Involvement of Retinoic Acid/Retinoid Receptors in the Regulation of Murine αB-crystallin/Small Heat Shock Protein Gene Expression in the Lens*

Crystallins are a diverse group of abundant soluble proteins that are responsible for the refractive properties of the transparent eye lens. We showed previously that Pax-6 can activate the αB-crystallin/small heat shock protein promoter via the lens-specific regulatory regions LSR1 (−147/−118) and LSR2 (−78/−46). Here we demonstrate that retinoic acid can induce the accumulation of αB-crystallin in N/N1003A lens cells and that retinoic acid receptor heterodimers (retinoic acid receptor/retinoid X receptor; RAR/RXR) can transactivate LSR1 and LSR2 in cotransfection experiments. DNase I footprinting experiments demonstrated that purified RAR/RXR heterodimers will occupy sequences resembling retinoic acid response elements within LSR1 and LSR2. Electrophoretic mobility shift assays using antibodies indicated that LSR1 and LSR2 can interact with endogenous RAR/RXR complexes in extracts of cultured lens cells. Pax-6 and RAR/RXR together had an additive effect on the activation of αB-promoter in the transfected lens cells. Thus, the αB-crystallin gene is activated by Pax-6 and retinoic acid receptors, making these transcription factors examples of proteins that have critical roles in early development as well as in the expression of proteins characterizing terminal differentiation.

The refractive properties of the transparent eye lens depend on a diverse group of globular proteins called crystallins that comprise approximately 90% of the water-soluble proteins of this tissue (1,2). Despite their specialized function in the lens, crystallins are surprisingly diverse and may differ among species. Moreover, crystallins often play more than one biological role, a situation called gene sharing (3), with many being related or identical to metabolic enzymes or stress proteins (4 -6). These multifunctional crystallins are expressed very highly in the lens and to a lesser extent in other tissues, where they have nonrefractive roles.
We have been studying mouse ␣B-crystallin, a conserved small heat shock protein (29,30) that is constitutively expressed highly in the lens and more moderately in many other tissues (31,32). ␣B-crystallin is also induced by stress (33) and overexpressed in numerous diseases (34,35). The differential constitutive expression of the murine ␣B-crystallin gene is developmentally and transcriptionally controlled (32,36,37). Transgenic mouse experiments have established that the sequences downstream of Ϫ164 are sufficient to direct lens-specific gene expression (38). This 5Ј-flanking sequence contains two lens-specific regulatory regions called LSR1 (Ϫ147/Ϫ118) and LSR2 (Ϫ78/Ϫ46). Pax-6 can interact with both LSR1 and LSR2 and activate the ␣B-crystallin promoter in transient transfection experiments (39). In the present study, we show by DNase I footprinting, antibody/electrophoretic mobility shift assay (EMSA), site-directed mutagenesis, and transient-cotransfection experiments that RAR/RXR heterodimers can interact at retinoic acid-responsive elements (RAREs) within LSR1 and LSR2 and can activate the ␣B-crystallin promoter in lens cells either alone or collaboratively with Pax-6.

EXPERIMENTAL PROCEDURES
Nucleic Acid Isolation-For transfection experiments, plasmid DNA was isolated and purified using the Qiagen plasmid kit according to the manufacturer's instructions (Qiagen Inc., Chatsworth, CA).
Northern (RNA) Analysis-Total RNA was isolated from N/N1003A cells (40) treated with RA (Sigma) by using the RNA Isolation Kit (Stratagene, La Jolla, CA) and subsequently fractionated by electrophoresis through a 1.5% agarose-formaldehyde gel. The RNA was transferred to a Duralon membrane (Stratagene, La Jolla, CA) and hybridized to a 230-base pair HindIII-BamHI restriction fragment from exon 3 of the mouse ␣B-crystallin gene (32). The probe was labeled by using the Ready-To-Go Random Prime Labeling System (Amersham Pharmacia Biotech). Prehybridizations were performed at 60°C for 30 min, and hybridizations were carried out at 60°C for 90 min by using QuickHyb (Stratagene, La Jolla, CA) according to the manufacturer's instructions. Membranes were washed and autoradiographed as described previously (41). Methylene blue staining was performed as earlier (42) to monitor the integrity of RNA, the relative amounts of RNA loaded on the gel, and the efficiency of transfer to Duralon membranes. Membranes were exposed for autoradiography on Kodak XAR5 film at Ϫ80°C with an intensifying screen for 12 h.
Nuclear Extracts, Oligonucleotides, and Antisera-Nuclear extracts (9) were prepared from ␣TN4 -1 (43) and N/N1003A lens cells. Complementary oligodeoxynucleotides were synthesized (model 380A synthesizer; Applied Biosystems) and annealed at a 1:1 molar ratio as described previously (44). The oligodeoxynucleotides were labeled on one strand using T4 polynucleotide kinase, and electrophoretic mobility * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
EMSA and DNase I Footprinting-A polymerase chain reactiongenerated fragment corresponding to the Ϫ190 to ϩ40 sequence of the ␣B-crystallin gene was used for footprinting experiments with purified mouse RAR/RXR receptors. DNA and protein were incubated and treated with DNase I as described previously (44). The RAR/RXR proteins were kindly provided by Drs. Keiko Ozato and Jorge Blanco (NICHD, National Institutes of Health, Bethesda, MD). RAR␥ was obtained from Santa Cruz Biotechnology. End-labeling, EMSA, and DNase I footprinting were performed as described earlier (44).
Western (Protein) Analysis-Nuclear extracts prepared from ␣TN4 -1 and N/N1003A lens cells were fractionated by electrophoresis in a Tris-glycine polyacrylamide gel; the separated proteins were transferred to nitrocellulose membranes using a Trans-Blot (Bio-Rad). Immunoblotting was performed according to the manufacturer's instructions (Vector Labs, Burlingame, CA).
Site-directed Mutagenesis-Plasmids containing mutations generated previously (38) within the Ϫ164/ϩ44 EcoRI/PstI fragment of the mouse ␣B-crystallin gene, cloned in pRD30A (36), were used for transient transfection experiments and EMSAs. In brief, site-specific mutations (Mu-9762 and Mu-9763) (38) were introduced by using an oligodeoxynucleotide-directed mutagenesis kit (Sculptor in vitro mutagenesis kit, Amersham Pharmacia Biotech). Mutated oligodeoxynucleotides contained the substitution sequence TCTAGA (XbaI site) and 20 bases on each side complementary to the ␣B-crystallin promoter sequence. The resulting mutated restriction fragments were subcloned into pRD30A at the unique BamHI site (36). All constructs were confirmed by sequencing the ligated junctions and mutated regions.
Cell Culture, Transient Transfections, and CAT Assays-Mouse COP-8 fibroblasts (46) and N/N1003A lens cells were grown in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) containing 10% fetal calf serum and 50 g/ml of gentamicin in 10% CO 2 . The cells were propagated on 60-mm diameter plastic dishes. 10 g of wild-type ␣Bpromoter-cat plasmids (p65-7 and p11-3) (36) or mutated test plasmids (Mu-9762 and Mu-9763) (38); increasing amounts (0.25-1 g) of pSV40RAR␤ and pRSVRXR␤ (gifts from Drs. Keiko Ozato and Jorge Blanco), which express the wild-type RAR␤ and RXR␤, respectively (47); and 2 g of internal control pCH110, which expresses ␤-galactosidase (Amersham Pharmacia Biotech), were cotransfected for 6 h by the calcium phosphate method as described previously (44). Cells were treated with 100 l of 0.1 ng/ml of RA in the morning following transfection for 1 h. The cells were harvested, and extracts were prepared 48 h after transfection. CAT activities were determined by the biphasic assay (48), and ␤-galactosidase activities were determined as described previously (44). The transfection data represent the means of three separate experiments, with each experiment being conducted in duplicate.

RESULTS
Northern Blot Hybridization of ␣B-crystallin mRNA-In order to test whether retinoid signaling can induce endogenous ␣B-crystallin gene expression in lens cells, Northern blot hybridizations were performed with total RNA isolated from N/N1003A cells treated with increasing concentrations of RA (Fig. 1). The intensity of hybridization of the labeled probe to ␣B-crystallin mRNA was approximately 3 times greater in the cells treated with higher concentrations of RA ( Fig. 1, lanes  1-3). Control tests showed no increase in glyceraldehyde-3phosphate dehydrogenase and ␣-actin mRNAs after treatment with RA (data not shown).
Western Blot Analysis for RAR/RXR Receptors-Nuclear extracts from N/N1003A and ␣TN4 -1 lens cells were used to test for the presence of RAR/RXR receptors. Western blot analysis using anti-RAR and anti-RXR antibodies showed that both DNase I Footprinting with Lens Nuclear Extract and RAR/ RXR-We next examined the possibility that heterodimers of retinoic acid receptors can bind to the lens-specific sites LSR1 and LSR2. Three retinoic acid receptor heterodimers (RAR␤/ RAR␥, RXR␤/RAR␥, and RAR␤/RXR␤) were tested for the ability to protect the Ϫ190/ϩ40 fragment of the ␣B-crystallin gene from digestion with DNase. Fig. 3 shows that three regions were protected by each of the heterodimers tested, with the weakest footprint generated by RXR␤/RAR␥. The protected regions comprised LSR1 (Ϫ132/Ϫ110), LSR2 (Ϫ73/Ϫ54), and a region between LSR1 and LSR2 (Ϫ106/Ϫ87). The LSR1 and LSR2 sequence was footprinted on both DNA strands; however, the intervening region (Ϫ106/Ϫ87) was not footprinted on the upper (sense) strand (data not shown).
The footprinted sequences and the surrounding nucleotides are shown in Fig. 4A. Regions that resemble the consensus binding sequence for RAR/RXR (12) are designated RARE in Fig. 4B; the nucleotides represented by the larger uppercase letters conform with the RAR/RXR consensus binding sequence, while the nucleotides represented by the smaller letters deviate from the consensus binding sequence. In general, the DNase I-footprinted sequences in both LSR1 and LSR2 match quite well with the RAR/RXR consensus binding sequence.
EMSAs Using Competitor Oligodeoxynucleotides and RAR/ RXR Antibodies-Next we examined the binding of nuclear proteins derived from the N/N1003A lens cells to the LSR1 and LSR2 regions of the ␣B-crystallin promoter. Incubation of double-stranded oligodeoxynucleotide LSR2 (Ϫ78/Ϫ28) with the N/N1003A nuclear extract resulted in the formation of three major complexes (Fig. 5A, lane 2, C1-3). These complexes were abolished by competition with self-oligodeoxynucleotide LSR2 (Fig. 5A, lane 3) and diminished with a double-stranded oligodeoxynucleotide containing LSR1 (Fig. 5A, lane 4). Doublestranded oligodeoxynucleotides RARE and RAR␤, which both contain the consensus RARE site, competed for the formation of complexes 1 and 3 (Fig. 5A, lanes 5 and 6, respectively). By contrast, double-stranded oligodeoxynucleotide 9718/9719, which contains the chicken ␣A-crystallin Pax-6 binding site, and double-stranded oligodeoxynucleotide 926/927, which contains a consensus binding sequence for Pax-6, competed for the formation of complex 2 but not complexes 1 and 3 (Fig. 5A,  lanes 7 and 8, respectively). Although we cannot be certain, we believe that the two complexes migrating faster than complex 3 are nonspecific, since they were not significantly affected by competition with oligodeoxynucleotides containing consensus RARE and Pax-6 binding sites.
Functional Cotransfection Tests with RAR␤/RXR␤ and Pax-6 cDNA Expression Plasmids-To test whether RAR␤/ RXR␤ receptors can activate the ␣B-crystallin promoter, transient cotransfection experiments were performed using cDNA expression plasmids in N/N1003A lens cells. Vector alone (pRSV) did not activate the Ϫ164/ϩ44 ␣B-crystallin promoter fused to the cat gene (p65-7) in cotransfected N/N1003A cells (data not shown). p65-7 contains LSR1 and LSR2. By contrast, cotransfection with a mixture of pSV40RAR␤ and pRSVRXR␤ caused a 5-6-fold RA-dependent stimulation of CAT activity in the cells transfected with p65-7 (Fig. 7A). Cotransfection with either pSV40RAR␤ or pRSVRXR␤ alone, however, did not stimulate the reporter gene expression to a significant level (data not shown). pSV40RAR␤ and pRSVRXR␤ stimulated CAT expression approximately 3-fold in cotransfection experiments using p11-3, which contains LSR2 but lacks LSR1 (Fig. 7A). The absolute amount of CAT activity produced from p11-3 was at least 3-fold lower than that resulting from p65-7 (data not shown). Site-specific mutations Mu-9762 and Mu-9763 that were generated previously in the Ϫ147/Ϫ118 sequence (38) of LSR1 were used to verify that RAR␤/RXR␤ stimulates ␣Bpromoter activity through LSR1. Thus, the mutated promoter constructs (Mu-9760 and Mu-9761) were compared with the wild-type construct (p65-7) for their ability to direct expression of the cat gene in N/N1003A cells cotransfected with pSV40RAR␤ and pRSVRXR␤. The ␣B-crystallin promoters containing the Mu-9762 and Mu-9763 mutations were only about half as responsive as the wild type promoter in p65-7 to stimulation by pSV40RAR␤ and pRSVRXR␤ in the cotransfected cells (Fig. 7B).
Finally, since both Pax-6 (39) and RAR␤/RXR␤ (above) can stimulate ␣B-crystallin promoter activity, we tested whether these transcription factors have an additive or synergistic effect in cotransfection experiments. First, we confirmed that the Pax-6 expression plasmid, pKW10-Pax-6 (49), can stimulate CAT expression in N/N1003A lens cells transfected with p65-7, since our earlier experiments showing a maximum 5-fold increase in ␣B-crystallin promoter activity were performed in the COP-8 fibroblast cell line (39). The data in Fig. 8A show that pKW10-Pax-6 has a similar effect on p65-7 activity in the cotransfected N/N1003A cells as previously in COP-8 cells, even with respect to the decrease at higher concentrations of pKW10-Pax-6. Similar results were obtained with pKW10-Pax-6 and p11-3, which contains only LSR2 (data not shown). Cotransfection experiments using pKW10-Pax-6, pSV40RAR␤, and pRSVRXR␤ in conjunction with p11-3 (Fig. 8B) or p65-7 (Fig. 8C) showed that Pax-6 and RAR␤/RXR␤ have an additive stimulatory effect on ␣B-crystallin promoter activity. The additive stimulation was approximately twice as great with p65-7 as with p11-3. These data suggest that LSR1 and LSR2 are regulatory regions that utilize both Pax-6 and RAR/RXR for ␣B-crystallin promoter activity. DISCUSSION We have shown previously in transgenic mice using reporter transgenes that the high lens and lower nonlens expression of the mouse ␣B-crystallin gene are developmentally controlled at the transcriptional level (36,37). Lens-specific expression has been localized to LSR1 (Ϫ147/Ϫ118) and LSR2 (Ϫ78/Ϫ46), with the minimal lens-specific ␣B-crystallin promoter fragment identified being the Ϫ115/ϩ44 sequence (38,39). The combined presence of LSR1 and LSR2 in the Ϫ164/ϩ44 promoter fragment is approximately 30 times more active in the lens than is LSR2 alone in the minimal promoter fragment in transgenic mice (39). Promoter activity is augmented approximately 7-fold in the transgenic mouse lens when the Ϫ426/ϩ44 promoter fragment is used, which includes an enhancer at positions Ϫ426/Ϫ259 (36,38). Thus, high lens activity of the mouse ␣B-crystallin gene depends on coupling multiple cis-control elements and their cognate transcription factors.
We have shown earlier and confirm here that Pax-6 can activate the ␣B-crystallin promoter via both LSR1 and LSR2 in transient transfection experiments (39). Pax-6 also contributes to the lens expression of the mouse (9) and chicken (8) ␣A-, the chicken ␦1- (10), and the guinea pig -(11) crystallin genes (7). In recent transgenic mouse experiments, a mutant TATA-like sequence associated with LSR2 preferentially reduced ␣B-crystallin promoter activity in the lens in a Pax-6-independent fashion, indicating that, as expected, multiple factors contribute to the specialized activity of the ␣B-crystallin promoter in the lens (50). The present investigation shows that retinoic acid receptors, especially RAR␤/RXR␤ heterodimers, can also bind LSR1 and LSR2 and activate the ␣B-crystallin promoter. The activation of the ␣B-crystallin promoter by the simultaneous presence of RAR␤/RXR␤ and Pax-6 is additive rather than synergistic and leaves unresolved whether or not these factors physically interact with each other or through co-factors. Indeed, the mechanism of gene activation by retinoid receptors has not been established and involves chromatin alterations as well as direct interactions with DNA sequences (51). It remains to be determined if retinoic acid receptors play a role in the nonlens or stress-induction of the ␣B-crystallin/small heat shock protein (35).
Retinoic acid receptors are members of the superfamily of nuclear factors (thryroid hormone, steroid hormone, and vita-min D3 receptors) and are involved in a wide array of developmental processes (52)(53)(54)(55)(56). The importance of retinoid signaling for eye development in mice has been established by application of exogenous RA (57) and by deleting various combinations of RAR and RXR genes (18,19,58,59). The existence of retinoic acid receptors in cultured lens cells was shown in our Western blots in the present experiments. A role of retinoid signaling for lens differentiation is implied by the generation of abnormal lens phenotypes by ectopic expression of cellular RA-binding protein 1 (15) and RAR (60) and by the expression of reporter genes driven by RAREs of the RAR␤ gene in the presumptive (61) and developing (20) lens of transgenic mice. It has also FIG. 6. A, EMSA analyses of protein-DNA interactions using purified RAR/ RXR proteins and 32 P-labeled LSR2 oligodeoxynucleotide (Ϫ78/Ϫ28). Free and protein-complexed oligodeoxynucleotides were resolved by 5% polyacrylamide gel electrophoresis; supershifts were performed with monoclonal (lanes 3-6) or polyclonal (lanes 8 -10) antibodies. Anti-RXR(␣,␤,␥) and anti-RAR(␣,␤,␥) recognize all the forms of their respective antigens. Antibodies are described under "Experimental Procedures." C, specific complex; S, supershifted complex. B, autoradiograms of EMSAs indicating that ␣TN4 -1 lens nuclear proteins are antigenically related to RAR/RXR. Components used in the binding reactions are indicated above the lanes. 0.1 ng of labeled LSR2 binding site oligodeoxynucleotide containing sequences from positions Ϫ78/Ϫ48 of the ␣B-crystallin promoter (all lanes) was used. The arrow indicates the probe complexed with nuclear protein. been demonstrated that a minimal promoter-lacZ reporter gene fused to the RARE from the human RAR␤-2 gene is expressed in the zebrafish as early as embryonic day 9.5 in specific embryonic regions including the optic cup (62). Thus, retinoic acid receptors and Pax-6 are both examples of general factors that play essential roles in the early development of the lens as well as in the regulated expression of crystallin genes, which encode the major proteins of the terminally differentiated lens (1). This is consistent with the idea that one of the selective mechanisms used for recruiting the multifunctional crystallins is their responsiveness to transcription factors required for the development and maintenance of the transparent lens (5,6,63).
So far our data show only that RXR␤ and at least one of the RARs (␣, ␤, or ␥) are present in the N/N1003A and ␣TN4 -1 lens cells. With respect to the intact lens, a broad complex forms with the LSR2 oligodeoxynucleotide and lens nuclear extract; however, this complex was unaffected by the addition of the set of RAR and RXR antibodies used in the experiments with the cultured cells (data not shown). Because of the overlap between the Pax-6 and RAR/RXR binding sites, it is possible that Pax-6, RAR/RXR, and other factors present in the lens nuclei bind simultaneously at LSR2 and leave the antibodyinteractive sites for the retinoic acid receptors unavailable. In any case, further experiments are necessary to establish unequivocally which retinoic acid receptors may be involved in the activation of the ␣B-crystallin gene in the cultured lens cells and in the intact lens.
The present results add to previous experiments indicating that RA and its receptors play a critical role in the regulation of crystallin genes in the lens. RA has been shown to activate the ␦1-crystallin gene in stably transformed mouse teratocarcinoma stem cells (64) and in cultured lens epithelial cells from newly hatched chickens (17). Recent cotransfection experiments using reporter genes in recombinant plasmids have provided more direct evidence implicating retinoic acid receptors in the control of the chicken ␦1-crystallin gene (14). Unlike the FIG. 7. Cotransfection of ␣B-crystallin promoter-cat and RAR␤ and RXR␤ cDNA constructs in the N/N1003A lens cell line. A, relative CAT levels in N/N1003A cells cotransfected with pSV40RAR␤ and pRSVRXR␤ (wild type RAR␤ and RXR␤ cDNAs, respectively) and either p11-3 (wild type Ϫ115/ϩ44 ␣B-crystallin promoter fragment fused to cat gene) or p65-7 (wild type Ϫ164/ϩ44 ␣B-crystallin promoter fragment fused to cat gene). The CAT levels expressed are relative to that obtained in parallel cells cotransfected with pSV40RAR␤, pRS-VRXR␤, and pRD30A (the promoterless parent vector) (36). B, relative CAT levels in N/N1003A cells co-transfected with p65-7 or the Ϫ164/ ϩ44 promoter fragment containing the Mu-9762 or Mu-9763 and pSV40RAR␤ and pRSVRXR␤. CAT levels are relative to parallel tests using pRD30A instead of p65-7. Cells were harvested 48 h after DNA removal; CAT activity was determined by the biphasic assay (48) and normalized with respect to the activity of ␤-galactosidase, which resulted from cotransfection of pCH110 (see "Experimental Procedures"). The transfection data represent the means of three separate experiments, with each experiment being conducted with duplicate dishes.
FIG. 8. Transfection of ␣B-crystallin promoter-cat and Pax-6 cDNA and RAR/RXR cDNA constructs in the N/N1003A lens cell line. A, relative CAT levels in N/N1003A cells co-transfected with pKW10-Pax-6 (wild type Pax-6 cDNA) and either p11-3 or p65-7. B, relative CAT levels in N/N1003A cells cotransfected with pKW10-Pax-6 and pSV40RAR␤ and pRSVRXR␤ with either p11-3 or p65-7 (C). CAT levels are relative to parallel tests using pRD30A instead of p65-7 or p11-3. Cells were harvested 48 h after DNA removal; CAT activity was determined by the biphasic assay (48) and normalized with respect to the activity of ␤-galactosidase, which resulted from cotransfection of pCH110 (see "Experimental Procedures"). The transfection data represent the means of three separate experiments, with each experiment being conducted with duplicate dishes. ␦1-crystallin promoter/enhancer, the ␦2-crystallin promoter/ enhancer is not stimulated by RAR␤ in the cotransfected primary lens epithelial cells. This differential responsiveness is particularly interesting, since ␦2-crystallin, an active argininosuccinate lysase, is present at a relatively low concentration in the chicken lens, while enzymatically inactive ␦1-crystallin is the major ␦-crystallin in the lens (3,65).
Extensive experiments have demonstrated that the mouse ␥F-crystallin gene is controlled by retinoid signaling. It was first shown that RAR/RXR heterodimers bind to a novel everted RARE (called ␥F-HRE) consisting of two half-sites separated by eight base pairs in the 5Ј-flanking sequence (66). The regulation of the ␥F-crystallin gene by retinoic acid receptors appears very complex, inasmuch as ␥F-HRE is activated by T 3 R/RXR as well as RAR/RXR, yet is repressed by T 3 R/RAR␣ (13). There is also an RAR-related orphan receptor, ROR␣1, that is expressed in the mouse lens and binds as a monomer to the ␥F-HRE 3Ј-half site and spacer sequences (66). ROR␣1 stimulates ␥F-crystallin promoter activity in transfected primary chicken lens epithelial cells. Moreover, ROR␣1 occupancy and promoter activation are blocked by competing RAR/RXR heterodimers in the absence of RA; the blockage of ROR␣1 activation of ␥F-HRE by RAR␣ is dose-dependent and similar to the repression of the T 3 response from ␥F-HRE reporter plasmids. The novelty of ␥F-HRE (12) and the involvement of ROR␣1, which does not compete for binding to ␤-RARE or TRE (66), raises the possibility that the stimulation of the ␥Fand ␣B-crystallin promoters operates by different pathways.
Many invertebrates have complex eyes with cellular lenses containing abundant crystallins (67). Virtually nothing is known about the developmental pathways controlling the development of these lens-containing invertebrate eyes or about the regulatory mechanisms used for expressing their crystallin genes. Recent experiments have raised the possibility that retinoid signaling may extend to crystallin gene expression in invertebrates. Two novel lens crystallin genes (J1A-and J1Bcrystallin) cloned from the cubomedusan jellyfish (Tripedalia cystophora) (68) have RARE half-sites in their promoter regions, and these sequences bind a cloned RXR homologue derived from the same species. 2 If a causal connection can be made between retinoic acid receptors and J-crystallin gene expression, it would provide strong evidence that retinoid signaling is a conserved pathway for crystallin gene expression throughout the animal kingdom.