The physics of retinal delamination Tom Chou Depts. of Biomathematics and Mathematics, UCLA, Los Angeles, CA 90095-1766 (Dated: February 25, 2012) Retinal detachments occur when adjacent layers of tissues comprising the retina delaminate from each other or from the underlying choriocapillaris substrate that feeds it. Taking into account experimentally measured mechanical and transport properties of retinal tissue, we build a physical model by treating the retina as a stack of layered, active, water permeable tissues bound by cellular adhesion forces. Our model provides a mechanistic and physiological understanding of how changes in active fluid pumping or tissue conductivity affect overall retinal stability. Theoretical predictions on the location of first retinal delamination may help guide prevention strategies and the development of new, physically-motivated retinal reattachment protocols. The retina is a multilayer tissue lining the inside of the eye. It is composed of layers of photoreceptor cells (rods and cones) and ganglion nerve cells that conduct the electrical signals from the rods and cones. Normally, the retina is attached to a monolayer of cells, the retinal pigment epithelium (RPE), which in turn is attached to an collagen layer (Bruch’s membrane). The Bruch’s membrane covers the choriocapillaris substrate, a capillaryrich tissue that supplies nutrients to the retina. Figure 1 depicts the main histological features of an intact retina [1, 2]. (a) (b) nerve fiber layer ganglion cell layer inner plexiform layer 11111111111111111111111111111 00000000000000000000000000000 00000000000000000000000000000 11111111111111111111111111111 00000000000000000000000000000 11111111111111111111111111111 000000000000000 111111111111111 000000000000000 111111111111111 choriocapillaris inner nuclear layer outer plexiform layer outer nuclear layer cones and rods RPE Bruch’s membrane blood vessels FIG. 1: (a) Histology of the retina (adapted, with permission from http://www.anatomyatlases.org/). (b) The multiple cell layers are labeled. The retina sits atop the retinal pigment epithelium (RPE), a monolayer of cells that is both permeable and that actively pumps fluid and waste products back into the choriocapillaris. Separating the RPE from the underlying vascularized choriocapillaris substrate that feeds the retina is the Bruch’s membrane, a 2-4µm thick collagen sheet. Retinal detachments disrupt the integrity of the layered retinal structure. Separation of any two retinal layers is a common mode of injury that can lead to tissue death and vision loss [2, 3]. Often, the retina detaches from the RPE, with fluid accumulating between the photoreceptor and RPE cell layers. However, the location of the delamination can be highly variable. Fig. 2(a) shows a retina that has suffered two separations with the lower one occurring between the RPE and Bruch’s membrane, another relatively frequent site of delamination. In Fig. 2(b), the separation has occurred at the inner (more towards the eye interior) layer of cells, leaving the rest of the retina and the RPE layer (thin red line) largely intact. (a) (b) FIG. 2: Optical Coherence Tomography (OCT) images of retinal separation. The lateral extent of each image is approximately 3mm. (a) Retinal detachment between the RPE and Bruch’s membrane. Here, the RPE is lifted off the choroidal substrate. Degeneration and crazing can also be seen interior to the outer nuclear layer. (b) Delamination of the nerve fiber layer (NFL). Here, the outer layers, including the RPE remain largely intact on the choriocapillaris substrate. Detachments can be rhegmatogenous, where the retina suffers a tear, or exudative, where fluid accumulates between layers of the retina. Traction forces on the retina mediated through the polymeric vitreous humor in the eye may contribute to both types of detachments. Detachments and cell layer separations can also be caused by a variety of other factors including age, diabetes [4], myopia, cataract surgery, infections, and tumors [2, 3], which manifest themselves through changes in the physical properties of the relevant tissue. For example, changes in hydraulic conductivities have been thought to affect retinal stability [6, 7]. Many clinical procedures have been developed to repair or prevent separated retinas. Besides vitrectomy (surgery to remove the vitreous that may be connected to and pulling on the retina), pneumatic retinopexy (injection of gas into the eye) can provide a gas-vitreous interface allowing surface tension to reseal torn retinas [3]. Scleral buckles have also long been used to indent the eye near the location of retinal separation. The physical mechanism of this procedure is unclear, but is thought to involve hydrodynamic forces as the vitreous flows over the indentation [8]. Finally, photocoagulation or cryopexy can be used to induce microinjuries, allowing the result- 2 ing scar tissue to help re-adhere the tissue [2, 3]. Many material properties of retinal tissue have been experimentally quantified. For example, the hydraulic conductivities of the layers of the retina have been measured by successive laser ablation of the layers [5]. Adhesion energies between the RPE and photoreceptor cells have been determined by classic pulling experiments [9, 10]. Flows across various tissues, especially the active RPE pump flux, have also been measured [11]. Although these measured physical properties have been known to qualitatively influence retinal stability [12, 13], how they conspire to influence retinal stability has not been quantitatively explored. Given the wide variety of tissue types, causative factors, locations of delamination, and treatment protocols, a mathematical model of the underlying physical mechanisms associated with delamination is desirable. Here, we build and analyze a theoretical model for retinal delamination that incorporates the venous pressure in the choriocapillaris Pc , the intraocular pressure PIOP (both measurable and to some degree, controllable quantities), the hydraulic conductivity of the cell layers Lk , active fluid pumping, and the intrinsic maximum yield stress of the cellular adhesion bonds holding the layers together (see Fig. 3). First, consider the molecular forces holding adjacent tissue layers together in the absence of fluid flow and external traction forces. In the absence of fluid flow, we define σk∗ as the maximum separation force per area that can be sustained between tissue layers k and k − 1. In equilibrium, externally applied normal forces lifting the retina off the choroid (labeled T in Fig. 3(b)) act throughout all layers and the effective yield stress between layer k and k − 1 becomes σk∗ − T . If the remaining stress arising from fluid flow and exceeds σk∗ , tissue layers k and k − 1 will spontaneously delaminate. We next include fluid flows and associated hydraulic forces on the separate tissue layers. Figure 3 labels all the relevant pressures and hydraulic conductivities of the main tissue layers. The pressure-driven volume flux Jk across tissue layer k is given by PIOP L10 L8 L4 L9 L7 L5 L3 L6 T J { { { { { { P9 P8 P6 {111111111111111111111111111111 000000000000000000000000000000 Jp 000000000000000000000000000000 111111111111111111111111111111 P4 { 1111111111111111 0000000000000000 0000000000000000 { 0000000000000000 1111111111111111 L2 1111111111111111 { L1 P5 P3 P2 P0 L0 P7 P1 Pc FIG. 3: Tissue layers of an attached retina (not to scale), with the hydraulic conductivity of each layer labeled Li , and the interlayer fluid pressures denoted Pi . The net flux J in the inward direction and the outward RPE pump flux Jp are indicated by the long blue and short green arrows, respectively. The most common type of delamination occurs between the photoreceptor layer (L4 ) and the RPE (L3 ), labelled with the extracellular pressure P3 . “RPE detachments” can also occur at the interface between the RPE and Bruch’s membrane (L2 ), labelled with pressure P2 . the only tissue that actively pumps fluid, the total inward volume flow across it, instead of (1), is now Jn = Ln ∆n − Jp (∆n ), (2) where the metabolically-driven outward pump flux Jp depends on both the physiological state of the RPE cells and possibly the intercellular pressure difference ∆n = Pn−1 − Pn across the RPE. At steady-state, fluid incompressibility requires that the total volume flux across each layer is constant. Summing (1) and (2) over all k and imposing Jk = J = constant, we find the relation ∆P ≡ Pc − PIOP = J N X Rj + Rn Jp (∆n ), (3) j=0 Jk = Lk (Pk−1 − Pk ) ≡ Lk ∆k , 0 ≤ k ≤ N, (1) where P−1 ≡ Pc , PN ≡ PIOP , Lk is the hydraulic conductivity of tissue or cell layer k, and ∆k = Pk−1 − Pk is the local pressure difference in the intercellular space across layer k. By treating the flow resistance of each layer in series, we can relate the total current J with a known pressure difference ∆P ≡ Pc − PIOP . Besides passive, pressure-driven flows, tissues can actively pump fluid across cell layers. In the eye, the RPE monolayer of cells underneath the retinal photoreceptors actively pumps fluid outward through the Bruch’s membrane back into the choriocapillaris. If the RPE is defined as the nth structural layer (n = 3 in Fig. 3(b)), and is where the resistance of layer j is Rj ≡ 1/Lj . The total inward volume flux toward the vitreous space is thus J= ∆P − Rn Jp (∆n ) , RT (4) PN where RT ≡ j=0 Rj is the total hydraulic resistance. Typically, ∆P ≃ 5 − 10mm Hg [15] so that in the absence of active pumping, there is a passive inward volume flow. Substituting (4) in ∆k = Rk J (Eq. 1), we find the pressure difference across the k th layer: ∆k = Rk [∆P − Rn Jp (∆n )] , RT k 6= n. (5) 3 For the pressure difference ∆k=n across the active RPE layer, we use J in (2) to find ∆n = Rn Rn > 0. ∆P + Rn Jp (∆n ) 1 − RT RT (6) Eqs. (5) and (6) represent the flow-induced stresses pulling apart layer k from layer k − 1. By comparing ∆k against the total yield stresses σk∗ − T between layers, we determine the conditions for retinal stability. For example, layer j is unstable to separation from layer j − 1 as soon as ∆j > σj∗ − T . Since at equilibrium, the vitreous traction force −T is transduced equally through all the layers, increasing external pulling forces alone will not change which layer first becomes unstable. However, layer-specific changes in adhesion energy, hydraulic conductivity Lj , or RPE pump function can control which layer first becomes unstable. The stabilizing influence of the RPE pump flux on all interfaces other than the RPE-Bruch’s membrane junction is evident from (5), where Jp > 0 decreases the pressure difference tending to separate layer k from layer k − 1. In fact, ∆P − Rn Jp can become negative for sufficiently large RPE resistance Rn and pump flux Jp , giving rise to an net outward volume flux and a hydraulic compression of all layers k 6= n. While a leaky RPE with low resistance Rn can short-circuit the pump flux and diminish the stabilizing effect of the pumping, the overall magnitude of the stress on layer k is maximal when its own conductivity vanishes. Here, the entire hydraulic stress ∆P −Rn J0 head is supported by layer k. However, from (6), any pump flux Jp > 0 will tend to destabilize the photoreceptor-RPE interface k = n. In detachments where the retina first detaches from the RPE, the stress ∗ − T while ∆k < σk∗ − T for k 6= n + 1. ∆n+1 > σn+1 By specifying a functional form for Jp (∆n ), we can self-consistently solve (6) to find an explicit solution to ∆n and determine all fluxes and pressures as functions of known parameters. Since the RPE is highly metabolic, its pump activity may be insensitive to the pressure difference across the cell layer. The active flux Jp (∆n ) across isolated dog RPE at different trans-RPE pressures has been shown to be insensitive to a wide range of pressures ∆n < 0 [11]. However, other measurements show that the in vitro pump flux across isolated RPE is significantly greater than that across the intact, in vivo RPE, where ∆n is probably positive. Given these observations, we propose a piece-wise linearly decreasing pump flux Jp (∆n ): Jp (∆n ) = 1− 0, ∆n ∆∗ n J0 , ∆n ≤ ∆∗n (7) ∆n > ∆∗n where ∆∗n is the maximum pressure against which the RPE can pump fluid, and J0 is the pump amplitude when ∆n = 0. Other simple forms such as Jp ∝ e−∆n or one where Jp vanishes above a threshold value of ∆n can also be used but will yield qualitatively similar results. Using (7) in (6) we find an explicit expression for ∆n > ∆∗n , ∆n = ∆∗n (1 − (1 Rn Rn RT )Rn J0 + RT ∆P Rn )Rn J0 + ∆∗n −R T . (8) Using (8) in (7) yields an expression for Jp which can be used to explicitly find ∆k6=n : ∆k6=n = Rk ∆∗n (∆P − Rn J0 ) + Rn J0 ∆P Rn RT )Rn J0 + ∆∗n (1 − R (9) T From (8), if ∆∗n < Rn ∆P/RT (∆n > ∆∗n ), the RPE pump flux vanishes and ∆j = Rj ∆P/RT . Note that for strong RPE pumps such that ∆∗n ≫ Rn ∆P/RT , ∆n+1 < ∆n , and retinal-RPE detachments are likely to be favored over RPE-Bruch’s membrane delaminations ∗ (“RPE detachments”) only if σn+1 < σn∗ . The stresses ∆k are functions of the parameters Rj , ∆P, J0 , and the threshold ∆∗n . Although tissues from different animals, measured using different experimental protocols display a wide range of parameter values, estimates are available. The entire retina, excluding RPE and choroid has a hydraulic conductivity of approximately 3 × 10−10 m/(sPa) [5], with the inner and outer plexiform layers providing nearly all of the resistance. Another measurement of dog RPE-choroid gave a conductivity of ∼ 1.7 × 10−11m/(sPa), and an intrinsic RPE pump flux J0 ∼ 2 × 10−8 m/s [11]. For ∆P = 5mm Hg ≈ 650Pa, the active pump flux J0 & LT ∆P ≈ 1.7 × 10−8 , indicating that at this pressure difference, the active RPE pumping slightly over-compensates the passive, pressure-driven inward volume flow. The combined resistance of the Bruch’s membrane-choroid tissue in humans decreases with age from ∼ 10−8 m/(sPa) to ∼ 10−10 m/(sPa). Using these experimental observations, we estimate that under typical conditions rn ≡ Rn /RT ∼ 0.75 − 0.95, rn−1 ≡ Rn−1 /RT ∼ 0.001 − 0.2. We now explore the features of the normalized hydraulic stresses ˜ n ≡ ∆n /(∆P ) and ∆ ˜ n+1 ≡ ∆n+1 /(∆P ) with these ∆ parameter estimates in mind. ˜ n and ∆ ˜ n+1 /rn+1 (rj ≡ Rj /RT ), the Figure 4 plots ∆ hydraulic stresses pulling apart the RPE-Bruch’s membrane and photoreceptor-RPE interfaces, respectively, as ˜ ∗ ≡ ∆∗ /(∆P ) functions of the dimensionless variables ∆ n n ˜ and J0 ≡ RT J0 /(∆P ). Since an active RPE pump evacuates the photoreceptor-RPE space by pumping fluid into the space between the RPE and Bruch’s membrane, increasing the pump rate J˜0 destabilizes the RPE-Bruch’s ˜ n ) and stabilizes the membrane interface (increases ∆ ˜ n+1 ). photoreceptor-RPE interface (decreases ∆ Age-related decreases in the hydraulic conductivity of the Bruch’s membrane (Ln−1 = L2 in Fig. 3) have 4 ˜ n ≡ ∆n /(∆P ) FIG. 4: (a) Normalized hydraulic stress ∆ tending to separate the RPE from Bruch’s membrane, plotted as functions of the normalized pump threshold ∆∗N /(∆P ) and J˜0 ≡ RT J0 /(∆P ). (b) The normalized hydraulic stress ˜ n+1 /rk+1 separating the photoreceptor-RPE interface. The ∆ flat portions of both plots correspond to regions where rn ∆P > ∆∗n and Jp = 0. The value rn ≡ Rn /RT = 0.8 as assumed in both plots. been hypothesized to block resorption of fluid into the choroid and increase the incidence of RPE detachments in elderly patients [16]. In Fig. 5(a), we display the ˜ n as a function of the normalized, indebehavior of ∆ pendent hydraulic resistances of the RPE and Bruch’s membrane, fn ≡ Rn /R′ and fn−1 ≡ Rn−1 /R′ (where R′ ≡ RT −Rn −Rn−1 ), respectively. Contrary to this hypothesis, our model shows the effect of increasing fn−1 on ˜ n is more subtle. Fig. 5(a) shows, for reasonable values ∆ ˜ n as a function of fn = Rn /R′ and fn−1 = FIG. 5: (a) ∆ Rn−1 /R′ (R′ ≡ RT − Rn − Rn−1 ), the normalized RPE and ˜ n+1 /fn+1 Bruch’s membrane resistances, respectively. (b) ∆ is largely negative, indicating that the active pump flow compresses the photoreceptor layer to the RPE. Only for small ˜ n+1 > 0, allowing the photoreceptors RPE resistance fn is ∆ to possibly spontaneously separate from the RPE. Stresses are ˜ ∗n = ∆∗n /(∆P ) = 3 and R′ J0 /(∆P ) = 1. plotted for fixed ∆ ˜ ∗n = 3 and R′ J0 /(∆P ) = 1, very little destabilizaof ∆ tion of the RPE-Bruch’s interface occurs as the Bruch’s membrane resistance fn−1 is increased. While increasing fn−1 hinders fluid pumped under the RPE from resorbing back into choroid, it also blocks most of the capillary pressure Pc from being transduced past the Bruch’s membrane. Therefore, the dominant destabilizing factor is an increase in the normalized RPE resistance fn allowing significant pressure to build underneath the RPE. In contrast, the photoreceptor-RPE interface is destabilized when the pressure can build up underneath the photoreceptor layer. Therefore destabilization always occurs when the RPE becomes leakier (decreasing fn ), allowing pumped fluid to easily reenter the photoreceptorRPE space. Increasing the resistance of Bruch’s membrane fn−1 hinders fluid from reentering the choroid, but if fn is large, the RPE provides the main resistance to pressure build-up underneath the photoreceptor cells due to passive flows. Therefore, for tight RPE layers, an increase in fn−1 has a relatively greater effect in destabilizing the photoreceptor layer from the RPE than in destabilizing the RPE-Bruch’s membrane inter˜ n and ∆ ˜ n+1 increase with fn−1 face. Specifically, both ∆ ∗ ∗ ˜ ˜ ˜ when fn J0 > ∆n /(∆n − 1), but otherwise decreases with decreasing Bruch’s membrane flow resistance fn−1 . Our theoretical framework for retinal delamination is analogous to a simple series of resistors bridging a voltage drop, but with an interior (possibly voltage gradientdependent) current source representing the RPE pumps. Voltage gradients describe the amplitudes of local electric fields that contribute to dielectric breakdown. Although we have only considered the effects of RPE and Bruch’s membrane conductivities on photoreceptor and RPE delaminations, our model can be used to address other possible physiological causes of detachment. For example, inflammation of choroidal tissue and drugs such as acetazolamide that reduce it [14], can be modeled by increasing and decreasing the tissue permeabilities L0 or L1 . Small blood vessels that invade the retina can also be described by imposing a venous pressure within a specific tissue layer. These simple adalptations allow our model to address numerous clinical scenarios associated with retinal delamination. This work was supported by the Gerald Oppenheimer Family Foundation, the Army Research Office through grant 58386MA, and the NSF through grants DMS1032131 and DMS-1021818. [1] H. Wässle and B. B. Boycott, Functional architecture of the mammalian retina, Physiol Rev., 71, 447480, (1991). [2] M. Yanoff and J. S. Duker, Ophthalmology, [3] T. H. Williamson, Vitreoretinal Surgery, (SpringerVerlag, Berlin, 2008). [4] H.-P. Hammes and M. Porta, editors, Experimental Approaches to Diabetic Retinopathy, (Karger, Basel, 2010). [5] R. J. Antcliff, A. A. Hussain, and J. Marshall, Arch. Ophthalmol., 119, 539-544, (2001). [6] D. J. Moore, A. A. Hussain, and J. Marshall, Invest. Ophthalmol. & Vis. Sci., 36, 1290-1297, (1995). [7] A. C. Bird and J. Marshall, Trans. Ophthalmol. Soc., 105, 674682, (1986). [8] W. J. Foster, N. Dowla, S. Y. Joshi and M. Nikolaou, 5 [9] [10] [11] [12] Graefe’s Arch. Clin. and Exptl. Ophthalmol., 248, 31, (2009). D. M. Maurice, J. Salmon, and H. Zauberman, 12, 212217, (1971). H. Zauberman and E. R. Berman, Exptl. Eye Res., 8, 276-283, (1969). S. Tsubai, Invest. Ophthalmol. & Vis. Sci., 28, 17761782, (1987). I. Fatt and K. Shantinath, Exp. Eye Res. 12, 218-226, (1971). [13] A. Negi, S. Kawano and M. F. Marmor, Invest Ophthalmol Vis Sci., 28, 2099-2102, (1987); M. F. Marmor, Graefe’s Arch. Clin. and Exp. Ophthalmol., 226, 548-552, (1988). [14] B. Moldow et al., Graefe’s Arch. Clin. and Exp. Ophthalmol., 236, 881-889, (1998). [15] O. Maepea, Exp. Eye Res., 54, 731-736 (1992). [16] C. Starita, et al., Exp. Eye Res., 62, 565-572, (1996).