Abstract
Diseases culminating in photoreceptor loss are a major cause of untreatable blindness. Transplantation of rod photoreceptors is feasible, provided donor cells are at an appropriate stage of development when transplanted. Nevertheless, the proportion of cells that integrate into the recipient outer nuclear layer (ONL) is low. The outer limiting membrane (OLM), formed by adherens junctions between Müller glia and photoreceptors, may impede transplanted cells from migrating into the recipient ONL. Adaptor proteins such as Crumbs homologue 1 (Crb1) and zona occludins (ZO-1) are essential for localization of the OLM adherens junctions. We investigated whether targeted disruption of these proteins enhances donor cell integration. Transplantation of rod precursors in wild-type mice achieved 949 ± 141 integrated cells. By contrast, integration is significantly higher when rod precursors are transplanted into Crb1rd8/rd8 mice, a model of retinitis pigmentosa and Lebers congenital amaurosis that lacks functional CRB1 protein and displays disruption of the OLM (7,819 ± 1,297; maximum 15,721 cells). We next used small interfering (si)RNA to transiently reduce the expression of ZO-1 and generate a reversible disruption of the OLM. ZO-1 knockdown resulted in similar, significantly improved, integration of transplanted cells in wild-type mice (7,037 ± 1,293; maximum 11,965 cells). Finally, as the OLM remains largely intact in many retinal disorders, we tested whether transient ZO-1 knockdown increased integration in a model of retinitis pigmentosa, the rho-/- mouse; donor cell integration was significantly increased from 313 ± 58 cells without treatment to 919 ± 198 cells after ZO-1 knockdown. This study shows that targeted disruption of OLM junctional proteins enhances integration in the wild-type and degenerating retina and may be a useful approach for developing photoreceptor transplantation strategies.
Introduction
Most retinal degenerative diseases culminate in the loss of photoreceptors. Cell transplantation offers a strategy to replace these cells and restore visual function. Both adult wild-type and diseased retinae are capable of integrating transplanted cells, and these cells develop unambiguous characteristics of mature photoreceptors (3, 21, 24). By using a fluorescent marker linked to the promoter sequence of Nrl, a transcription factor expressed in rods shortly after terminal mitosis (2), we showed that the cells that possess this capacity to migrate and functionally integrate are postmitotic rod photoreceptor precursors, rather than stem/progenitor cells (24). Thus, rod photoreceptor transplantation is feasible, but the developmental stage of the donor cell appears critical to its success. Moreover, transplantation efficiency is low. Migration of transplanted murine neural progenitor cells occurs into the immature opossum retina, but is lost with increasing maturity (34, 49), while in the adult retina, extensive migration is restricted to severely degenerated retina (55) or to areas of significant physical disruption (15, 56). This suggests that barriers exist within the recipient retina that prevents greater numbers of cells from integrating (40). The outer limiting membrane (OLM) may be one such barrier.
The OLM comprises a series of zonula adherens junctional complexes between photoreceptors and Müller glia (Fig. 1A) and is first discernible by postnatal day (P) 5 in the mouse (47). Adherens junctions are transmembrane cadherin–catenin complexes that mediate cell–cell adhesion. These junctions interact with a cytoplasmic plaque comprised of adaptor proteins, including, among others, Crumbs homologue 1 (Crb1) and zonula occludins 1 (ZO-1), which is essential for their maintenance. Pharmacological interference or genetic disruption of a number of components of the adherens junction and/or the supporting scaffold leads to significant impairment in OLM integrity [see (48, 50)].
Integration of photoreceptors transplanted into the Crb1rd8/rd8 retina is higher than in wild-type controls. (A) Left, schematic showing the adherens junction complexes (red) between Müller glia terminal processes and photoreceptor inner segments, and right, some of the proteins involved in adherens junction formation and localization [adapted from (28)]. (B) Schematic of a subretinal injection. (C) Confocal projection montage showing integration into the 6-week-old Crb1rd8/rd8 mouse. Some cells remain in the subretinal space at 3 weeks posttransplantation but large numbers migrate into the ONL. (D) Histogram showing the number of integrated photoreceptors found at 3 weeks posttransplantation into 3-, 6-, or 12-week-old Crb1rd8/rd8 recipients, compared with 6-week-old wild-type recipients. Error bars: ±SEM. p-values: ANOVA with Bonferonni's correction for multiple comparisons. (E, F) Integrated cells were frequently observed in large clusters of 10–30 cells in the Crb1rd8/rd8 mouse (E). This contrasts integration into the wild-type retina, which is usually sporadic with single cells, or small clusters of 2–5 cells (F). Scale bars: 50 μm.
Previously, we showed a twofold increase in the number of integrated cells after intravitreal administration of a pharmacological toxin, alpha-aminoadipic acid (AAA), prior to cell transplantation (51). AAA disrupts Müller glial function and, indirectly, OLM integrity in wild-type mice, suggesting that reduced OLM integrity enhances cell integration. However, AAA administration was recently reported to cause Müller cells to dedifferentiate and migrate into the ONL (43), and our previous study did not exclude the possibility that these or other effects of AAA enhance integration. Here, we directly test the hypothesis that the OLM is a barrier to transplanted cell migration and integration by investigating integration in models with targeted genetic and molecular disruption of specific proteins of the OLM junctional complex. We show that it is possible to induce a transient disruption of OLM integrity using RNA interference. When combined with cell transplantation, these strategies lead to significant increases in cell integration in both wild-type mice and, importantly, in models of retinal degeneration.
Materials and Methods
Animals
C57/Bl6 (Harlan, UK), Nrl.gfp+/+ (A. Swaroop, University of Michigan), Crb1rd8/rd8 (Jackson Laboratory), and rho-/- (P. Humphries, Trinity College Dublin) mice were maintained in the animal facility at University College London. Animals were kept on a standard 12-h light/dark cycle. All experiments have been conducted in accordance with the United Kingdom Animals (Scientific Procedure) Act of 1986.
Dissociation of Retinal Cells and Transplantation
Dissociated cells were prepared from P4 Nrl.gfp+/+ mice. The use of Nrl.gfp mice provides a genetic marker that identifies donor rod photoreceptors. Cells were dissociated using a papain-based kit (Worthington Biochemical, Lorne Laboratories, UK) and resuspended at a concentration of 200,000 cells/μl in sterile EBSS and DNase (0.005%). Surgery was performed as previously described (24). Recipient mice were anesthetised with an intraperitoneal injection of 0.2 ml of a mixture of Dormitor (1 mg/ml medetomidine hydrochloride, Pfizer Pharmaceuticals, Kent UK), ketamine (100 mg/ml, Fort Dodge Animal Health, Southampton, UK), and sterile water for injections in the ratio of 5:3:42. Cells were transplanted via a single injection (1 μl), made at an oblique angle through the superior sclera into the subretinal space, using a sterile 34-gauge hypodermic needle (Hamilton, Switzerland) and injected slowly to produce a standard and reproducible retinal detachment in the superior hemisphere (in the same region as the siRNA injection, where appropriate) (Fig. 1B). The needle was slowly withdrawn, leading to a self-sealing of the wound tunnel.
Preparation and Administration of siRNAs
Targeting siRNA against ZO-1 was generated using the sequence 5′-AAGATAGTTTGGCAGCAAGAG-3′ (Invitrogen, UK), as reported previously (39). The siRNA was resuspended and diluted to the appropriate concentration in sterile buffer containing oligofectamine using RNAase-free plasticware. Controls comprised a 1-μl injection of buffer/oligofectamine containing an equivalent concentration of a proven nontargeting siRNA (All-Stars negative control siRNA; Cat. No. 1027281; Qiagen, UK), or sterile buffer alone (vehicle). siRNAs were introduced by subretinal injection to the superior retina 48 h prior to cell transplantation, unless stated otherwise.
Immunohistochemistry and Histology
Eyes were dissected out, leaving the superior rectus muscle in place to provide a landmark for the superior retina, and placed in 4% (for cell counting) or 1% (immunohistochemistry) paraformaldehyde (PFA) in PBS for 1 h. Eye cups were prepared by removing the front of the eye, including the lens, before being cryoprotected in 20% sucrose/PBS, cryo-embedded in OCT (TissueTek), and cut as transverse sections 20 μm thick.
Retinal cryosections were air-dried for 15–30 min and washed in Tris-buffered saline (TBS). For immunohistochemistry, sections were preblocked in TBS containing normal goat serum (2%), bovine serum albumin (1%), and 0.05% Triton X-100 for 1 h before being incubated with primary antibody overnight at 4°C. After rinsing with TBS, sections were incubated with secondary antibody for 2 h at room temperature (RT), rinsed, and counterstained with Hoechst 33342. Negative controls omitted the primary antibody. The following antibodies were used: pancadherin (1:500; rabbit polyclonal; Sigma, UK); ZO-1 (1:500; rabbit polyclonal; Zymed); β-catenin (1:2000; rabbit polyclonal, Abcam, UK); GFAP (1:500; rabbit polyclonal; DAKO); Crb1 (1:100; kind gift of Jan Wijnholds).
Confocal Microscopy
Retinal sections were viewed on a confocal microscope (Zeiss LSM510 or Leica SP2), as previously described (24). Images show either single confocal sections or merged projection images of an xyz confocal stack through retinal sections, approximately 10 μm thick.
Electron Microscopy
Mice were sacrificed 48 h after siRNA administration. The eyes were fixed overnight (3% glutaraldehyde/1% PFA) and then the cornea and lens were removed and the eye cups orientated and processed, as previously described (45, 51). Ultrathin sections were collected on copper grids (100 mesh, Agar Scientific), contrast stained with 1% uranyl acetate and lead citrate and analyzed using a JEOL 1010 Transmission Electron Microscope (80 kV), fitted with a digital camera for image capture.
Cell Counts
Counts of integrated cells were taken 3 weeks after transplantation using a fluorescence microscope (Observer Z.1, Zeiss, UK). No substantial changes were observed in the number of integrated cells between 2 and 4 weeks posttransplantation (data not shown). The use of the Nrl.gfp mouse as a source of donor cells provides a genetic marker for the identification of rod photoreceptors (2, 26, 41). Cells were considered to be integrated if the whole cell body was correctly located within the outer nuclear layer, and at least one of the following was visible: spherule synapse, inner/outer processes, inner/outer segments (Fig. 1a). Animals were omitted from analysis if there was clear evidence of an injection having been administered intravitreally, rather than subretinally, or if there was an absence of cells in the subretinal space, which is an indication that reflux of the donor cell suspension had occurred at the time of injection (<5%).
Apoptotic Cell Counts
The total number of apoptotic cells was determined by counting all TUNEL-positive profiles in each layer of the retina in alternate serial sections. Only regions around the site of injection are shown. Apoptosis was minimal or absent in regions away from the site of detachment.
Electrophysiology
ERGs from injected animals were recorded in a standardized fashion, 7 weeks after treatment using an Espion E2 system with a ColourDome stimulator (Diagnosys LLC, Lowell, MA). Animals were dark adapted overnight and Ganzfeld ERGs were obtained simultaneously from both eyes to provide an internal control. Mice were anesthetized as before and the pupils were dilated using one drop of Tropicamide 1%. Viscotears were placed on each cornea to keep it moistened after corneal contact electrodes and midline subdermal reference and ground electrodes were placed. Single flash recordings were obtained at light intensities of 0.001, 0.01, 0.1, 1, 3, 5, and 10 cd s m−1 using a sampling frequency of 5 kHz, a flash duration of 4 ms, and a frequency stimulus of 0.5 Hz. Data were recorded from 10 ms before stimulus onset to 400 ms poststimulus. For the first three intensities 10 responses were averaged to obtain the final ERG, whereas for the last two intensities 5 responses were averaged. The bandpass filter was set between 0 and 1 kHz. For ERG analysis, the a- and b-wave values (a-wave trough to b-wave peak) of the treated eyes were paired with the untreated contralateral eyes to provide an internal control. This method controls for interanimal variance.
Statistics
All means are stated ±SEM, unless indicated otherwise. N = number of eyes and n = number of sections examined, where appropriate. All experiments concerning siRNA injections were conducted as paired experiments, whereby the contralateral eye served as a within-animal control. Statistical significance was assessed using Graphpad Prism 5 software, and applying paired t-test, ANOVA with Bonferonni's correction for multiple comparisons, where appropriate.
Results
Increased Integration of Rod Photoreceptors in the Crb1rd8/rd8 Mouse, a Model of OLM Disruption
Crb1 is a protein essential for the formation and localization of the OLM adherens junction and expressed by Müller glia (27, 48, 50). Mutations in the human CRB1 gene cause retinal diseases including retinitis pigmentosa and Lebers congenital amaurosis (23). The Crb1rd8/rd8 mouse has a single base deletion in Crb1, which causes a frameshift and premature stop codon, truncating the protein upstream of the transmembrane domain. Loss of functional CRB1 protein affects OLM integrity causing retinal degeneration (27) and provides a genetic model in which to evaluate the role of the OLM in impeding transplanted cell integration. This naturally occurring mutant was first described by Mehalow and coworkers (27), who noted that the OLM displayed significant fragmentation, as assessed by electron microscopy and immunohistochemistry, apparent from 2 weeks of age.
To determine the impact of OLM disruption in Crb1rd8/rd8 on transplanted rod photoreceptor cell migration and integration, we transplanted unsorted dissociated P4 neural retinal cells from Nrl.gfp+/+ mice, in which GFP expression is driven by the promoter of Nrl, a transcription factor specific for postmitotic rod precursors that persists in adult rods [see (2, 24) for details]. Cells were injected into the subretinal space of 3-, 6-, and 12-week-old Crb1rd8/rd8 mice or 6-week-old wild-type mice. Three weeks postinjection, the animals were sacrificed and the total number of integrated cells determined. Cell integration was significantly increased in both the 3- and 6-week-old Crb1rd8/rd8 recipients, compared with wild-type controls (Table 1, Fig. 1C, D). A maximum of 10,227 and 15,721 integrated cells was observed in 3- and 6-week-old Crb1rd8/rd8 mice, respectively. This contrasts with a maximum number of 1,078 cells in 6-week-old wild-type mice. However, donor cell integration was only twofold higher than wild-type controls following transplantation into 12-week-old Crb1rd8/rd8 recipients, a time when GFAP upregulation is maximal (of the time points studied here).
Photoreceptor Integration in Crb1rd8/rd8 Mice
Number of integrated photoreceptors found at 3 weeks posttransplantation into 3-, 6-, or 12-week-old Crb1rd8/rd8 recipients, compared with 6-week-old wild-type recipients, ±SEM. N = number of eyes.
Previously, we had observed that in wild-type animals integrated cells are often found in small columns or clusters. While having the appearance of clonal clusters of cells, these groups do not arise from proliferation of a single donor stem or progenitor cell (24). A possible explanation is that cells migrate into the retina at specific locations, such as areas of OLM disruption or at the site of specific cell types. Upon further examination of the pattern of integration occurring in the Crb1rd8/rd8 mice, it was notable that integrated cells were frequently located in large clusters, often of 20–50 cells and sometimes hundreds in a group (Fig. 1E). This was in contrast to the sporadic integration seen in the wild-type retina, where cells are usually found singularly or in small clusters of only 2–5 cells (Fig. 1F). The spread of integrated cells was not qualitatively significantly different from that observed in wild-type mice (i.e., the majority of cells integrating did so around the site of cell injection).
As increased integration levels were seen at earlier stages in the Crb1rd8/rd8 mice but declined by 12 weeks, we used immunohistochemistry to assess the level of OLM disruption, ONL disorganization, and glial cell activation at each time point (3, 6, and 12 weeks of age). Glial cell activation is a feature of the degenerating retina and has been proposed to impede cell migration (13, 31). In the wild-type mouse, Crb1 and a second adapter protein, ZO-1, are visualized as a continuous line at the outer margin of the neural retina (Fig. 2A, B, far right). Conversely, in the neural retina of 3-week-old Crb1rd8/rd8 mice, CRB1 expression was no longer correctly located at the OLM, consistent with loss of the transmembrane domain of the nonfunctional protein (Fig. 2A), while that of ZO-1 was discontinuous and fragmented (Fig. 2B). ZO-1 expression became increasingly disorganized with advancing age, with significant gaps in expression at the OLM by 6 weeks (Fig. 2B). Similarly, pancadherin, which labels cadherin cell–cell adhesion molecules, is strongly expressed at the OLM in wild-type mice (Fig. 2C, far right) but became progressively more diffuse and disorganized over time in the Crb1rd8/rd8 mouse (Fig. 2C). Over the same time period, GFAP, which is expressed by activated glial and astrocyte populations, revealed an increasing level of glial cell activation in the Crb1rd8/rd8 mice (Fig. 2D, E). In wild-type mice, GFAP-positive processes were seen only occasionally (Fig. 2D, E, far right), and then were largely limited to the peripheral margins of the retina. A similar pattern was observed in the 3-week-old Crb1rd8/rd8 mice. By 6 weeks GFAP expression was increased and GFAP-positive fibers could be seen extending throughout the thickness of the neural retina. Significantly, by 12 weeks GFAP was extensively expressed throughout the retina (Fig. 2E). Furthermore, Müller cell hypertrophy, seen as processes extending along the outer edge of the ONL (22), was frequently observed (Fig. 2D, insert). Thus, the abnormal OLM in the Crb1rd8/rd8 mouse leads to a progressive disorganization of the ONL, together with increasing Müller glial cell activation, with age, which is widespread by 12 weeks. These data suggest that cell integration at 12 weeks may be adversely affected by the marked glial cell activation. In summary, lack of CRB1 and an abnormal OLM enhance photoreceptor cell integration, but only at early stages of degeneration prior to widespread glial cell activation.
The abnormal OLM in the Crb1rd8/rd8 mouse leads to a progressive disorganization of the ONL and increasing Müller glial cell activation with age. (A–E) Immunohistochemistry showing confocal projection images of staining for proteins associated with adherens junctions, Crb1 (A) and ZO-1 (B), the cell adhesion marker, pancadherin (C), glial fibrilliary acidic protein (GFAP) (D, E), which labels activated astrocytes and glia and vimentin (E), which labels Müller glia regardless of activation (37). Staining is shown in 3-, 6-, and 12-week-old Crb1rd8/rd8 mice and 6-week-old wild-type mice. Insets in (A) and (B), single confocal sections from 6-week-old Crb1rd8/rd8 and wild-type mice showing the fragmented staining pattern (arrows) of ZO-1 and Crb1 in the Crb1rd8/rd8 mouse. Inset in (D), GFAP staining shows GFAP-positive fibers extending along the edge of the ONL at 12 weeks. GFAP staining increased markedly between 6 and 12 weeks. All images shown are taken from the superior retina. Scale bars: 50 μm and 10μm (insets).
The OLM Can Be Transiently Disrupted Using siRNAs Targeted Against a Component of the Adherens Junction Complex
As cell integration is significantly increased when precursors are transplanted into the Crb1rd8/rd8 mouse, a model in which the integrity of the OLM is compromised due to lack of CRB1 protein, we next sought to determine if the same effect was observed if an alternative protein involved in adherens junction formation was knocked down. In order to minimize any secondary effects on retinal structure that arise following OLM disruption, we designed a strategy for transiently and reversibly disrupting the OLM in the adult retina. A number of reports have demonstrated that siRNAs can produce a knockdown of their target proteins in the eye (6, 32, 38, 44). Here, we used siRNA targeted against ZO-1, one of the adaptor proteins involved in adherens junction formation, and one whose expression was affected in the Crb1rd8/rd8 mouse.
In order to establish the optimum dose of siRNA required for a transient downregulation of ZO-1 in vivo, several doses (10, 15, and 20 μM) of either ZO-1 siRNA or a nontargeting control siRNA were administered either via subretinal or intravitreal injection to adult (6–8 weeks old) wild-type mice. The retinae were then examined via immunohistochemistry at 48 h postinjection (Fig. 3). Intravitreal administration failed to affect ZO-1 expression at the OLM (data not shown). Similarly, subretinal injection of 10 μM siRNA failed to knock down ZO-1, as demonstrated by continuous ZO-1 immunostaining at the OLM (Fig. 3A). Conversely, ZO-1 staining was discontinuous around the site of detachment following application of 15 μM siRNA (Fig. 3B), and largely abolished following application of 20 μM siRNA (Fig. 3C, D). ZO-1 expression remained intact in the opposite hemisphere, away from the injection site (data not shown).
ZO-1 siRNA knocks down ZO-1 expression in the OLM. (A–C) Confocal images showing ZO-1 staining of the OLM following subretinal injection of 10 μM (A), 15 μM (B), or 20 μM (C). (D) Confocal montage showing extent of ZO-1 knockdown following application of 20 μM siRNA. Limits are marked by arrows. Significant OLM disruption was seen following application of the 15 μM dose, and knockdown was extensive with the 20 μM dose. (E) Apoptosis counts at 48 h postinjection in eyes receiving 20 or 15 μM ZO-1 siRNA, and over time in those receiving 15 μM siRNA, compared with nontargeting controls. (F) Electron micrographs showing disruption of the OLM 48 h after application of 15 μM ZO-1 siRNA or nontargeting control siRNA. Adherens junctions are highlighted by arrows. (G–I) Confocal images showing immunostaining for crb1 (G), pancadherin (H), and β-catenin (I), 48 h after application of 15 μM ZO-1 siRNA or nontargeting control siRNA. Wild-type sections are also included for comparison (left). Scale bars: 50 μm and 5 μm (F).
While the highest dose of siRNA was the most effective at reducing ZO-1 expression, it was also frequently associated with significant levels of cell death in the ONL 48 h after administration, compared with nontargeting siRNA-injected controls (Table 2, Fig. 3E; p < 0.05 ANOVA with Bonferroni's correction). By contrast, application of 15 μM ZO-1 siRNA did not lead to a significant increase in cell death (p > 0.05) (Table 2). The detachment process itself, made with vehicle alone, is also associated with a degree of apoptosis, similar to that seen following injection of either 15 μM ZO-1 siRNA or nontargeting siRNA (p > 0.05) (Table 2). In each case, apoptotic cells were evenly distributed across the ONL. The level of apoptosis fell significantly by 4 weeks postinjection, suggesting that the apoptosis observed was transient and not indicative of an induced degeneration. ERG recordings were taken at 7 weeks postinjection. There was no significant difference in either a- or b-wave amplitude in eyes treated with ZO-1 siRNA compared with either nontargeting or uninjected controls (Fig. 4A, B). Thus, ZO-1 siRNA application does not appear to have a detrimental effect on visual function.
Retinal function is not impaired by ZO-1 siRNA administration. (A) Histogram showing average a- and b-wave amplitudes at a flash intensity of 0.1 cd s m−1. (B, C) ERG intensity series at 8 weeks following subretinal injection of either 15 μM ZO-1 siRNA (B, left) or 15 μM nontargeting control siRNA (C, left), compared with uninjected contralateral control eyes (A, right and B, respectively). ERG traces shown in each chart were recorded at intensities of 0.001, 0.01, 0.1, 1, 3, 5, and 10 cd s m−1. Error bars: ±SEM.
TUNEL Labeling Following siRNA Application
Number of TUNEL-positive cells per eye at 48 h and 4 weeks postinjection of either ZO-1 siRNA, nontargeting control siRNA, or vehicle alone, ±SEM. N = number of eyes.
We next sought to confirm that knockdown of ZO-1 using subretinal injection of 15 μM ZO-1 siRNA conferred disruption to the OLM itself. Electron microscopy images show that at 48 h post-siRNA administration, the OLM was intact in retinae treated with the nontargeting control siRNA (Fig. 3F, arrows, top). Conversely, OLM integrity was lost in many regions around the injection site of the ZO-1 siRNA-treated eyes, with significant distances between the remaining adherens junctions (Fig. 3F, arrows, bottom). These findings are further supported by the reduced or disrupted expression of pancadherin (Fig. 3G), β-catenin (Fig. 3H), and Crb1 (Fig. 3I), in ZO-1 siRNA-treated eyes, compared with only limited disturbance following injection of nontargeting control siRNA or vehicle alone.
We next established the time course of ZO-1 knockdown to determine the optimal time for transplanting cells following OLM disruption (Fig. 5). Following subretinal injection of 15 μM ZO-1 siRNA, ZO-1 expression was reduced between 48 and 72 h postinjection, but had largely recovered by 1 week post-siRNA injection (Fig. 5A). Pancadherin (Fig. 5D) presented a similar pattern of disruption and recovery over the same time course. The introduction of vehicle either with or without nontargeting control siRNA causes a transient detachment of the neural retina, but had little obvious effect on OLM integrity, as assessed by ZO-1 (Fig. 5B, C) and pancadherin (Fig. 5E, F) expression. Occasionally, localized disruption was seen at 48 h (Fig. 5B, insert, arrow) both in nontargeting and vehicle-injected retinas, but this had largely resolved by 1 week postinjection. GFAP expression was increased compared to uninjected retinas but was similar in each of ZO-1 siRNA-treated, nontargeting-treated, and vehicle-treated eyes (Fig. 5G–I). Thus, these changes are most likely due to the physical disruption caused by the subretinal detachment itself. These data show that ZO-1 siRNA treatment can achieve a transient, reversible disruption of the OLM in the adult wild-type mouse while maintaining recipient retina integrity.
Time course of OLM disruption following subretinal injection of ZO-1 siRNA. Confocal images showing staining for ZO-1 (A–C), pancadherin (D–F), and GFAP (G–I) at 48 and 72 h and 1 week postinjection of 15 μM ZO-1 siRNA (A, D, G), 15 μM nontargeting control siRNA (B, E, H), or vehicle (C, F, I). Scale bar: 50 μm. Insets show high-power images of single confocal sections of ZO-1 or pancadherin immunolabeling, where appropriate. Arrowheads depict level of OLM. Scale bar: 10 μm.
Increased Rod Photoreceptor Integration Following OLM Disruption by ZO-1 Knockdown in the Wild-Type Mouse
Having determined the time course of OLM disruption, we examined the impact of a transient downregulation of ZO-1, and subsequent disruption of the OLM, on cell integration. Wild-type animals received a 1-μl injection to the superior retina containing either 10, 15, or 20 μM siRNA, 48 h prior to the injection of dissociated P4 neural retinal cells (200,000 unsorted cells/μl) to the same region. The contralateral eye received a control injection of nontargeting siRNA, as before, followed by an equivalent cell injection 48 h later. Three weeks after cell transplantation, eyes were collected and the retinae were assessed for cell integration (Fig. 6). All animals receiving either the 20 or 15 μM dose of ZO-1 siRNA demonstrated a significant increase in integration, compared both with the contralateral, nontargeting siRNA-treated eye and standard control injections (no preinjection) (Fig. 6A–C, Table 3). Cell integration was unchanged by application of 10 μM siRNA compared both with the corresponding dose of nontargeting siRNA and with wild-type controls (Fig. 6A). The maximum number of integrated cells observed following siRNA treatment was 11,965, seen following application of 20 μM ZO-1 siRNA. Those eyes receiving nontargeting siRNA (all doses) had significantly higher levels of integration (approximately twofold) than was seen in a standard injection in which the animal received no pretreatment (1,842 ± 131 vs. 949 ± 141 cells, combining all nontargeting data sets; N = 15/N = 13; p < 0.05, unpaired t-test). We have also found a similar enhancement when combining predetachments made with vehicle alone with subsequent cell transplantation, suggesting that the predetachment process itself leads to a modest but significant increase in integration (manuscript in preparation). Of note also, is that the pattern of cell integration in the ZO-1 siRNA-treated animals is similar to that observed in the Crb1rd8/rd8 mouse; cells frequently integrated in clusters, perhaps correlating with areas of OLM disruption (Fig. 6D). Together these data show that loss of two different proteins essential for correct adherens junction localisation, CRB1 and ZO-1, enhances donor cell integration into the ONL.
Enhanced integration of transplanted photoreceptors following ZO-1 siRNA treatment. (A) Histogram showing the number of integrated rod photoreceptors found at 3 weeks posttransplantation of unsorted P4 Nrl.gfp+/+ donor cells into either 6-week-old wild-type recipients or 4-week-old rho-/- recipients, pretreated with ZO-1 siRNA (gray bars), compared with contralateral control eyes treated with equivalent concentrations of nontargeting control siRNA (white bars), and standard wild-type or rho-/- controls (no preinjection; black bars). Error bars: ±SEM. p-values: regular type = ANOVA with Bonferonni's correction for multiple comparisons; italic type = paired t-test. (B, C) Confocal projection montage showing integration of unsorted P4 Nrl.gfp+/+ donor cells into a 6-week-old wild-type mouse treated with nontargeting control siRNA (B) in one eye and ZO-1 siRNA in the other (C). (D) Integrated cells were frequently observed in large clusters of 10–30 cells in the ZO-1 siRNA-treated eyes.
Photoreceptor Integration in Wild-Type and rho-/- Mice Following OLM Disruption
Number of integrated photoreceptors found at 3 weeks posttransplantation into wild-type or rho-/- recipients, receiving either ZO-1 siRNA or nontargeting siRNA, compared with control recipients, ± SEM. N = number of eyes. Bold type indicates significant increases, compared with the appropriate contralateral nontargeting siRNA-treated controls, paired t-test.
Increased Rod Photoreceptor Integration Following OLM Disruption by ZO-1 Knockdown in the Rhodopsin Knockout Mouse
Next we sought to determine whether or not such findings translate to the degenerating retina. The rho-/- mouse is a moderately severe model of rod photoreceptor degeneration with almost all rods lost by 3 months of age (17). In contrast to previous reports (4, 5), we found the OLM to be intact even at late stages of degeneration (Fig. 7). Expression of ZO-1, Crb1, and pancadherin were as robust as that seen in wild-type (Fig. 7A-C). Moreover, assessment at the EM level revealed the adherens junctions to be intact (Fig. 7E). GFAP was also extensively upregulated, as is typical of degenerating retinas (Fig. 7D). We therefore examined the impact of OLM disruption on cell integration in this model. One-month-old rho-/- animals received 15 μM ZO-1 siRNA in one eye and nontargeting control siRNA in the contralateral eye 48 h prior to the injection of cells to the same region. All ZO-1 siRNA-treated animals demonstrated a significant increase in integration, compared with both the contralateral, nontargeting siRNA-treated eye and standard rho-/- control injections (no preinjection) (Fig. 6A, Table 3).
The OLM is intact in the 4-week-old rho-/- mouse. (A–D) Confocal images showing staining for ZO-1, Crb-1, pancadherin, and GFAP in the 1-month-old rho-/- mouse. Scale bar: 50 μm. (E) Electron micrograph of the OLM in the 3-month-old rho-/- mouse. Adherens junctions are highlighted by arrows. Scale bars: 5 μm and 1 μm (inset).
Discussion
Cell transplantation for the treatment of neurodegenerative diseases is an area of intense current investigation. However, transplantation into the mature CNS faces a major hurdle—the limited ability of transplanted cells to survive, migrate, and establish morphological and synaptic connectivity with the recipient tissue. There has been considerable interest in trying to generate specific cell types for transplantation, but for these findings to be successfully translated into clinical therapies we also require a much better understanding of how the donor cells interact with the recipient microenvironment. A number of reports have demonstrated the potential for a variety of cell types to migrate into the immature retina (35, 42, 49). However, this ability is largely lost as the recipient animal matures, whereupon stem cell populations appear able to migrate only when the recipient retina is damaged in some way (20, 29, 49). Here, we provide strong evidence demonstrating that the OLM of the recipient retina is a physical barrier to donor cell migration and integration and presents a potential target for improving transplantation into both the wild-type and, importantly, the degenerating retina. Moreover, we present a novel way of inducing a transient disruption of this barrier that, when combined with cell transplantation, led to significant improvements in integration in models of retinal degeneration.
Recessive mutations in the human CRB1 gene cause retinal diseases including retinitis pigmentosa and Lebers congenital amaurosis (7, 10). Currently, there is no effective treatment for patients with CRB1 mutations. The Crb1rd8/rd8 mouse has a phenotype similar to human patients with missense CRB1 mutations. We found that donor cell migration was significantly improved in these animals compared with wild-type animals and that the level of integration increased with increasing OLM disruption, up to 6 weeks of age. This contrasts with the very low levels of cell integration observed following transplants into other models of retinal degeneration, such as the rho-/- and the Prph2rd2/rd2 (rds) mouse, where typically only a few hundred cells integrate (24). Together these data demonstrate that the OLM presents a physical barrier to transplanted cell migration and integration [and it is important to note that the OLM is likely to be one of a number of such barriers (40)]. Secondly, they show that the degenerative apoptotic environment itself is unlikely to be promoting transplanted cell integration. Thirdly, the Crb1rd8/rd8 mouse provides a relevant model of a specific type of retinal dystrophy, which may be more amenable to cell transplantation therapy than other inherited retinal degenerations.
While retinal dystrophies involving CRB1 may prove ideal candidates for cell transplantation therapy, retinal degeneration caused by many other gene defects, such as rd1 and rds, is not associated with marked OLM disruption until very late stages of the disease (12, 16, 36). This suggests that the OLM might present a significant barrier to transplanted photoreceptor precursor cell migration and integration in the majority of retinal degenerations, particularly at early stages. Therefore, we must seek ways to overcome the physical barrier presented by the OLM. The pharmacological induction of OLM disruption by aminoadipic acid (AAA) we have reported previously (51) would not be suitable due to detrimental side effects within the recipient retina, including very significant apoptosis in both the Müller glia and photoreceptor populations (18, 30, 33). Furthermore, the overall increase in integration, while consistently higher than the vehicle-injected eyes, was only similar to that seen for standard control injections. However, the results presented here suggest that if appropriate methods of OLM disruption are used, significant improvements in cell integration above that previously reported (3, 24) can be achieved. Moreover, such improvements are also possible in models of retinal degeneration.
RNAi strategies have been successfully applied to the treatment of mouse and rat models of hepatitis and cancers (25, 53). A feature of siRNAs that is that they are typically unstable, being removed by endogenous RNases, and so have a short duration of action. The strategy employed here utilized this characteristic and used naked siRNA to demonstrate a proof of principle, namely that the OLM could be reversibly disrupted by siRNAs in both wild-type and degenerating retinas. ZO-1 was the first tight junction protein identified and functions as a junctional adaptor that interacts with multiple transmembrane proteins, components of the junctional plaque and actin filaments (1, 11, 46). Because ZO-1 is also expressed at the tight junctions of the RPE, it is possible that a subretinal injection of ZO-1 siRNA will also impact on these junctions, although we did not observe any obvious signs of leakiness or oedema. For this strategy to be taken forward, it is therefore important to identify other targets for siRNAs that may have less impact on the overlying RPE tight junctions. Recent work suggests that there are differences in the expression of tight junction proteins in the RPE compared with the adherens junctions of the OLM (8), which may provide alternative targets. Indeed, Crb1 itself is worthy of further consideration, because its expression in the eye is restricted to the OLM (9).
Although ZO-1 proved an effective target for inducing reversible OLM disruption here, high concentrations were associated with a moderate level of cell death. Recently, Yang and colleagues demonstrated that double-stranded RNAs greater than 21 nucleotides can bind with the Toll-Like receptor 3 (TLR3) and induce apoptosis (52). It is unlikely that the apoptosis observed in this study was due to interactions with TLR3, because the levels observed were similar in both the siRNA-treated (targeting and nontargeting) and vehicle-treated eyes. Moreover, the cell loss declined with time, unlike the progressive degeneration observed by Yang and colleagues. Nonetheless, further work is required to optimize the surgical procedures involved to minimize any associated cell loss.
Of interest is the finding that integration in the 12-week-old Crb1rd8/rd8 mouse was not significantly better than wild-type controls, despite the continued disruption of the OLM. We also observed very high levels of GFAP and vimentin staining at the edge of the ONL in these older animals, with numerous Müller glial processes running along the outer limits of the ONL. This level and pattern of expression is similar to that seen in 4–6-week-old rho-/- mice and strongly indicative of glial scarring. Such scarring is often associated with neural injury and is known to obstruct regenerative processes, such as axonal regrowth, by forming physical and diffusional barriers that separate undamaged regions from the area of injury (13, 14, 31). We propose that in the Crb1rd8/rd8 mouse there is a trade-off between the progressive disruption of the OLM, permitting increased integration and an increase in glial scarring, ultimately inhibiting cell migration, despite the disruption of the OLM. There is thus a window of opportunity around 6 weeks of age, at which cell integration is increased. GFAP expression itself has been both positively and negatively correlated with cell migration. Kinouchi and colleagues (19) demonstrated that elimination of both GFAP and vimentin in wild-type mice permitted increased integration following transplantation of retinal progenitor cells. We have also observed that integration is negatively correlated with increasing gliosis in a number of models of degeneration (A.C.B and R.A.P, unpublished observations). Conversely, Perez and colleagues (54, 55) suggest that cells migrate at sites of GFAP upregulation following transplantation of fetal retinal sheets. Here, we observed similar levels of glial cell activation, as indicated by GFAP staining, following application of either ZO-1 siRNA, nontargeting siRNA, or vehicle alone, but integration was only significantly improved those eyes receiving the ZO-1 siRNA. Similarly, integration was significantly lower in control rho-/- mice, a model in which gliosis is prominent, and only enhanced in animals receiving ZO-1 siRNA. Together with the findings from the Crb1rd8/rd8 mouse and the AAA work reported previously, these results point strongly towards the improvements being as a result of OLM disruption. However, it is important to remember that it is highly unlikely that any one single barrier is responsible for preventing transplanted cell integration per se. Thus, any effective therapeutic treatment must encompass a number of different factors, including the reduction of reactive gliosis and subsequent glial scarring and transient disruption of the OLM.
In conclusion, we have provided strong evidence that the OLM represents a significant physical barrier to transplanted photoreceptor cell integration. Moreover, we present a novel use of RNAi technology to introduce a transient and reversible disruption of this barrier that when used in combination with cell transplantation leads to substantial increases in the number of successfully transplanted cells in both wild-type and degenerating retinas.
Footnotes
Acknowledgments
This work was supported by grants from the Wellcome Trust (082217) and Medical Research Council UK (G03000341). R.A.P. is a Royal Society University Research Fellow. R.E.M. is a Health Foundation Clinician Scientist Fellow. R.R.A. and R.E.M. were partially funded by the Department of Health's National Institute for Health Research Biomedical Research Centre at Moorfields Eye Hospital. We would like to thank A. Georgiadis, N. Gent, P. Munro, A. Edduadi, and J. Metelo for technical assistance.