Lipocalin-Type Prostaglandin D Synthase ( (cid:98) -Trace) Is Located in Pigment Epithelial Cells of Rat Retina and Accumulates within Interphotoreceptor Matrix

Glutathione-independent prostaglandin D synthase, identical to (cid:98) -trace, (a major CSF protein), is localized in the CNS. This enzyme, lipocalin-type prostaglandin D synthase, is a member of the lipocalin family of secretory proteins that transport small lipophilic substances. This enzyme’s activity in adult rat retina was enriched sixfold in retinal pigment epithelium (RPE) and even more in interphotoreceptor matrix (IPM), all higher than brain. Western blots with anti-lipocalin-type prostaglandin D synthase showed three distinct immunoreactive bands. In the retinal cytosolic fraction, only one band was observed ( M r 25,000); in IPM, the larger component occurred ( M r 26,000). The RPE membrane-bound fraction showed two bands ( M r 20,000 and 23,000), indicating synthesis, and the cytosolic fraction contained two bands ( M r 23,000 and 26,000), indicating modiﬁcation for release into IPM. At least two glycosylation sites occurred on the prostaglandin D synthase moiety, ex-plaining the three immunoreactive bands in Western blots. with polyclonal antibodies against this lipocalin-type enzyme showed intense localization in RPE, but less in photoreceptor outer and inner segments. In situ hybridization showed mRNA speciﬁcally expressed in RPE. Thus, lipocalin-type prostaglandin D synthase is predominantly expressed in RPE and actively accumulated in IPM. This may demonstrate gene sharing because, while catalyzing prostaglandin D 2 synthesis, it may perform an additional, unrelated role in IPM. This enzyme is secreted from the RPE into IPM from which it is then taken up by photoreceptors. However, the nature of its ligand(s) is not known; they may be retinoids and/or docosahexanoic acid.

The lipocalin-type PGD synthase, the only member of the lipocalin family (Nagata et al., 1991) that has been shown to have enzymatic activity, has been localized to the leptomeninges, choroid plexus, and oligodendrocytes in brain and spinal cord (Urade et al., , 1993. This enzyme also shares characteristics similar to other lipocalins: it is secreted into the CSF (Clausen, 1961) and binds retinoids with affinities similar to those of other lipocalins such as retinol-and retinoic acid-binding protein (Tanaka et al., 1995). Thus, this enzyme is predicted to be bifunctional, acting as a PGD 2 -producing enzyme as well as an intercellular transporter of retinoids or other lipophilic substances. In fact, this may represent another example of gene sharing in which a gene sequence codes for a protein that serves two completely different functions (Piatigorsky et al., 1988;Urade et al., 1995a). For example, in the case of crystallins, several metabolic enzymes can also function as structural components in the lens (Piatigorsky and Wistow, 1989). However, a bifunctional nature of lipocalin-type PGD synthase in the CNS remains to be demonstrated. PGD 2 is the major prostaglandin formed in the eye (Goh et al., 1987). Also, lipocalin-type PGD synthase has been detected in retina (Goh et al., 1987), and mRNA for the human PGD 2 receptor was recently found to be highly expressed in retina (Boie et al., 1995). This suggests that PGD 2 is actively involved in retinal function. Thus, the retina may be used to study the functional significance and regulatory mechanisms of PGD 2 and lipocalintype PGD synthase.
In this study, we investigate the cellular localization of lipocalintype PGD synthase in retina and show that this enzyme is predominantly produced in retinal pigment epithelium (RPE), and that it accumulates in the interphotoreceptor matrix (IPM). Moreover, lipocalin-type PGD synthase is also found in photoreceptors, whereas its mRNA is present only in RPE cells. Together, these results suggest that RPE-secreted lipocalin-type PGD synthase in the IPM is subsequently taken up by photoreceptors.
Portions of this work have appeared in abstract form Marcheselli et al., 1996).

MATERIALS AND METHODS
Tissues. Male and female adult rats (Sprague Dawley and Long-Evans; 225-300 gm) were decapitated under deep ether anesthesia. Eyes were collected on ice, anterior segments and lenses were quickly removed, and vitreous was drained.
For Western blot analyses, retinas and RPE were obtained from Long-Evans pigmented rats because RPE cells can be visualized during the separation process. Retinas were removed from eyecups and incubated in buffer containing (in mM): 50 Tris-HCl, pH 7.4, 2 EGTA, 2 MgCl 2 , and 250 sucrose at 37ЊC for 40 min. Retinas (n ϭ 5 per sample point) were gently separated from RPE cells and then homogenized in 300 l of the same buffer. RPE and the soluble components of IPM were separated by centrifugation of the incubation medium at 10,000 ϫ g for 10 min. The pellet obtained here was combined with the pellet from the next centrifugation step, and then resuspended for the membrane-bound fraction. The supernatants were centrifuged at 100,000 ϫ g for 30 min, and the fluids obtained constituted the soluble fraction or cytosol.
For histological studies, eyecups (n ϭ 12) from albino Sprague Dawley rats were fixed at 4ЊC for 5 hr in PBS containing 4% (w/v) paraformaldehyde. Tissue was then cryoprotected with PBS containing 30% (w/v) sucrose at 4ЊC for 3 hr, cut into halves along the optic nerve, embedded in ornithine carbamoyl transferase compound, and frozen on dry ice. Cryosections of 20 m thickness were cut, mounted onto poly-L-lysinecoated or gelatin-subbed glass slides, and stored at Ϫ80ЊC until used for immunohistochemistry and in situ hybridization.
Immunohistochemistry. All dilutions were made in PBS containing 0.3% Triton X-100 (Sigma) and 10% normal sheep serum (Chemicon). Sections were incubated with 10% normal sheep serum at 25ЊC for 1 hr and subsequently with polyclonal rabbit anti-rat brain PGD synthase serum (1:5000 dilution) (Urade et al., 1985) at 25ЊC overnight. Controls were made with normal rabbit serum or antiserum preabsorbed with excess amounts of recombinant rat brain PGD synthase (Urade et al., 1995a). Tissue sections were next incubated with anti-rabbit serum or antiserum preabsorbed with excess amounts of recombinant rat brain PGD synthase (Urade et al., 1995a), then were treated with anti-rabbit IgG F(ab) 2 fragments (20 m/ml) coupled with fluorescein or rhodamine (Boehringer Mannheim) at 25ЊC for 2 hr. Sections were examined in bright field by Nomarski differential interference contrast and with epifluorescence microscopy (Nikon Optiphot-2, Melville, NY; Olympus IX70, Tokyo, Japan) or by confocal laser scanning microscopy (MCR 600, Bio-Rad).
Western blotting. Tissues (n ϭ 5 retinas per sample) were homogenized in PBS with glass Potter homogenizers, and the soluble fraction was collected by centrifugation of homogenates at 100,000 ϫ g for 30 min. The cytosolic fraction was incubated with an excess amount of antibody at 4ЊC overnight to collect the immunoreactive protein (Urade et al., 1985). The immune complex was then recovered by incubation with protein A-Cellulofine (10 l) (Chisso, Kumamoto, Japan) at 25ЊC for 2 hr, solubilized in 1% SDS, and applied to SDS-PAGE.
Measurement of enzyme activity. PGD synthase activity was determined with 40 M [1-14 C] PGH 2 in the presence of 1 mM dithiothreitol (Urade et al., 1985), except when otherwise stated. Protein concentration was determined by the method of Lowry et al. (1951) as modified by Bensadoun and Weinstein (1976) using bovine serum albumin as a standard.
In situ hybridization. A 567 bp fragment corresponding to positions 40 -606 of the cDNA for rat brain PGD synthase was subcloned into pBluescript II KS(ϩ) vector. For in vitro transcription, template DNA was linearized with HindIII (for the antisense probe) or EcoRI (for the sense probe). Probes were labeled with 11-digoxigenin-uridine triphosphate (Boehringer Mannheim) using a Gemini System II kit (Promega). Retinal sections were fixed and acetylated as reported previously (Urade et al., 1993). Sections were hybridized at 50ЊC overnight with digoxigeninlabeled antisense or sense RNA probes (100 ng/l). After treatment with RNase A (40 g/ml) at 37ЊC for 6 min, the RNA probe bound to the tissue was visualized by immunostaining with a digoxigenin detection kit (Boehringer Mannheim).

Immunohistochemical localization of the lipocalin-type PGD synthase in rat retina
The cellular localization of lipocalin-type PGD synthase was examined in the retina of adult rats by immunofluorescence staining with polyclonal antibodies against this enzyme (Fig. 1). Immunofluorescence was clearly observed in RPE and photoreceptor outer segments, but not in other retinal regions ( Fig. 1 A,B,E,F ). No positive fluorescence was detected when we used normal rabbit serum, antibodies against hematopoietic PGD synthase, or the anti-lipocalin-type PGD synthase serum preabsorbed with excess amounts of the recombinant enzyme (data not shown). Higher magnification revealed that immunofluorescence was distributed throughout the cytoplasm of the epithelial cells, outlining each nucleus (Fig. 1C), and confocal laser scanning microscopy demonstrated that these positive signals were located in the perinuclear regions and in distinct domains within the cytoplasm (Fig. 1 D). Weak immunofluorescence was also detected in outer and inner segments of photoreceptor cells (Fig. 1 E,F ). The lipocalin-type PGD synthase mRNA was localized in RPE cells by in situ hybridization with digoxigenin-labeled antisense RNA (Fig.  1G). Control sections incubated with sense RNA did not show positive staining in RPE cells (Fig. 1 H).

Biochemical and molecular biological characterization of PGD synthase in retina
When PGD synthase distribution levels in rat ocular tissues were studied, the largest accumulation (Fig. 2) was found in the IPM soluble fraction, followed by the RPE cytosolic fraction. Western blot analysis with polyclonal antibodies against PGD synthase demonstrated a twofold increase per milligram of total protein between the IPM and the RPE soluble fractions, and an eightfold increase between the soluble and membrane-bound fractions of RPE. In retina, levels were very low, remaining about the same between the two fractions.
Interestingly, when PGD synthase activity was measured, even higher activity was detected in IPM (9.3 nmol ⅐ min Ϫ1 ⅐ mg Ϫ1 of protein) compared with the RPE soluble fraction (3.5 nmol ⅐ min Ϫ1 ⅐ mg Ϫ1 protein). However, retina demonstrated very low PGD synthase activity (0.58 nmol ⅐ min Ϫ1 ⅐ mg Ϫ1 protein). Moreover, it is important to note that ocular tissue expresses a fivefold higher activity for PGD synthase than brain tissue. Western blots demonstrated consistent immunoreactivity of the polyclonal antibody in association with three, tightly colocalized bands (Fig. 3), suggesting that the PGD synthase moiety can undergo glycosylation (Urade et al., 1989). The distribution of PGD synthase-immunoreactive protein between the three bands was noticeably different when retina, IPM, and RPE soluble and membrane-bound fractions were compared. The IPM, which is a soluble fraction, demonstrated components of the three bands, but the heaviest, the most highly glycosylated form, was the most enriched (Fig. 3). A sizeable amount of this heavy band was also found in the cytosolic fraction of RPE cells. The medium-weight Figure 1. Detection of lipocalin-type PGD synthase-immunoreactive protein (A-F ) and its mRNA (G, H ) in rat retina. A, Nomarski differential interference contrast micrograph. C, Choriocapillaris; RPE, retinal pigment epithelium; OS/IS, outer segments/inner segments; ONL, outer nuclear layer; OPL, outer plexiform layer; INL, inner nuclear layer; IPL, inner plexiform layer; GCL/NFL, ganglion cell layer/nerve fiber layer. B, Immunofluorescence micrograph of A with anti-lipocalin-type PGD synthase antiserum, detecting immunoreactivity in RPE cells and photoreceptor outer segments ( green band). Scale bar, 50 m. C, Fluorescence micrograph of immunoreactive RPE cells at higher magnification ( green band). Note typical cuboidal shape of RPE cells. Scale bar, 20 m. D, Confocal laser scanning micrograph of immunoreactive RPE cells. Immunoreactivity is localized to the perinuclear space and distinct domains within the cytoplasm (bright white spots). Cell nuclei are dark ellipsoids (arrows). Scale bar, 5 m. E, Nomarski differential interference contrast micrograph. RPE cell layer (above) and photoreceptor layer (below) are separated mechanically. F, Fluorescence micrograph of E showing the presence of lipocalin-type PGD synthase immunoreactivity in both RPE cells and photoreceptors (red bands). Scale bar, 50 m. G, In situ hybridization with digoxigenin-labeled antisense RNA for this enzyme. mRNA for lipocalin-type PGD synthase is detected in RPE cells (dark band above arrows), but not within photoreceptors. H, Hybridization with digoxigenin-labeled sense RNA to label nonspecific background. No positive signal was detected. RPE cells are located just above arrows. Scale bar, 20 m.
band was the most abundant component in both the cytosolic and membrane-bound fractions of RPE (Fig. 3). In retina, both fractions contained only the medium-weight band. To correlate the mobility shifts of the immunoreactive proteins with degrees of glycosylation, IPM and retina were incubated with N-glycanase for different lengths of time. After 15 hr of treatment, bands appearing at M r 25,000 and 26,000 shifted partially to M r 20,000 and M r 23,000 (Fig. 4). After 40 hr, both retina and IPM showed these bands reduced to M r 20,000. These results indicate that the immunoreactive protein is N-glycosylated in at least two positions, which is also consistent with the properties of the enzyme purified from rat brain (Urade et al., 1989). Additionally, no immunoreactivity was observed in retinal extracts or IPM (data not shown) when antibodies against the hematopoietic-type PGD synthase were used (Urade et al., 1989).
Northern blot analysis of whole retina (retina plus RPE) revealed that mRNA for lipocalin-type PGD synthase also existed in this tissue (Fig. 5B). An 850 bp cDNA probe hybridized with RNA from the retina, demonstrating the presence of lipocalin-type PGD synthase, but no hybridization was observed for hematopoietic-type PGD synthase (Fig. 5C). The expression levels of PGD synthase in retina were relatively lower than those of the arachnoid membrane and total brain RNA. Thus, it can be    inferred that the PGD synthase mRNA present in retinal tissues (Fig. 1G,H ) originated from RPE.

DISCUSSION The lipocalin-type PGD synthase as a lipophilic substance transporter in the IPM
Lipocalin-type PGD synthase is a member of the superfamily composed of the lipophilic molecule transporters, and recent work has demonstrated its ability to bind retinoic acid and retinal with affinities comparable to those of other retinoid-binding proteins (Tanaka et al., 1995). IPM participates in a very active retinoid recycling system to replenish the photosensitive chromophores of rhodopsin within photoreceptor outer segments. In fact, at least four retinoid-binding proteins that participate in the visual cycle have been identified in association with photoreceptors, IPM, and RPE (Bok, 1985). Moreover, the supply of retinoids from systemic circulation is strictly controlled by the bloodretina barrier at the RPE level, demonstrating the importance of this layer of cells to the overall health and maintenance of retina . PGD synthase may also be involved in the transport of retinoids or retinoid-like compounds through IPM toward photoreceptor cells. For example, one lipophilic candidate as a ligand for this transport enzyme is docosahexanoic acid, an essential fatty acid that is supplied systemically to retina by liver (Scott and Bazan, 1989) for RPE and photoreceptor cells Bazan, 1990, 1993). Interestingly, this molecule is also linked to a recycling pathway within the retina (Gordon et al., 1992), and interphotoreceptor retinoid binding protein (IRBP) has been implicated as a carrier of this essential fatty acid (Bazan et al. 1985;Bazan, 1990).
The lipocalin-type PGD synthase as an enzyme in the eye PGD 2 is a major prostaglandin formed in the eye (Birkle and Bazan, 1984), and PGD synthase has the highest activity among other prostaglandin synthases in ocular tissues, especially within the retina (Goh et al., 1987). Recently, the human PGD 2 receptor was cloned, and its mRNA tissue distribution showed the highest abundance in retina (Boie et al., 1995), although its specific cellular localization in retina has not yet been shown. These observations, together with our findings, indicate that PGD 2 may be important in RPE function. However, to answer the questions about PGD 2 production in retina, it will be necessary to locate cycloxygenases 1 and 2, which catalyze the conversion of arachidonic acid to PGH 2 at the cellular level in retina. Because PGH 2 has a very short half-life in aqueous solutions, the cyclooxygenases should be very closely associated with PGD synthase as part of the PGD 2 -producing pathway.
There are structural and functional similarities between retinal and CNS lipocalin-type PGD synthases. This and its mRNA are found in epithelial cells of the arachnoid membranes and choroid plexus in brain (Urade et al., 1993) and are secreted into CSF within its closed compartment in large amounts (Hoffman et al., 1993;Watanabe et al., 1994). RPE and the leptomeninges are ontogenically and functionally analogous to each other, forming the blood-retina and blood-brain barriers with tight junctions between cells to seal the closed compartments of retina and brain. Furthermore, although we found the immunoreactive protein of this enzyme in photoreceptor cells, no mRNA could be detected in these cells (Fig. 1G,H ). A similar observation has also been reported in rat brain, in which immunoreactivity was detected in most neurons of infant rats and several neurons in layers I-II of the cerebral cortex of adult rats Urade et al., 1993), but no mRNA was detected in these neurons (Urade et al., 1993). Photoreceptor cells, which are highly specialized neurons, seem to respond in a similar manner as the immunoreactive neurons detected in brain. Therefore, these results suggest that lipocalin-type PGD synthase is secreted by brain epithelial cells into the closed compartment containing the CSF and by retinal RPE cells into the IPM, followed by subsequent uptake from these compartments by specific neurons.
In this study, we have shown by enzyme activity, Western blot analysis (Fig. 2), immunofluorescence staining ( Fig. 1 A-F ), and in situ hybridization (Fig. 1G,H ) that lipocalin-type PGD synthase is highly enriched in RPE cells and accumulates in large amounts within the IPM. Also, glycosylation of the PGD synthase moiety may be associated with the process of mobilization from the endoplasmic reticulum (the site of synthesis) to the IPM, where the most abundant component appears to be the highly glycosylated form. PGD synthase activity and protein levels found in IPM and RPE soluble fractions positively correlate (despite different glycosylation levels), in agreement with previous findings, indicating that the levels of glycosylation of the PGD synthase moiety do not affect enzymatic activity (Urade et al., 1989). Because retina does not express PGD synthase mRNA (Fig. 1G,H ) and contains only a glycosylated form of PGD synthase (Fig. 3), these observations strongly suggest directional movement from the RPE membrane-bound fraction (site of synthesis) and the RPE cytosolic fraction (transport), through the IPM soluble fraction (transport), to photoreceptors ( Fig. 1 E,F ).
Among the many questions that arise from the findings reported in this paper is the possible presence of a lipocalin-type PGD synthase selective uptake mechanism (or receptor) in photoreceptors. Is this located in the inner segments? What is the nature of the molecule, if any, that is being transported? Besides retinoic acid and/or docosahexanoic acid, can PGD 2 itself be attached to the enzyme of its synthesis and then be intercellularly transported to the photoreceptor cells where it may elicit a function? This sequestration of PGD 2 can be a regulatory mechanism to prevent further metabolism to PGJ 2 , a potent bioactive modulator (Forman et al., 1995;Kliewer et al., 1995). Moreover, lipocalin-type PGD synthase may be a unidirectional carrier molecule from RPE to the photoreceptor, because it is synthesized in RPE. IRBP, on the other hand, is synthesized in photoreceptors and then secreted into the IPM (Pepperberg et al., 1993). One may argue that carrier molecules present in the IPM could transport ligands in either direction.
Our study suggests that lipocalin-type PGD synthase may have multiple functions, acting as both enzyme and intercellular transporter, and the high concentration of lipocalin-type PGD synthase in the IPM strongly argues that this molecule plays an important function in retina. Finally, we have shown that retina-RPE can serve as an excellent model to study the function and metabolism of PGD 2 and the bifunctional ability of the lipocalin-type PGD synthase as a PGD 2 -producing enzyme and a potential lipophilic ligand transporter.