Molecular Vision 2007; 13:1259-1273 <>
Received 19 March 2007 | Accepted 20 June 2007 | Published 23 July 2007

Analysis of the RPE transcriptome reveals dynamic changes during the development of the outer blood-retinal barrier

Lawrence J. Rizzolo,1,2 Xiang Chen,3 Matthew Weitzman,1,2 Ru Sun,1,2 Heping Zhang3

1Department of Surgery, Yale University School of Medicine, New Haven, Connecticut; 2Department of Ophthalmology and Visual Sciences; 3Department of Epidemiology and Public Health

Correspondence to: Lawrence J. Rizzolo, Department of Surgery, Yale University School of Medicine, PO Box 208062, New Haven, CT 06520-8062; Phone: (203) 785-6277; FAX: 203 785-5155; email:


Purpose: The morphology of the RPE shows minimal change as the neural retina and choriocapillaris differentiate. Nonetheless, initial studies of proteins related to the outer blood-retinal barrier suggest extensive remodeling of the retinal pigment epithelium (RPE) in response to this changing environment. A genomic approach was used to investigate the extent of this remodeling.

Methods: RPE was isolated from E7, E10, E14, and E18 chick embryos and total RNA extracted for probing the entire genome on Affymetrix microarray chips. Statistical parameters using ANOVA were adjusted to yield a theoretical false discovery rate of 5%. STEM software was used to cluster genes into statistically related patterns of expression. Gene ontology clustering, using Affymetrix software was used for functional clustering. The database was used as a source of known biological pathways.

Results: Of the 37,694 probesets on the microarray, 17,199 were absent. Of the 20,495 expressed probes, the expression of 8,889 was developmentally regulated. 4,814 of these could be clustered into 12 patterns of expression that were statistically significant. Minimal contamination by surrounding tissues was detected. The developmental patterns of 22 tight and adherens junction proteins were compared using hybridization to the microarray and quantitative PCR. Only two showed small variations from the patterns revealed by the microarray. The data indicate extensive remodeling of the extracellular matrix, cell surface receptors, cell-cell junctions, transcellular ion transport, and signal transduction pathways throughout development. Notably, the appearance of the mRNAs for claudin 20, ZO-3, and cadherins 13 and 20 very late in development suggest barrier properties continue to change after functional junctions are formed.

Conclusions: The data reveal a far more dynamic view of the RPE and its interactions with its environment than would be expected from morphological examination. The remodeling of junctional complexes, extracellular matrix interactions and transcellular transport capabilities indicates a continuous remodeling of the blood-retinal barrier as the retina develops. These data provide a standard whereby culture models of RPE function and regulation may be judged.


The retinal pigment epithelium plays a central role in the physiology of the retina [1]. As the outer-blood retinal barrier, it regulates transport between the neural retina and fenestrated capillaries of the choroid. Among its functions, retinal pigment epithelium (RPE) regulates the ionic environment of the subretinal space, phagocytizes shed outer segments of photoreceptors and participates in the visual cycle by converting trans-retinal to the photosensitive cis isoform. Diseases of the RPE result in retinal degeneration and degeneration of the choriocapillaris [2-4]. Transplantation of RPE into diseased eyes has met with limited success, perhaps because the diseased environment was unable to provide the environmental signals that regulate normal RPE function. To explore that possibility, our lab has studied how tissue-tissue interactions are established during normal development to regulate RPE's function as the outer blood-retinal barrier. The barrier consists of two components: Transcellular mechanisms regulate transport through the cells of the monolayer and establish transepithelial gradients. By contrast, tight junctions regulate diffusion through the paracellular spaces, which prevents transepithelial gradients from dissipating.

Aside from an increase in melanin and minor changes in morphology, the differentiation of RPE was thought to be completed early in retinal development. More recent studies demonstrate changes in cell polarity, cellular metabolism and the expression of intercellular junctional proteins [5-13]. These tight and adherens junctions bind the monolayer together and regulate transepithelial diffusion through the paracellular spaces. Development of the RPE can be divided into three stages that relate to the development of the inner and outer segments of photoreceptors. Across vertebrate species, these developmental milestones of the neural retina appear to be linked to developmental milestones of the choroid, Bruch's membrane and RPE [14,15]. Previous studies related the formation of tight junctions to these milestones [16]. In the early phase, before inner segments penetrate the outer limiting membrane, a rudimentary adherens junction binds the RPE monolayer together, but tight junctions are absent. Near the end of this phase, isolated tight junctional strands begin to appear. In the intermediate phase, which ends when outer segments begin to form, these strands gradually coalesce into a discontinuous network that encircles the cell. By mid-intermediate phase, the network becomes continuous and functional, but it continues to increase in depth and strand number throughout this period. During the late phase, the complexity (density of anastomotic connections between strands) increases. Throughout development, progressive morphological changes were also observed in the adherens junctions [17]. Throughout development, changes have been demonstrated in the expression and steady-state levels of some proteins that form these junctional complexes [6,7,12,16,18,19]. In vitro experiments indicate the choroid and neural retina regulate this remodeling process [16,20].

The only genomic analysis of RPE development that has been reported used zebrafish. That study focused on a single time point that corresponds to the intermediate to late phase transition [21]. To determine how extensive molecular remodeling of the RPE might be, we used a microarray of the chick genome to monitor how gene expression changed from the early through the late phase of development. Approximately 40% of the transcriptome that was expressed on embryonic day 7 (E7) changed by E18. The analysis focused on intracellular junctions and genes related to the barrier functions of the RPE.


Isolation of retinal pigment epithelium and total mRNA

Sheets of RPE were isolated from chicken embryos on E7, E10, E14, and E18 and stored in RNAlater (Qiagen, Valencia, CA), as described [16]. To isolate total RNA, the RNeasy Protect kit (Qiagen) was used according to the manufacturer's protocols. For each age, 3-4 independent preparations were used for analysis on Affymetrix microarrays of the chicken genome (Santa Clara, CA). For each preparation, sheets of RPE were pooled from 20-30 eyes. The quality of the total RNA was assessed by the Keck Center, Yale University using formamide gels and a 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA).

Hybridization and quantification by the reverse-transcriptase polymerase-chain-reaction

Hybridization of the microarrays and initial statistical analysis for quality control was performed by the Keck Center at the Yale University School of Medicine. For quantitative and semi-quantitative assays of specific mRNAs, we used real time PCR, as described [16] with the exception that GAPDH was used in place of 18S RNA to normalize the samples. For each age, 5-6 independent preparations of RNA were analyzed, and the standard error is reported. The following primers were used: Claudin 20, 5'-TAA CGC AGA TGC AAG GAC TG-3', 5'-GCA GAC TCC TCC AGC AAA AC-3'; ZO-1 α+; 5'-AAC CCA GCA ACC TCA TCA AC-3', 5'-GGA TCT ATA TGC GGC GGT AA-3'; ZO-1 α-; 5'-AAC TGC TTC TCA GCC GGT AT-3',5'-CTG CTC GTA CTC CCT ACT TGG-3'; ZO-2, 5'GAC AGG GCA GAC TTC TGG AG3',5'-TTG CCT CAC AGT GTT CAA GC-3'; ZO-3, 5'-GAC ACA AAC ATG GAC GAT GC-3', 5'-AAT GCG TCC GGA TGT AGA AG-3'. All assays were performed in triplicate.

Data processing

The data were first filtered to remove those probes not expressed at any time point. A probe is considered to be expressed at a time point if it is detected (i.e., labeled as "P" by GCOS software) in at least half of the replicates. For those probes expressed, the raw expression signal was log-transformed. To identify significantly differentially expressed probes, we applied classic one-way ANOVA analysis. A probe is considered developmentally regulated if its p-value is less than or equal to a threshold that yields a theoretical false discovery rate of 5% [22].


The Short Time-series Expression Miner (STEM, version 1.2.2b) software [23] with default parameters was used for analyzing the set of regulated probes. Briefly, STEM implements a novel clustering method that depends on a set of distinct and representative short temporal expression profiles and each probe in the dataset is assigned to a profile with closest match. The expected number of probes assigned to each profile is estimated by permutation and the statistically significantly over-expressed profiles are then identified.

Identification of protein pathways

Protein Lounge was used to determine the membership of the protein pathways described in this report. Affymetrix, Ensembl, and DAVID software were used to correlate probeset identifier numbers with the mRNA that they represent. DAVID and the Affymetrix cluster analysis tools, both based on the Gene Ontology (GO) classification system, were used to identify functional clusters within the statistical clusters identified by STEM software.

Results & Discussion

General results

Of the 37,694 probesets on the microarray, 20,495 were detected and 8,889 of those represented genes that were developmentally regulated (for raw data see the Gene Expression Omnibus, GSE7176). Using STEM software, the regulated probesets were clustered into groups according to their pattern of expression. Of the 50 clusters, 12 were judged to be statistically significant, and potentially contain genes that are coordinately regulated. These 12 clusters included 4,814 probesets. Expression increased for 6 clusters that could be distinguished according to the phases of development in which increases occurred (early: E7-E10, intermediate: E10-E14, and/or late: E14-E18). Similarly, expression decreased in the remaining 6 clusters. In the descriptions that follow, genes were grouped according to whether the expression of the mRNAs was stable over time, increased, decreased or were regulated but not included in a statistically significant cluster. Memberships and a graphic representation of the statistically significant clusters are included in the Appendix 1.

The validity of these findings would be compromised if the neural retina or the choroid contaminated the preparations of RPE. To minimize contamination, the RPE was isolated in large sheets that were readily distinguished by their pigment from the neural retina and choroid. Several lines of evidence suggested contamination was minimal. We examined 6 choroidal markers and 12 retinal markers. Only 4 were detected in our preparations. These were expressed at low levels, had large standard deviations, did not vary with age, or were detected on ages when they are not expressed in vivo [24]. For the choroid, the endothelial cell markers: Fli-1, CD144 (VE-cadherin), vascular cell adhesion molecule (VCAM) and CD34 were undetected. Further, the neural cell adhesion molecule (NCAM), a prominent component of chick choroid, but not of chick RPE, was undetected [13]. However, the endothelial cell marker, CD31 (PCAM) was evident. For the retina, undetected markers included opsin-1 (iodopsin), opsin-3 opsin-4, opsin-5, rhodopsin (opsin 2), rhodopsin kinase, and several cyclic nucleotide-gated channels. Although violet and blue cone opsins were detected, they were also evident on E7, when they were not detected by PCR [24]. These data suggest that some low hybridization signals may reflect non-specific binding and indicate that the RNA preparations were minimally contaminated with the surrounding tissues.

Correspondence to anticipated and known patterns of expression

Early differentiation and withdrawal from the cell cycle: Another way to validate the data is to examine patterns of expression that would be anticipated from the literature. The differentiation of the outer layer of the optic cup into RPE is initiated by the interaction of three transcription factors, Pax6, MitF and Otx2 [25-27]. Pax 6 expression decreases in the RPE during development and becomes restricted to the neural retina. In contrast, Otx2 is specific for the RPE. All three were detected on the microarray even though the E7 time point is after differentiation has begun. By contrast, Chx10, a negative regulator of RPE differentiation, and Rx1, an initiator of ocular development were not detected, indicating minimal contamination by the neural retina. The hybridization signal for Pax6 decreased 3x (p<0.0001) and MitF decreased 2x (p<0.002) between E7 and E18. As would be expected, Otx2 increased 1.5x (p<0.03). We were able to determine small changes with statistical confidence, because the ANOVA encompassed all 4 time points. As discussed at the end of this section, these small changes detected by hybridization translated into large changes, as measured by quantitative RT-PCR. Altogether, 47 transcription factors and regulators were found to decrease expression during development, while 48 others increased (Appendix 1).

Proliferation of the RPE decreases substantially in the central zone by E5 and by E12 is very low around the entire globe [28]. Accordingly, one would expect a decrease in the expression of genes that promote progression through the cell cycle and an increase in those that repress it. Figure 1 includes 30 cell-cycle genes that were developmentally regulated; 21 decreased. Of the nine than increased, proteins such as protein phosphatase 1, cyclin D1 and Cdc42 participate in a variety of cellular processes [29-31].

Visual cycle: The proteins of the visual cycle might be expected to increase in expression about the time that outer segments are made [8,32]. For some visual cycle proteins, the increase in mRNA was gradual, but for lecithin retinol acyltransferase and retinol dehydrogenase 12 a major increase was observed in the latter phases of development (Figure 2). The largest increase was for RPE65 (50x). Similar large increases in the late phase for other enzymes might have been obscured, because the microarray was saturated. There appeared to be stable expression of the mRNA for the interphotoreceptor retinoid binding protein (IRBP), a carrier protein for retinoids that is secreted in to the subretinal space. In contrast to the other mRNA's, the signal for IRBP was low and the standard error was large. IRBP might be expressed by chick RPE, as it is in zebrafish [33]. It is more likely that low signals with large errors were the hallmarks of non-specific binding or minor contamination by the neural retina.

Phagocytosis: Lysosomal proteins and phagocytic pathways should be expressed throughout development, but increased expression might be expected in the late phase when outer segments are being formed and RPE microvilli are lengthening in response. Shedding of outer segment discs would increase the phagocytic load on the RPE and there would be a need for membrane receptors (and associated signaling pathways) that are specific for the outer segments. Sixty seven genes associated with phagocytosis were examined (Figure 3). The expression of 60% of these was developmentally regulated and most of these increased their expression of mRNA. Notably, the mRNAs of receptor proteins for shed outer segments, CD36 and integrin subunit αv, increased during development. Although the magnitude of the changes was not as great, a similar pattern was observed for the 25 lysosomal enzymes that were examined (Figure 4).

Correlation of hybridization to the microarray with quantitative PCR: Table 1 summarizes the comparison of the two methods. With the exception of claudin 10, both methods agreed on whether an mRNA was present or absent. The next comparison examined whether the methods agreed that significant changes did, or did not, occur between the E7 to E10, E10 to E14 and E14 to E18 time points. Hybridization failed to detect a 7x decrease in AF-6 mRNA between E7 and E10 or a 5x increase in Par 3 mRNA between E14 and E18 [18]. This is consistent with the observation that a 35x increase in claudin 4L2, measured by PCR, registered as only a 1.7x increase on the microarray (see Figure 5). JAMs B and C were not detected by either method (JAM-A was not represented in the microarray). The biological significance of these data is discussed in the next section. These limited data demonstrate a high correlation between the qualitative patterns of expression exhibited by the hybridization and PCR techniques (Table 1). The presence or absence of only 1 of these 22 mRNAs was misidentified by hybridization. When neighboring time-points were compared, relative expression was misrepresented only twice out of 45 comparisons. It is unclear why hybridization to the microarray was insensitive to changes <7x, as measured by PCR. This insensitivity illustrates the power required of the statistical methods, as small changes measured on the microarray reflected changes large enough to be of biological significance. By using ANOVA to analyze the entire time-course, it was possible to make the small distinctions included in the following figures with a theoretical false discovery of rate 5% [22]. P-values for the figures are included in the Appendix 2.

These data indicated the mRNA preparations and statistical methods were of sufficient quality to examine the mRNAs for the proteins that enable RPE to interact with its environment, regulate its morphology and function as a blood-retinal barrier. The time-course of expression focused attention on genes that were both developmentally regulated and members of clusters whose pattern of expression was statistically correlated [23]. The next section further verifies and quantifies predictions based on hybridization to the microarray by using quantitative, real-time PCR of tight junctional proteins.

Remodeling of the apical junctional complex

Barrier function requires an apical junctional complex composed of adherens and tight junctions. The complex allows transepithelial concentration gradients to form and transmits signals that regulate cell size, shape, polarity and proliferation [34-39]. The transmembrane proteins of the complex are linked by adaptor proteins to filamentous actin and a variety of effector proteins [38]. The gradual remodeling of the complex during RPE development has been described at the morphological and molecular level for the adherens [6,7,17] and tight [12,16,18,19] junctions. The current study revealed alterations in gene expression that imply more extensive modulation of the complex's signaling capacity than was previously recognized.

Tight junctions: Earlier studies on the development of RPE tight junctions focused on the claudins. The claudin family of transmembrane proteins form the tight junctional strands observed by freeze-fracture electron microscopy and determine the selectivity of the paracellular pathway [40]. Changes in the expression of the claudins correlated with morphological changes in the fine-structure of RPE tight junctions. Several claudins were regulated at the level of transcription, protein stability and subcellular localization [16]. The qualitative agreement between PCR and hybridization that was noted in Table 1 is shown in detail in Figure 5 and Figure 6. Claudins 3, 4, and 15 were not detected by either method, but a weak signal was detected for claudin 10 on the microarray. The claudins that were detected by both methods exhibited very similar developmental patterns. The microarray revealed an additional claudin not previously reported for RPE, claudin 20. A large increase in claudin 20 mRNA occurred late in development, which was confirmed by real-time RT-PCR (Figure 6). By E18, the copy number for claudin 20 mRNA was greater than claudin 1, which suggests these are the most prominent claudins in the RPE tight junction. The magnitude of this late-phase increase was surprising given the minimal changes in the fine-structure of the tight junctions [16,41]. By contrast, the expression of claudins 1 and 5 did not increase after E14 (Figure 5). Although the mRNA for claudins 2 and 4L2 did increase in the late phase, their level of expression was lower than claudins 1 and 20. Nonetheless, minor claudins can modulate the function of the RPE tight junction [42-44]. Although the steady-state level of claudin 20 protein and its subcellular distribution remain to be determined, the changing ratio of the claudin mRNAs during development implies significant changes in function. This maturation of function occurred after the structure of the tight junctions was established [16], but during the time photoreceptors matured with the formation of outer segments.

The transmembrane proteins of the tight junction are linked to effector proteins and the cytoskeleton by adaptor proteins [38]. We focused on the ZO family of adaptors. Chick RPE expresses ZO-1 and ZO-1 like protein (ZO-1LP), which by Mr, immunogenicity and ability to bind occludin appear to be the orthologs of mammalian ZO-1α- and ZO-1α+ [12,19]. The difference between the two variants is an internal sequence of 80 amino acids, the alpha region. ZO-1α+ is more specific for epithelia. In RPE, ZO-1α- was expressed at high steady-state levels in the early phase and steadily declined during development. By contrast, ZO-1α+ protein was absent during the early phase but increased until E12 to become the dominant ZO-1 variant, whereupon its expression also decreased [19]. Unlike the proteins, the microarray data indicated that the expression of ZO-1 mRNA was constant (Figure 6). To confirm its presence in chick, primers that span the alpha region were used to subclone and sequence it. RT-PCR confirmed that ZO-1α- and ZO-1α+ mRNAs were expressed in RPE throughout development (data not shown). To pursue this finding, primer pairs were synthesized that were specific for either ZO-1α- or ZO-1α+. Although the ZO-1α- variant was constant, the ZO-1α+ variant increased throughout development (Figure 6). These data suggest two mechanisms that regulate expression of ZO-1. Regulation of RNA splicing would lead to a change in the ratio of ZO-1α- and ZO-1α+, and regulation of translation or degradation would affect each splice variant equally to decrease the steady-state protein levels observed during development.

Like ZO-1, steady-state levels of ZO-2 fell during development [12], even though ZO-2 mRNA did not decrease (Figure 6). By contrast, both hybridization and PCR detected an increase in ZO-3 mRNA during the late phase, between E14 and E18. This paralleled the late-phase increases of claudins 2, 4L2 and 20 that occur after the morphology of the tight junction is established [16].

In the early stage, there are few tight junctional strands despite the presence of many proteins implicated in the assembly of tight junctions, such as JAM-A, AF-6, PAR 3 and 6, occludin (not represented on the array), ZO-1 and ZO-2 [12,18,19]. This list of pre-existing proteins appears to be more extensive, given the mRNAs identified in Figure 7. Notably, the mRNAs for the adaptor proteins MAGI 1 and MUPP1 decreased, but for the adaptors MAGI 2 and MAGI 3, mRNAs were stably expressed.

It is counter-intuitive that assembly and adaptor proteins of the tight junction decrease when tight junctions are forming. This counter-intuitive behavior can be understood in terms of the assembly of the entire apical junctional complex. Many of these proteins also assemble the adherens junction. They first form a primordial complex that reorganizes to segregate different proteins into an adherens and a tight junction [45-50]. Unlike these experimental models, RPE differentiation is spread over a long period of time, which facilitates a more detailed examination of the remodeling process. In RPE, many of the proteins required to make a tight junction are present in advance, in association with an adherens junction, awaiting the appearance of the claudins and ZO-1α+. When the apical junctional complex remodels, the entire complex apparently requires less of these adaptor and assembly proteins. By contrast, the mRNA for other adaptor proteins, MAGI 2 and 3, are stably expressed and ZO-3 increases after assembly of tight junctions appears to be morphologically complete. Between human and chick, ZO-3 is least conserved of the ZO proteins and is not functionally redundant with ZO-1 and ZO-2 [15,51,52]. Therefore, the assembly and maturation of the apical junctional complex appears to involve a condensation of diverse proteins into a primordial complex, followed by a process of segregation and pruning, as new proteins are added and others are reduced to mature levels. Because these adaptor proteins express a large number of protein binding sites [15,53,54], and because the transcription of affiliated effector proteins are also transcriptionally regulated (Figure 7 and Figure 8), it is reasonable to assume that all the diverse functions of the apical junctional complex are modulated during development.

Adherens junctions: The remodeling of the adherens junction during chick RPE development has been documented by morphology and changes in the expression of the cadherins [6,7,17]. The expression of N-cadherin decreases throughout development, R-cadherin peaks at E14, and B-cadherin decreases after E14. In the current study, these changes were mirrored by the changes in mRNA levels (Figure 8). However, the expression of additional cadherins was observed. As B-cadherin expression declined in the late phase, cadherin-13 (heart cadherin) and cadherin-20 each increased 5X (p<0.00001). In other tissues, cadherins 11 and 13 are regulated in opposite directions, with cadherin 11 more typical of mesodermal tissues [35,55]. In RPE, expression of cadherin 11 decreased only transiently (p<0.00003). Cadherin 20 has been reported in melanocytes and neuroepithelial tissues, including the optic vesicle in mice [56]. Consistent with this change in receptors, there was regulation of members of their signaling pathways, such as Fyn, Zyxin and Lef1 (Figure 8).

Remodeling of the extracellular matrix

The modulation of the apical junctional complex occurs in the context of dramatic environmental changes and alteration of the RPE's ability to respond to those changes [15,16,20,37]. Besides known effects on the retina and choroid, the RPE modulated components of its extracellular matrix. The mRNA for the laminin chains required to form laminin 3 (α1β2γ1) and laminin 7 (α3β2γ1) were present throughout development (Figure 9 and Table 2). Laminin 1 (α1β1γ1) presumably decreased because the mRNA for the β1-chain decreased throughout development. By contrast, laminin 11 (α5β2γ1) presumably increased because the mRNA for the α5-chain increased throughout development. During the intermediate and late phases, there was also an increase in the expression of the β3-chain. However, one of its partners, the γ2-chain, was not detected. Therefore, the significance of β3 (and βx) remains to be explained. These changes represented a shift from an embryonic matrix enriched in laminin 1 chains to a more mature matrix that contains laminins 3, 7, and 11.

The transition from an immature to a mature basal lamina was also evident in the expression of collagen IV mRNAs (Figure 10). Six isoforms have been reported for the α-chain of collagen IV. The α6-chain, and the embryonic α2-chain decreased during development, but this was countered by an increase in the α4 chain (found in differentiated basal laminae). The α5 chain of collagen 4 appeared to be regulated, but it was not a member of any of the statistically significant clusters. Together with the changes in laminin mRNA, there was the potential to form a basal lamina that resembled the specialized basal laminae of lung, amniotic and placental membranes [57,58]. Notably, amnionic membranes and an extracellular matrix derived from placenta appeared to promote RPE differentiation in other species [59-62], whereas a tumor-derived, embryonic-like matrix failed to support normal differentiation [63].

The mRNAs for various other collagens were evident. These were presumably destined for the collagenous layers of Bruch's membrane that form in the latter stages of development. Evidence was found for the stable expression of collagen types 7, 9, 17, 18, and 20. The expression of types 1, 5, and 11 appeared to decrease during development, whereas types 3, 6, and 8 increased.

Remodeling of surface receptors

Receptors and signaling pathways determine how the RPE would respond to its remodeled environment. Matrix receptors, and their intracellular binding partners, shifted their expression (Figure 11). For example, integrins are αβ dimers. The mRNAs for the α1, α2, α6, and α5 integrin chains were stably expressed, but there was a decrease for α8 and β1. The latter mRNAs were replaced by α7, αv, and β8. The significance of these changes is best understood for integrin αvβ5 and CD36, which signal the phagocytosis of shed photoreceptor outer segments [64,65]. The significance of the other changes noted in Figure 11 remains to be investigated, but likely forms the basis for the differential responsiveness E7 and E14 RPE to environmental stimuli. In culture, E7 RPE responds to different retinal factors than E14 RPE [66]. Retinal condition medium affects the expression of some claudin family members in E7 RPE, but other members in E14 RPE [16].

The apical junctional complex and extracellular matrix receptors activate signal transduction pathways. Gene ontologies were used to identify signal transduction pathways that were represented in the statistically significant clusters. Several identified pathways related to RPE differentiation. In the TGFβ family, activin receptor type I mRNA increased during the early phase 1.2x (p<0.005), although activin receptor IIB mRNA decreased 5.3x (p<0.00001) throughout development. Activin βB (5x, p<0.0009) and BMP2 (1.4x, p<0.0001) mRNAs increased during the early phase. Similarly, a number of wnt and frizzled (wnt receptors) mRNAs were developmentally regulated. Altogether the expression of 175 genes involved in signal transduction changed during this period of development (Appendix 2).

The junctional complexes and many extracellular matrix receptors interact with the cytoskeleton. Actin and myosin are important regulators of apical junctional complex [67-70]. During RPE development, regulated expression was observed for the mRNA of 40 actin-myosin associated proteins (Figure 12). The distribution and function of the actin-based cytoskeleton is regulated by small G-proteins and their effector proteins. The mRNA for 51 proteins of this class were identified by the microarray, with the expression of 60% regulated during development (Figure 13).

Microtubules help mediate the intracellular trafficking of vesicles, the polarized distribution of proteins and phagocytosis. The expression of mRNA for 13 microtubule related proteins were regulated during development (Figure 14). There was very little change in the expression of intermediate filaments. The predominant intermediate filament expressed in chick RPE is vimentin, which was stably expressed in this study. Evidence was found for keratins B4, 14, and 15, glial fibrillary acidic protein, restin and desmuslin. The only intermediate filament whose expression was regulated was restin, which increased 2.1x (p<0.02) in the intermediate and late phases.

Transcellular components of the blood retinal barrier

An examination of the outer blood-retinal barrier should also consider the transcellular transport pathway. The transcellular pathway also develops gradually. As tight junctions form, the paracellular pathway for transepithelial glucose transport closes, which requires RPE to increase glucose transport through the transcellular pathway. Previous studies indicated that the housekeeping transporter, GLUT3 (CEF-GT3) was stably expressed, but that GLUT1 (SLC2A1) increased during development [71]. These data were confirmed and extended by the current study (Figure 15), but the story proves to be more complicated. There was stable expression of the mRNA for GLUT 10 (SLC2A10), but continuous or late phase increases for GLUT 8, 9, and 11 (SLC2A8, 9, 11). These increases corresponded to the time when functional tight junctions are forming. Besides these facilitated transporters, the mRNA for a sodium-coupled glucose transporter, SGLT1 (SLC5A1) appeared on E10 and increased 2.5x by E18. This type of transporter has not been previously reported in RPE. Together, these data indicate several mechanisms that would enable RPE to satisfy the retina's need for glucose after tight junctions become functional.

The RPE establishes a polarized distribution of ion pumps, channels and transporters to mediate active transcellular transport of ions and organic solutes. Transport physiologists have determined what types of transporters must exist in the apical and basolateral membranes of RPE, and many of these have been identified [72,73]. The microarray data suggested a more complicated story. Figure 16 lists the types of transport proteins identified by physiologists and the mRNA for candidates represented in the chick transcriptome. There were often multiple candidates for a given transporter. Further for a given transporter, expression of one candidate might be up-regulated when another candidate might be down-regulated or stably expressed. The data also suggested more transporters than have been characterized by the transport physiologists. Further investigation is needed to understand how this remodeling of transporter expression affects function and relates to the changing ion selectivity of the tight junctions.

A few observations are of note. In contrast to most epithelia, the Na,K-ATPase is partially polarized towards the apical membrane, and there is some suggestion that the beta subunit may be important for this atypical localization [74-76]. The β1 subunit mRNA was stably expressed, but the α1 and β2 subunit mRNAs increased during the intermediate and late phases, when microvilli were extended and the distribution of the ATPase became polarized [11]. Bestrophin, a member of a novel class of chloride channel, is found in the basolateral membrane of RPE [77]. Genetic defects in bestrophin result in Best vitelliform macular dystrophy. The mRNA for bestrophin increased during the intermediate and late phases of development. The monocarboxylate transporters are important for transporting lactate out of the retina [5,78]. A polarized distribution has been described for family members 1 and 3. The apical family member (member 1) increased in the intermedia and late phases, but the basolateral family member (member 3) decreased. The story becomes more complicated, as the microarray identified 3 additional players (family members 5, 8, and 12), two of which increased as the other decreased during development. The polarized distribution of these 3 family members remains to be determined.

The molecular remodeling of the RPE is far more extensive than ever imagined. As the neural retina and choriocapillaris differentiate, the RPE demonstrates its capacity to remodel the extracellular matrix and to update the signal transduction pathways that respond to the new signals that the apical and basal environments will provide. As the demands of the differentiating neural retina change, the RPE appears to reformat its functions as the outer blood-retinal barrier. The mRNAs are expressed for a new complement of transmembrane transporters and for proteins that would alter the specificity of its tight junctions. Although not the focus of this report, it stands to reason that there are corresponding changes in metabolic and catabolic pathways. There are several limitations to this analysis. The annotation of the chick database is a work in progress, and the microarray data must be considered provisional until confirmed by more rigorous methods. Further, studies of protein expression and subcellular distribution are needed for a more complete understanding. Nonetheless, the hypotheses generated by this analysis give a direction to further studies. Hopefully, this overview will encourage investigators to mine the data set to address additional functions of the RPE.


The Authors thank Drs. Yan Luo and Masayuki Fukuhara for help with preliminary studies. This work was supported by National Institutes of Health grant EY08694 (LJR) and CORE grant EY00785 (Department of Ophthalmology and Visual Science, Yale University).


1. Marmor MF, Wolfensberger TJ, editors. The retinal pigment epithelium: function and disease. New York: Oxford University Press; 1998.

2. Aramant RB, Seiler MJ. Progress in retinal sheet transplantation. Prog Retin Eye Res 2004; 23:475-94.

3. Del Priore LV, Hornbeck R, Kaplan HJ, Jones Z, Valentino TL, Mosinger-Ogilvie J, Swinn M. Debridement of the pig retinal pigment epithelium in vivo. Arch Ophthalmol 1995; 113:939-44.

4. Litchfield TM, Whiteley SJ, Lund RD. Transplantation of retinal pigment epithelial, photoreceptor and other cells as treatment for retinal degeneration. Exp Eye Res 1997; 64:655-66.

5. Deora AA, Philp N, Hu J, Bok D, Rodriguez-Boulan E. Mechanisms regulating tissue-specific polarity of monocarboxylate transporters and their chaperone CD147 in kidney and retinal epithelia. Proc Natl Acad Sci U S A 2005; 102:16245-50.

6. Grunwald GB. Cadherin cell adhesion molecules in retinal development and Pathology. Prog Retin Eye Res 1996; 15:363-92.

7. Liu X, Mizoguchi A, Takeichi M, Honda Y, Ide C. Developmental changes in the subcellular localization of R-cadherin in chick retinal pigment epithelium. Histochem Cell Biol 1997; 108:35-43.

8. Manes G, Leducq R, Kucharczak J, Pages A, Schmitt-Bernard CF, Hamel CP. Rat messenger RNA for the retinal pigment epithelium-specific protein RPE65 gradually accumulates in two weeks from late embryonic days. FEBS Lett 1998; 423:133-7.

9. Marmorstein AD, Bonilha VL, Chiflet S, Neill JM, Rodriguez-Boulan E. The polarity of the plasma membrane protein RET-PE2 in retinal pigment epithelium is developmentally regulated. J Cell Sci 1996; 109:3025-34.

10. Rizzolo LJ. Basement membrane stimulates the polarized distribution of integrins but not the Na,K-ATPase in the retinal pigment epithelium. Cell Regul 1991; 2:939-49.

11. Rizzolo LJ, Zhou S. The distribution of Na+,K(+)-ATPase and 5A11 antigen in apical microvilli of the retinal pigment epithelium is unrelated to alpha-spectrin. J Cell Sci 1995; 108:3623-33.

12. Wilt SD, Rizzolo LJ. Unique Aspects of the blood-brain barrier. In: Anderson JM, Cereijido M, editors. Tight junctions. 2nd ed. Boca Raton: CRC Press; 2001. p. 415-43.

13. Rizzolo LJ, Zhou S, Li ZQ. The neural retina maintains integrins in the apical membrane of the RPE early in development. Invest Ophthalmol Vis Sci 1994; 35:2567-76.

14. Rizzolo LJ. Polarity and the development of the outer blood-retinal barrier. Histol Histopathol 1997; 12:1057-67.

15. Rizzolo LJ. Development and role of tight junctions in the retinal pigment epithelium. Int Rev Cytol 2007; 258:195-234.

16. Rahner C, Fukuhara M, Peng S, Kojima S, Rizzolo LJ. The apical and basal environments of the retinal pigment epithelium regulate the maturation of tight junctions during development. J Cell Sci 2004; 117:3307-18.

17. Sandig M, Kalnins VI. Morphological changes in the zonula adhaerens during embryonic development of chick retinal pigment epithelial cells. Cell Tissue Res 1990; 259:455-61.

18. Luo Y, Fukuhara M, Weitzman M, Rizzolo LJ. Expression of JAM-A, AF-6, PAR-3 and PAR-6 during the assembly and remodeling of RPE tight junctions. Brain Res 2006; 1110:55-63.

19. Williams CD, Rizzolo LJ. Remodeling of junctional complexes during the development of the outer blood-retinal barrier. Anat Rec 1997; 249:380-8.

20. Peng S, Rahner C, Rizzolo LJ. Apical and basal regulation of the permeability of the retinal pigment epithelium. Invest Ophthalmol Vis Sci 2003; 44:808-17.

21. Leung YF, Ma P, Dowling JE. Gene expression profiling of zebrafish embryonic retinal pigment epithelium in vivo. Invest Ophthalmol Vis Sci 2007; 48:881-90.

22. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society. Series B (Statistical Methodology) 1995; 57:289-300.

23. Ernst J, Nau GJ, Bar-Joseph Z. Clustering short time series gene expression data. Bioinformatics 2005; 21:i159-68.

24. Adler R, Tamres A, Bradford RL, Belecky-Adams TL. Microenvironmental regulation of visual pigment expression in the chick retina. Dev Biol 2001; 236:454-64.

25. Martinez-Morales JR, Rodrigo I, Bovolenta P. Eye development: a view from the retina pigmented epithelium. Bioessays 2004; 26:766-77.

26. Zhao S, Rizzolo LJ, Barnstable CJ. Differentiation and transdifferentiation of the retinal pigment epithelium. Int Rev Cytol 1997; 171:225-66.

27. Bharti K, Nguyen MT, Skuntz S, Bertuzzi S, Arnheiter H. The other pigment cell: specification and development of the pigmented epithelium of the vertebrate eye. Pigment Cell Res 2006; 19:380-94.

28. Stroeva OG, Mitashov VI. Retinal pigment epithelium: proliferation and differentiation during development and regeneration. Int Rev Cytol 1983; 83:221-93.

29. Ceulemans H, Bollen M. Functional diversity of protein phosphatase-1, a cellular economizer and reset button. Physiol Rev 2004; 84:1-39.

30. Etienne-Manneville S. Cdc42--the centre of polarity. J Cell Sci 2004; 117:1291-300.

31. Fu M, Wang C, Li Z, Sakamaki T, Pestell RG. Minireview: Cyclin D1: normal and abnormal functions. Endocrinology 2004; 145:5439-47.

32. Bridges CD. Distribution of retinol isomerase in vertebrate eyes and its emergence during retinal development. Vision Res 1989; 29:1711-7.

33. Stenkamp DL, Calderwood JL, Van Niel EE, Daniels LM, Gonzalez-Fernandez F. The interphotoreceptor retinoid-binding protein (IRBP) of the chicken (Gallus gallus domesticus). Mol Vis 2005; 11:833-45 <>.

34. Cereijido M, Anderson JM, editors. Tight junctions. 2nd ed. Boca Raton: CRC Press; 2001.

35. Halbleib JM, Nelson WJ. Cadherins in development: cell adhesion, sorting, and tissue morphogenesis. Genes Dev 2006; 20:3199-214.

36. Mandell KJ, Parkos CA. The JAM family of proteins. Adv Drug Deliv Rev 2005; 57:857-67.

37. Ban Y, Rizzolo LJ. Differential regulation of tight junction permeability during development of the retinal pigment epithelium. Am J Physiol Cell Physiol 2000; 279:C744-50.

38. Matter K, Aijaz S, Tsapara A, Balda MS. Mammalian tight junctions in the regulation of epithelial differentiation and proliferation. Curr Opin Cell Biol 2005; 17:453-8.

39. Van Itallie CM, Anderson JM. The molecular physiology of tight junction pores. Physiology (Bethesda) 2004; 19:331-8.

40. Van Itallie CM, Anderson JM. Claudins and epithelial paracellular transport. Annu Rev Physiol 2006; 68:403-29.

41. Kniesel U, Wolburg H. Tight junction complexity in the retinal pigment epithelium of the chicken during development. Neurosci Lett 1993; 149:71-4.

42. Van Itallie C, Rahner C, Anderson JM. Regulated expression of claudin-4 decreases paracellular conductance through a selective decrease in sodium permeability. J Clin Invest 2001; 107:1319-27.

43. Furuse M, Furuse K, Sasaki H, Tsukita S. Conversion of zonulae occludentes from tight to leaky strand type by introducing claudin-2 into Madin-Darby canine kidney I cells. J Cell Biol 2001; 153:263-72.

44. Van Itallie CM, Fanning AS, Anderson JM. Reversal of charge selectivity in cation or anion-selective epithelial lines by expression of different claudins. Am J Physiol Renal Physiol 2003; 285:F1078-84.

45. Ando-Akatsuka Y, Yonemura S, Itoh M, Furuse M, Tsukita S. Differential behavior of E-cadherin and occludin in their colocalization with ZO-1 during the establishment of epithelial cell polarity. J Cell Physiol 1999; 179:115-25.

46. Asakura T, Nakanishi H, Sakisaka T, Takahashi K, Mandai K, Nishimura M, Sasaki T, Takai Y. Similar and differential behaviour between the nectin-afadin-ponsin and cadherin-catenin systems during the formation and disruption of the polarized junctional alignment in epithelial cells. Genes Cells 1999; 4:573-81.

47. Martinez-Estrada OM, Villa A, Breviario F, Orsenigo F, Dejana E, Bazzoni G. Association of junctional adhesion molecule with calcium/calmodulin-dependent serine protein kinase (CASK/LIN-2) in human epithelial caco-2 cells. J Biol Chem 2001; 276:9291-6.

48. Miyoshi J, Takai Y. Molecular perspective on tight-junction assembly and epithelial polarity. Adv Drug Deliv Rev 2005; 57:815-55.

49. Rajasekaran AK, Hojo M, Huima T, Rodriguez-Boulan E. Catenins and zonula occludens-1 form a complex during early stages in the assembly of tight junctions. J Cell Biol 1996; 132:451-63.

50. Thomas FC, Sheth B, Eckert JJ, Bazzoni G, Dejana E, Fleming TP. Contribution of JAM-1 to epithelial differentiation and tight-junction biogenesis in the mouse preimplantation embryo. J Cell Sci 2004; 117:5599-608.

51. Adachi M, Inoko A, Hata M, Furuse K, Umeda K, Itoh M, Tsukita S. Normal establishment of epithelial tight junctions in mice and cultured cells lacking expression of ZO-3, a tight-junction MAGUK protein. Mol Cell Biol 2006; 26:9003-15.

52. McNeil E, Capaldo CT, Macara IG. Zonula occludens-1 function in the assembly of tight junctions in Madin-Darby canine kidney epithelial cells. Mol Biol Cell 2006; 17:1922-32.

53. Collins JR, Rizzolo LJ. Protein-binding domains of the tight junction protein, ZO-2, are highly conserved between avian and mammalian species. Biochem Biophys Res Commun 1998; 252:617-22.

54. Gonzalez-Mariscal L, Betanzos A, Nava P, Jaramillo BE. Tight junction proteins. Prog Biophys Mol Biol 2003; 81:1-44.

55. Nollet F, Kools P, van Roy F. Phylogenetic analysis of the cadherin superfamily allows identification of six major subfamilies besides several solitary members. J Mol Biol 2000; 299:551-72.

56. Moore R, Champeval D, Denat L, Tan SS, Faure F, Julien-Grille S, Larue L. Involvement of cadherins 7 and 20 in mouse embryogenesis and melanocyte transformation. Oncogene 2004; 23:6726-35.

57. Nguyen NM, Senior RM. Laminin isoforms and lung development: all isoforms are not equal. Dev Biol 2006; 294:271-9.

58. Gelse K, Poschl E, Aigner T. Collagens--structure, function, and biosynthesis. Adv Drug Deliv Rev 2003; 55:1531-46.

59. Stanzel BV, Espana EM, Grueterich M, Kawakita T, Parel JM, Tseng SC, Binder S. Amniotic membrane maintains the phenotype of rabbit retinal pigment epithelial cells in culture. Exp Eye Res 2005; 80:103-12.

60. Ohno-Matsui K, Ichinose S, Nakahama K, Yoshida T, Kojima A, Mochizuki M, Morita I. The effects of amniotic membrane on retinal pigment epithelial cell differentiation. Mol Vis 2005; 11:1-10 <>.

61. Voloboueva LA, Liu J, Suh JH, Ames BN, Miller SS. (R)-alpha-lipoic acid protects retinal pigment epithelial cells from oxidative damage. Invest Ophthalmol Vis Sci 2005; 46:4302-10.

62. Maminishkis A, Chen S, Jalickee S, Banzon T, Shi G, Wang FE, Ehalt T, Hammer JA, Miller SS. Confluent monolayers of cultured human fetal retinal pigment epithelium exhibit morphology and physiology of native tissue. Invest Ophthalmol Vis Sci 2006; 47:3612-24.

63. Ban Y, Rizzolo LJ. A culture model of development reveals multiple properties of RPE tight junctions. Mol Vis 1997; 3:18 <>.

64. Ryeom SW, Sparrow JR, Silverstein RL. CD36 participates in the phagocytosis of rod outer segments by retinal pigment epithelium. J Cell Sci 1996; 109:387-95.

65. Finnemann SC, Nandrot EF. MerTK activation during RPE phagocytosis in vivo requires alphaVbeta5 integrin. Adv Exp Med Biol 2006; 572:499-503.

66. Ban Y, Wilt SD, Rizzolo LJ. Two secreted retinal factors regulate different stages of development of the outer blood-retinal barrier. Brain Res Dev Brain Res 2000; 119:259-67.

67. Kapus A, Szaszi K. Coupling between apical and paracellular transport processes. Biochem Cell Biol 2006; 84:870-80.

68. Madara JL, Moore R, Carlson S. Alteration of intestinal tight junction structure and permeability by cytoskeletal contraction. Am J Physiol 1987; 253:C854-61.

69. Bruewer M, Hopkins AM, Hobert ME, Nusrat A, Madara JL. RhoA, Rac1, and Cdc42 exert distinct effects on epithelial barrier via selective structural and biochemical modulation of junctional proteins and F-actin. Am J Physiol Cell Physiol 2004; 287:C327-35.

70. Luo Y, Zhuo Y, Fukuhara M, Rizzolo LJ. Effects of culture conditions on heterogeneity and the apical junctional complex of the ARPE-19 cell line. Invest Ophthalmol Vis Sci 2006; 47:3644-55.

71. Ban Y, Rizzolo LJ. Regulation of glucose transporters during development of the retinal pigment epithelium. Brain Res Dev Brain Res 2000; 121:89-95.

72. Gallemore RP, Hughes BA, Miller SS. Retinal pigment epithelial transport mechanisms and their contributions to the electroretinogram. Prog Retin Eye Res 1997; 16:509-66.

73. Strauss O. The retinal pigment epithelium in visual function. Physiol Rev 2005; 85:845-81.

74. Mircheff AK, Miller SS, Farber DB, Bradley ME, O'Day WT, Bok D. Isolation and provisional identification of plasma membrane populations from cultured human retinal pigment epithelium. Invest Ophthalmol Vis Sci 1990; 31:863-78.

75. Quinn RH, Miller SS. Ion transport mechanisms in native human retinal pigment epithelium. Invest Ophthalmol Vis Sci 1992; 33:3513-27.

76. Rizzolo LJ. Polarization of the Na+, K(+)-ATPase in epithelia derived from the neuroepithelium. Int Rev Cytol 1999; 185:195-235.

77. Hartzell C, Qu Z, Putzier I, Artinian L, Chien LT, Cui Y. Looking chloride channels straight in the eye: bestrophins, lipofuscinosis, and retinal degeneration. Physiology (Bethesda) 2005; 20:292-302.

78. Philp NJ, Wang D, Yoon H, Hjelmeland LM. Polarized expression of monocarboxylate transporters in human retinal pigment epithelium and ARPE-19 cells. Invest Ophthalmol Vis Sci 2003; 44:1716-21.

Rizzolo, Mol Vis 2007; 13:1259-1273 <>
©2007 Molecular Vision <>
ISSN 1090-0535