise shape and size of the detachment. In this scenario, we assume that the blister grows until the tension is exactly balanced (the equality in Eq. 6) so that the operative + 1)1 f , leading value of the shape parameter is f = (U to the dimensionless control parameter

1 . [ 20 ] + 1 U Without loss of generality, further analytic progress can be made in a number of physiologically-relevant subcases. First, assume the RPE cell layer becomes locally more permeable to uid ow (L c > 1) and the adhesion outside the lesion is also weak. Solving Eq. 19 in the small limit for the physical root g , we nd a stable blister footprint radius
3/2 r + L (f ) = U

which depends on only two dimensionless parameters, the di mensionless healthy-tissue RPE pump ux J p , and the parameter combination . A real, stable solution to Eq. 22 exists only if 2 4Jp (1 Jp ). Once this condition is no longer . Furthermore, Eq. 22 is satised, the blister radius R 1. When the parameters are such that valid only when R the blister footprint shrinks to below that of the lesion, the problem becomes one of xed blister radius R and Eq. 15 applied, but with R interpreted as R . p > 1, uid in the subretinal space For RPE pump uxes J is alis suciently evacuated such that the blister radius R ways nite. The retina is held down by the healthy-tissue (and hence ) vanish, the RPE pumps and even when U J blister radius R p /(Jp 1). However, when Jp < 1, stability of the blister relies on the chorioretinal adhesion en , the blister becomes unstable and ergy. For suciently low U is decreased to the grows without bound. For example, if U p (1 J p /(1 J p ). If is point that = 2 J J p ), R

R = R R

f g

only when J p > 1. Figure 6(a) plots R as a function of Jp for various L J . As expected, detached blisters increase in c p size as lesion conductivity increases, and diminish in size as p increases. RPE pumping J

L c 1 + Jp Jp J p 1

[ 21 ]

The rst case arises when, for example, the retina is highly is large, f is small, stretchable and/or impermeable. Here, U and the blister is bulbous in shape. In the opposite limit of permeable and/or sti retinas, f 1, indicating that the blister is relatively at. Our results are summarized in Fig. 6(b), are shown in the J p plane. For a where values of R more detailed plots representing Eq. 22, see SI.

), further decreased (for example, by decreasing and/or U . the blister becomes unstable and R For suciently large , the retina is always stable. Once = 1, the blister radius contracts to that of the lesion, and no retina above healthy choroid tissue is detached. Further increasing does not aect the shape of this fully contracted blister. The nal shape is determined by the value of + 1)1 when = 1: f = (U r )2/3 r 1 ( L L = U r + 1))2/3 L r 1 ( (L

Discussion and Conclusions

Fig. 6.
0) and leaky (Eq. 21) in the weak adhesion (U U (a) Plot of R lesion (Lc > 1) limit. (b) Contour plot of R (Eq. 22) as functions of the RPE p and the dimensionless combination . Three qualitatively dierent pump ux J

are clearly indicated. For > 1, the physical solution g is such regimes for R = 1 indicating that the blister radius has contracted to the boundary of that R p < 1/2 1 + the lesion. For < 1, and J 1 2 , the blister pressure
P Pb IOP cannot be balanced by the retinal tension term and the blister is unstable . Only when J p > 1/2 1 + 1 2 to unbounded growth R is a

nite blister footprint radius stable.

A more complex situation arises if the choroidal permeability within the lesion is approximately that of healthy tissue (Lc = L c and Lc = 1) and the pump ux in the
Footline Author

We have developed a mathematical framework to describe exudative retinal detachments. Forces arising from hydraulic pressure dierences, active and passive uid ow, and cellular adhesion are balanced to determine the stability conditions of a uniform retinal layer. We nd that global retinal detachment occurs whenever Eq. 4 is satised. Using realistic parameter values, we nd that spontaneous, uniform delamination can occur only if both the RPE pumps fail and the retina-RPE adhesion energy are globally reduced. A more realistic scenario involves a localized exudative detachment, or blister, under which subretinal uid has accumulated [15]. One of our main ndings is that the shape of the blister depends only on the stretching elasticity of the detached retinal sheet and the adhesion energy per area between the retina and the RPE. To nd the size, or footprint radius
PNAS Issue Date Volume Issue Number 5

R , of the blister, the pressure dierence across the detached retina must be balanced with the geometry-dependent tension of the retina. We consider two clinically-relevant scenarios. The rst assumes that a circular patch of retina has been detached from an unspecied external force (such as vitreous shear forces or transient normal tension forces through collagen bers that may be attached to the retina). The physiological properties (adhesion strength, choroidal hydraulic conductivity, and RPE pump ux) under the detached region may be dierent from those in the normal undetached region. In this cr scenario, we nd a critical blister radius R above which the cr blister is unstable to further growth. Estimates for R range from millimeters to a centimeter, although the actual critical radius may be larger due to restoring forces from vitreous tamponading, which we neglect. In the second scenario, we assume a localized circular lesion of radius R , above which the retina is already detached from e.g. prior trauma. The physiological parameters in the lesion dier from the rest of the healthy tissue. Here, the blister footprint radius is determined by the physical solution to a cubic equation and other constraints. Two relevant subcases with simple analytic results arise. When the substrate tissue within the lesion is leaky and L c > 1 (for example, if the tight junctions of the RPE monolayer are locally disrupted), the blister radius in the weak adhesion limit ( 0) is determined by Eq. 21 and is plotted in Fig. 6(a). In the case where the RPE pump ux is not aected by the inamed tis-

sue, the resulting quadratic equation can be solved to nd the blister radius as a function of the dimensionless combination p = Jp /(Lc P ). Here, blister sizes are de (Eq. 20) and J p , )-space in Fig. 6(b), while the more detailed lineated in (J = R /R on and J dependences of R p are illustrated in Fig. S3 in the SI. Our model for exudative retinal detachment relates the physiological state of the retinal and choroidal tissue to physical parameters such as intraocular and vascular pressures, hydraulic conductivity, adhesion strength, retinal elasticity, and RPE pump ux. Some of these parameters can be experimentally or clinically altered using e.g., anti-inammatory drugs to decrease hydraulic conductivities, or intraocular pressure lowering drugs to decrease PIOP . Retinal detachments under dierent physiological and structural conditions can also be compared using the mechanical principles quantied in our model. Our analyses can potentially guide the clinical prediction, prevention, and repair of exudative retinal detachments. TC is grateful to Dr. Elena Bitrian of the Jules Stein Eye Institute for illuminating discussions and the Oppenheimer Foundation for a seed grant. Part of this project was conceived and developed in the Institute for Pure and Applied Mathematics, UCLA. The authors were also supported by the Army Research Oce through grant 58386MA (TC), and National Science Foundation through grants DMS-1032131 (TC), DMS-1021818 (TC), DMS-1009105 (MS), and DMS-1016406 (MS).

1. M. Yano and J. S. Duker, Ophthalmology, (Mosby International, Ltd., London, 1999) 2. T. H. Williamson, Vitreoretinal Surgery, (Springer-Verlag, Berlin, 2008). 3. W. J. Foster, N. Dowla, S. Y. Joshi, and M. Nikolaou, The Fluid Mechanics of Scleral Buckling Surgery for the Repair of Retinal Detachment, Graefes Arch Clin Exp Ophthalmol, 248, 31-36, (2010). 4. M. Stopa, A. Linco, and H. Linco, Analysis of forces acting upon submacular hemorrhage in pneumatic displacement, Retina, 27, 370-374, (2007). 5. M. F. Marmor, New hypotheses on the pathogenesis and treatment of serous retinal detachment, Graefes Arch Clin Exp Ophthalmol, 226, 548-552, (1988). 6. R. F. Brubaker, The ow of aqueous humor in the human eye, Trans Am Ophth Soc, 80, 391-474, (1982). 7. C. R. Ethier, M. Johnson, and J. Ruberti, Ocular Biomechanics and Biotransport, Ann Rev Biomed Eng, 6, 249-273, (2004). 8. M. Goel, R. G. Picciani, R. K. Lee, and S. K. Bhattacharya, Aqueous Humor Dynamics: A Review. The Open Ophthalmol J, 4, 52-59, (2010). 9. O. Maepea, Pressures in the anterior ciliary arteries, choroidal veins and choriocapillaris. Exp Eye Res 54, 731-736 (1992). 10. C. Starita, et al., Hydrodynamics of ageing Bruchs membrane: implications for macular disease, Exp Eye Res, 62, 565-572, (1996). 11. S. Tsubai, Measurement of the volume ow and hydraulic conductivity across the isolated dog retinal pigment epithelium. Invest Ophthalmol & Vis Sci, 28, 1776-1782, (1987). 12. I. Fatt, Flow and Diusion in the vitreous body of the eye, Bull Math Biol 37, 85-90, (1975). 13. I. Fatt and K. Shantinath, Flow conductivity of retina and its role in retinal adhesion, Exp Eye Res 12, 218-226, (1971). 14. H. Zauberman and E. R. Berman, Measurement of adhesive forces between the sensory retina and the pigment epithelium. Exp Eye Res, 8, 276-283, (1969). 15. H. L. Kain, A new model for examining chorioretinal adhesion experimentally, Arch Ophthalmol, 102, 608-611, (1984). 16. T. R. Powers, G. Huber, and R. E. Goldstein, Fluid-membrane tethers: Minimal surfaces and elastic boundary layers, Phys Rev E, 65, 041901, (2002). 17. D. M. Maurice, J. Salmon, and H. Zauberman, Subretinal pressure and retinal adhesion, Exp Eye Res 12, 212-217, (1971). 18. D. M. Maurice, Flow of water between aqueous and vitreous compartments in the rabbit eye, Am J Physiol Renal Physiol, 252, F104 - F108, (1987). 19. R. J. Antcli, A. A. Hussain, and J. Marshall, Hydraulic conductivity of xed retinal tissue after sequential excimer laser ablation. Arch Ophthalmol, 119, 539-544, (2001). 20. D. J. Moore, A. A. Hussain, and J. Marshall, Age-related variation in the hydraulic conductivity of Bruchs membrane. Invest Ophthalmol & Vis Sci, 36, 1290-1297, (1995). 21. I. L. Jones, M. Warner, J. D. Stevens, Mathematical modelling of the elastic properties of retina: a determination of Youngs modulus. Eye, 6, 556-559, (1992). 22. G. Wollensak and S. Eberhard, Biomechanical Characteristics of Retina. Retina, 24, 967-970, (2004).

23. A. Reichenbach, W. Eberhardt, R. Scheibe, C. Deich, B. Seifert, W. Reichelt, K. Dahnert, and M. Rodenbeck, Development of the rabbit retina. IV. Tissue tensility and elasticity in dependence on topographic specializations. Exp Eye Res, 53, 241-251, (1991). 24. K. Franze, M. Francke, K. Guenter, A. F. Christ, N. Koerber, A. Reichenbach and J. Guck, Spatial mapping of the mechanical properties of the living retina using scanning force microscopy, Soft Matter, 7, 3147-3154, (2011).

www.pnas.org

Footline Author

SUPPORTING INFORMATION
Relevant physiological parameters. The parameter values, including values for the elasticity of detached retina, are taken

from dierent experiments performed on dog, pig, rabbit, and human retina.

Table S1: Table of measured parameter values. Note that 1mm Hg is equivalent to 133Pa = 133N/m2 . Physical parameter intraocular pressure vascular pressure blister pressure retinal conductivity choriodal conductivity RPE pump ux retina-RPE adhesion retinal Youngs modulus retinal thickness Symbol PIOP Pc Pb Lr Lc Jp U E d Typical values 10 20mm Hg Pc > Pb > PIOP 107 , 1010 m/(Pa s) 10
8

Reference [6, 7] [9, 18] [9, 15] [19] [20] [7, 11] [14, 15, 17] [21, 22, 23] [7]

15 25mm Hg

, 10
8

m/(Pa s)

103 , 104 , 105 Pa 100 250m

0.1J/m

10

m/s
2

While the parameter values in Table S1 vary greatly, we nonetheless use them as rough guides to the magnitude of the forces detaled in our modeling. The variability in the parameters arises from measurements across dierent animals, possibly dierent tissue layers, using dierent experimental techniques and sample preparation protocols. For example, the retinal elasticity E appears to be highly sensitive on whether it was measured using mechanical stretching [23, 22], indentation force microscopy [24], with the force indentation results giving lower values of E 1kPa, more consistent with neural tissue. Similarly, the retina-RPE adhesion energy U is typically obtained by measuring the force required to separate the retina from the RPE. However, the force required to separate the layers will depend on the RPE pump activity, which would need to be carefully subtracted from the applied force in order to obtain the intrinsic, pump-free adhesion energy. The range of values observed is 0.01 0.1J/m2 which is on the order of that of an air-water interface. Angle-dependent detachment condition. The precise balancing of tension and adhesion forces at the edge of a blister will depend on the microscopic details of how macromolecular bonds are oriented and adhere the retinal cells to the sti choroid. For example, if the bonds are free to swivel, then the cells detach when the magnitude T of the tension exceeds that of the binding energy per area U . This case is descibed by Eq. 6 and the analyses subsequent to it. However, if the bond-rupturing coordinate is say, normal to the retina, then the binding energy per area is balanced by only the normal component T sin of the retina tension. Since the bonds (or the cells themselves) holding the retinal cells can be ruptured by forces along dierent bond-rupture coordinates we inFootline Author

terpolate between the total force and normal force rupturing modes by dening the parameter through the force balance equation T cos2 + sin2 U z (Pb PIOP ) U

[ S1 ]

where the angle is the contact angle the retinal makes at the perimeter of detachment and can be related to the blister radius through R /R = sin . Upon expressing the tension T and in terms of the shape factor f , we nd 1f f , + 4(1 )f (1 f ) U [ S2 ]

; ) As the case with Eq. 12, the shape parameter f f (U and . If = 1, the magnitude of the tendepends only U sion T contributes to cell detachment and we recover Eq. 12. However, for = 0, only the normal component of T acts to detach the retinal cells. In this case, we nd ; = 0) Ad f (U Ar 1/3 4/3 2 U U 1+ =1 1 [ S3 ] 6 27 2/3 U + 2 2/3 2 U 1+ 1 . 27

; ) provide For irreversible delamination, these solutions f (U a lower bound for the appropriate shape factor f and are plotted in Fig. S1(a).
PNAS Issue Date Volume Issue Number 7

Although for = 0, the complicated expression for ; = 0 does yield the simple analytic expressions for f (U = 1 derived in the main body of the paper, Figs. S1 and S2 shows that the dierent modes of detachment do not qualitatively alter our overall results except in the extreme case 3 and blisters have overhangs. where U

; ) Ad /Ar as a function of (a) The blister shape parameter f (U = U /(Ed) (solid curve). The the normalized retinal adhesion energy per area U shape parameter is also plotted for the case where cell bonding is ruptured only by normal forces (dashed curve, see Appendix A). (b) The maximum height h(0)/R for various . The dependence on is modest. plotted as a function of log U

Fig. S1.

( , J p ). For completeness, we plot slices of Analysis of R ( , J p ) in Figs. S3. Figure S3(a) plots the the solution R (Eq. 22) as expected normalized blister footprint radius R a function of the dimensionless combination and various p . R values of the normalized, rescaled, RPE pump ux J increases with decreasing , and in particular, for Jp < 1, , at there is an abrupt transition to instability and R the points indicated by the solid dots.

; = 0). InThe dashed curves in Fig. S1 shows f (U termediate values of yield ratios Ad /Ar that interpolate between the plotted curves. Fig. S1(b) shows the relative height of the blister for various . Note that the eect of the rupture mode on the blister shape is minimal except at large 3, where the relative dierence becomes appreciable. U To illustrate the eects of the retinal detachment mode on the allowed shapes, in Fig. S2 we plot h(y )/R at a xed for = 0, 0.5, 1. adhesion energy U

as a function of at varFig. S3. (a) The normalized blister radius R p = 0.8, 0.9, 1, 1.1, 1.5. This ious dimensionless rescaled RPE pump uxes J
3/2 (Lr + 1) 3/2 (for U 1) or U Lr (for U 1). (b) The normalized blister ra as a function of the normalized RPE pump ux J p for various values of dius R = 0.1, 0.3, 0.5, 0.7, 0.9.

plot is qualitatively similar to a plot where is replaced by U

= 0.2. When Change in the blister prole for varying and xed U only the normal component of tension induces detachment ( = 0), larger retinal stretching is required at the stable contact perimeter (red curve).

Fig. S2.

as a function of J p for xed . For Figure S3(b) shows R p is decreased, R increases. Less eective each xed , as J p , the RPE pumps lead to larger blisters. Below a critical J radius R growth without bound at a maximum permissible radius (dots).

www.pnas.org

Footline Author