Retina-specifically expressed novel subtypes of bovine cyclophilin.

The Drosophila ninaA gene encodes photoreceptor-specific cyclophilin thought to play a critical role in rhodopsin folding or transport during its synthesis or maturation in the most abundant subclass of photoreceptors. Cyclophilins comprise a highly conserved family of proteins which are the primary targets of the potent immunosuppressive drug, cyclosporin A (CsA), and which display peptidyl prolyl cis-trans-isomerase (PPIase) activity. In an attempt to identify mammalian cyclophilins with properties similar to the NinaA protein, a probe derived from the ninaA cDNA was used to screen bovine retina cDNA libraries. The screen identified two major alternatively spliced forms of cDNA that would encode proteins containing a region of high homology to other cyclophilins and that are expressed specifically in the retina. These proteins represent a new class of cyclophilins with novel structural features and greatly reduced PPIase and CsA binding activities in comparison to other known cyclophilins. Tissue in situ hybridization and immunolocalization of the proteins showed that the RNA and protein products are expressed in photoreceptors as well s other retinal neurons. However, among photoreceptors, the proteins are found predominantly in cones. Thus, mammalian retinas do contain cyclophilins that are retina-specifically and photoreceptor class-preferentially expressed. The results suggest that, in cones, the main function of these proteins is, like the NinaA protein, to facilitate proper folding or intracellular transport of opsins.

The Drosophila ninaA gene encodes a photoreceptor-specific cyclophilin thought to play a critical role in opsin/rhodopsin maturation (Schneuwly et al., 1989;Shieh et al., 1989). The ninaA gene was first identified from mutations which cause severe reductions in the amount of rhodopsin in the Drosophila compound eye (Pak, 1979;Stephenson et al., 1983). These mutations, named nina, fall into several complementation groups, ninaA, B, C, etc. From early on, the genes corresponding to two of these complementation groups, ninaA and ninaE, were of special interest because mutations in these genes affect the rhodopsin level in R1-6 photoreceptors and not in the other classes of photoreceptors in the compound eye (Larrivee et al., 1981). The ninaE gene was subsequently shown to encode R1-6 opsin (O'Tousa et al., 1985;Zuker et al., 1985). The ninaA gene, on the other hand, was found to encode a small membrane-bound protein with strong sequence homology to cyclophilin (Schneuwly et al., 1989;Shieh et al., 1989). The molecular basis of the photoreceptor class-specific effects of ninaA mutations is not understood.
Cyclophilins comprise a highly conserved family of proteins which are the primary targets of the potent immunosuppressive drug, cyclosporine A (CsA), 1 used in the treatment of graft rejection and immune disorders (Heitman et al., 1992;Sigal and Dumont, 1992). The in vivo functions of cyclophilins are poorly understood. The reports that the enzyme, peptidyl prolyl cis-trans-isomerase (PPIase), known to accelerate protein folding in vitro, is identical to a previously identified cyclophilin (Fischer et al., 1989;Takahashi et al., 1989) led to the hypothesis that the primary role of the NinaA protein is to act on rhodopsin as a folding catalyst (Schneuwly et al., 1989;Shieh et al., 1989;Stamnes et al., 1991). More recently, Baker et al. (1994) showed that NinaA forms a specific stable complex with R1-6 rhodopsin in vivo and suggested that it functions as a chaperone for R1-6 opsin/rhodopsin. Thus, to date, the NinaA protein is the only molecularly characterized member of the cyclophilin family for which the specific substrate and a probable function in vivo have been identified.
If a Drosophila rhodopsin requires a cyclophilin in its synthesis, transport, or maturation, one might ask whether mammalian rhodopsins also require cyclophilins in a similar role. Such cyclophilins, if present, would likely be expressed predominantly or specifically in the retina. However, most cyclophilins characterized to date are quite unlike the NinaA protein in that they are abundantly and ubiquitously expressed (Walsh et al., 1992). Stamnes et al. (1991) reported on a bovine protein that cross-reacts with NinaA antibodies, but the protein is not retina specific and was never characterized. The present work was undertaken to determine whether there are cyclophilin-like proteins expressed specifically or predominantly in the bovine retina.

EXPERIMENTAL PROCEDURES
Library Screening and cDNA Isolation-Two bovine cDNA libraries were constructed as described previously: library A with Ͻ1.5-kb average inserts and library B with Ͼ1.5-kb average inserts (Ferreira et al., 1993). Library A was screened under reduced stringency hybridization conditions using as probe a ϳ460-bp Sau3AI fragment of Drosophila ninaA cDNA, which corresponds to amino acid 29 to 183 of the NinaA protein (Schneuwly et al., 1989;Shieh et al., 1989). The SauAI fragment * This work was supported by Grant EY00033 from the National Eye Institute. Sequence analysis was supported by Grant AI27713 from the Purdue AIDS Center Laboratory for Computational Biochemistry. 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.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EMBL Data Bank with accession number(s) L41691 and L41692 § Supported in part by a North Atlantic Treaty Organization postdoctoral fellowship. Present address: Southwestern Medical School, NB4.218 5323 Harry Hines Blvd., Dallas, TX 75325-9111.
Southern blot analyses were carried out on these cDNA clones at the same (20% formamide, 42°C) and higher stringency hybridization conditions (30% formamide, 42°C) using the same SauAI fragment as probe. From these analyses, 11 of the cDNA clones were selected as being most likely to share high sequence homology with the ninaA probe. Partial sequence analysis of these 11 clones showed that they fell into five non-overlapping groups, A-E, three of which corresponded to either previously characterized cyclophilins or a new isoform of a previously identified one. These clones were not pursued any further except as probes in RNA blot analysis for comparison (Fig. 5,D and E). Instead, we focused on one of the remaining two groups, group C, represented by clones C47 and C13 (Fig. 1C), because of preliminary indications that they might be retina-specific.
To obtain the 5Ј sequences missing in clones C47 and C13, both library B constructed in this work and another bovine cDNA library provided by Dr. Jeremy Nathans (The Johns Hopkins University Medical School) were screened under high stringency hybridization conditions using clone C47 as probe. These screens yielded 20 additional positively hybridizing cDNA clones. The clones isolated from the Nathans library (labeled L in Fig. 1C), which had originally been cloned into gt10, were subcloned into the vector pGEM-3Z (previously called pGEM-Blue; Promega).
RNA Blot Analysis-Poly(A) ϩ RNA was prepared from bovine tissues (Chomczynski and Sacchi, 1987) obtained from a local slaughterhouse, fractionated on a 1% formaldehyde gel, 10 g/lane, blotted onto a nitrocellulose membrane (Amersham Corp.) (Sambrook et al., 1989), and probed with cDNA fragments a, b, c, and h (Fig. 1D). Also used as probes for comparison were cDNA clones from two of the other four groups of ninaA-hybridizing cDNA clones isolated. These clones corresponded to the previously characterized bovine thymus cyclophilin (Harding et al., 1986) and a bovine homolog of human- (Price et al., 1991) and chicken- (Caroni et al., 1991) secreted cyclophilin cDNA. The probes were labeled with 800 Ci/mmol of [␣-32 P]dCTP (Amersham) Vogelstein, 1983, 1984). The blots were prehybridized for 3 h and hybridized for 16 h, both at 65°C (Ferreira et al., 1993).
DNA Sequencing and Analysis-Double-stranded cDNA clones were sequenced by the dideoxy chain-termination method (Sanger et al., 1977) using the Sequenase system (U. S. Biochemical Corp.). For each cDNA clone to be sequenced, several overlapping fragments were prepared and subcloned, and each fragment was sequenced from both ends. Appropriate primers were used to fill in any gaps that might result from this approach. Clones C47 and C13 were completely sequenced in both strands, and 13 of the 20 subsequently isolated clones were partially sequenced. DNA and protein sequences were compared with those in the data bases using the GCG program (Devereux et al., 1984) and the programs provided by the Molecular Biology Computer Research Resource Center (MBCRR) at Harvard Medical School (now at Boston University) (Smith and Smith, 1990).
Tissue in Situ Hybridization-Approximately 12-m thick bovine retinal tissue sections were cut on a cryostat, as described in Ferreira et al. (1993). Sense and antisense cRNA probes were prepared from cDNA fragment c or h (Fig. 1D), which had been cloned into the pGEM-3Z vector, using either T7 or SP6 RNA polymerase (Stratagene). The probes were labeled with [␣-35 S]CTP.
Preparation and Purification of Retinal Cyclophilin-GST Fusion Proteins-The recombinant bovine retinal cyclophilin-GST fusion proteins were constructed using the bacterial expression vector pGEX-KG (Guan and Dixon, 1990). The three fusion constructs Cyp 1K-GST, Cyp BS-GST, and Cyp HS-GST were prepared, respectively, from the Klenowtreated BamHI-StuI fragment, the T4 polymerase-treated BstXI-StuI fragment, and the Klenow-treated HincII-StuI fragment of retinal cyclophilin cDNA (see Fig. 1A). These fragments were blunt-end ligated in frame into the pGEX-KG vector that was linearized by XbaI or EcoRI digestion and treated with Klenow polymerase and phosphatase. Two other cyclophilin-GST constructs, Cyp BH and Cyp B/S/N, were prepared for antibody production. The Cyp BH construct was prepared by a HindIII digestion followed by a partial BstXI digestion of Cyp 1K-GST, and the Cyp B/S/N construct was prepared by a NcoI digestion of Cyp BS-GST. Both constructs were then Klenow treated and blunt-end ligated into the pGEX-KG vector. All five fusion constructs were transformed into Escherichia coli. Transformed clones were checked for the size and orientation of the subcloned inserts and for the production of fusion proteins of expected size after treatment with 1 mM isopropyl-1thio-␤-D-galactopyranoside for ϳ4 h. For protein analysis, 30 l of bacterial cultures were pelleted in a microcentrifuge tube and resuspended in 30 l of 2 ϫ SDS-polyacrylamide gel electrophoresis sample buffer (1 ϫ SDS-polyacrylamide gel electrophoresis sample buffer: 50 mM Tris-Cl, pH 6.8, 100 mM dithiothreitol, 2% SDS, 10% glycerol). Samples were boiled for 5 min and loaded onto a 10% SDS-polyacrylamide gel (Laemmli, 1970) for analysis.
Large scale preparations of cyclophilin-GST fusion proteins were performed as described by Guan and Dixon (1990). Recombinant GST fusion proteins were purified by adsorption to glutathione-agarose (Sigma) under non-denaturing conditions and specific cleavage of GST by thrombin. To check for possible contaminations and/or degradation, protein samples were overloaded and run on a 10% reducing SDSpolyacrylamide gel and stained with Coomassie Blue (data not shown). The amounts of purified proteins were determined by amino acid composition analysis (Deutsher, 1990) for the proteins used in the PPIase kinetic and CsA binding studies and by spectrophotometric analysis (Deutscher, 1990) for the proteins used in antibody production.
PPIase and CsA Binding Activities of Bovine Retina Cyclophilin Constructs-The ability of the bovine cyclophilin constructs to catalyze the cis-trans-isomerization of the proline amide in a tetrapeptide substrate was evaluated by a modification of previously reported procedures (Fischer et el., 1989;Harrison and Stein, 1990). The reactions were monitored by a Cary model 2200 spectrophotometer in a 2-ml cuvette kept at a constant temperature of 8°C. A 0.1-ml cyclophilin sample and 1.84 ml of 78 mM ␣-chymotrypsin were added to the cuvette first. The bovine retinal cyclophilin constructs were tested at a concentration of 100 nM, and purified bovine thymus cyclophilin, used as positive control, was examined at a concentration of 1 nM. The reaction was initiated by adding 0.06 ml of 2.6 mM solution of the peptide cis-Suc-Ala-Ala-Pro-Phe-p-nitroanilide. The hydrolysis of the peptide was monitored by collecting the absorbance at 400 nm for 5 min with an IBM PC 52 microcomputer interfaced to the spectrophotometer. The data were analyzed by non-linear curve fitting to the equation, OD ϭ A(1 Ϫ e Ϫkt ) ϩ C and normalized by subtracting the calculated value of C from the absorbance readings. The first-order rate constants, k, were calculated using the Spectra Calc software (Spectra Calc, Galactic Industries, Salem, NH).
CsA binding was determined by a modification of the Sephadex LH-20 column assay described by Handschumacher et al. (1984). Briefly, 150 l of 20 mM Tris buffer, pH 7.2, containing 7.5% fetal bovine serum and 5 mM 2-mercaptoethanol, 10 l methanol or unlabeled CsA (Sandimmune), 20 ml of [ 3 H]CsA (5.0 Ci/ml, 0.515 g/ml in 40% ethanol) (DuPont NEN), and a 20-l cyclophilin test sample (approximately 0.50 M) were added to silicon-coated microcentrifuge tubes. The tubes were gently agitated and incubated at room temperature for 30 min. A 150-ml sample from each tube was then loaded onto a minicolumn containing 1.8 ml of Sephadex LH-20 resin (Pharmacia Biotech Inc.), which had been preequilibrated with 20 mM Tris buffer, pH 7.2, and 0.02% NaN 3 . The columns were eluted with 20 mM Tris buffer plus 7.5% fetal bovine serum and 5 mM 2-mercaptoethanol, and 1-ml fractions were collected. The first 1-ml fraction which contained the CsA⅐protein complex was assayed for radioactivity in 10 ml of Ready Protein Plus (Beckman Instruments, Inc., Fullerton, CA). Unlabeled CsA was used at 5 M concentration for displacement of [ 3 H]CsA binding activity of cyclophilin constructs.
Production of Antibodies-Each of the four New Zealand White rabbits used (two for each of the two antigens) was injected intradermally (Harlow and Lane, 1988) at a total of 10 sites with approximately 10 -50 g/site of purified Cyp BH or Cyp B/S/N recombinant proteins emulsified in Hunter's TiterMax TM adjuvant (CytRx Corp.). Booster shots were administered 2 and 4 weeks after the initial immunization. After approximately 12 weeks of immunization, the rabbits were bled, and the antisera were tested at various dilutions by immunodot staining of the recombinant proteins (antigens) and unrelated proteins (negative controls) and also by Western blot analysis of recombinant proteins and protein retinal extracts. Results of these analyses were consistent with their specificity for the respective antigens. Two of the four antisera generated, Ab Cy321 and Ab Cy318, directed against Cyp B/S/N and Cyp BH, respectively, were used for immunocytochemistry.
Immunocytochemistry-Frozen bovine retinal sections were prepared as described in Ferreira and Pak (1994). Dark-adapted eye cups were fixed in 2% paraformaldahyde, infiltrated with 30% sucrose, embedded in Tissue-Tek, frozen in liquid nitrogen, and sectioned on a cryostat (4 -5 m). The sections were mounted on gelatin-subbed slides and immunostained as described in the HistoMark kit (Kirkegaard & Perry Laboratories) except that all solutions contained 0.1-0.4% saponin (Sigma). Coverslips were placed on the slides with a drop of 80% glycerol in phosphate buffer, and slides were examined with a Zeiss photomicroscope equipped with Nomarski optics. "Mechanical teasing" of sections was achieved by gently rubbing the coverslip against the slide. The specificity of immunostaining was tested by omitting the primary antibodies, substituting the primary antiserum with preimmune serum, and/or blocking the primary antibodies with the respective antigen. Results of these tests indicated expected specificity.

Molecular
Characterization of cDNA Clones-Of the five nonoverlapping groups of cDNA clones isolated from screening a bovine cDNA library with a ninaA cDNA fragment as probe, we focused on Group C, consisting of clones C13 and C47 ( Fig.  1) and the 20 additional clones isolated using C47 as probe (see "Experimental Procedures").
Complete sequence analysis of clones C13 and C47 and partial sequence analysis of 13 of the 20 subsequently isolated clones showed that these cDNA clones fall into two major classes, type I and type II. Type I cDNA, represented by clone C14, is 1528 nucleotides long and contains a complete open reading frame with two potential methionine start codons at ϩ1 and ϩ55 ( Fig. 2A). The sequence surrounding the first Met codon displays downstrean mRNA secondary structures characteristic of eukaryotic translation start sites, while the sequence surrounding the second Met codon shows both the downstream secondary structures and the Kozak consensus sequence (GCCA/GCCAUGG) for translation start sites (Kozak, 1991a(Kozak, , 1991b. Several stop codons are found upstream of these putative translation initiation sites in all possible reading frames. Type II cDNA is represented by the remaining clones (C8, C9, C12, C15, etc.). These two types of cDNA differ from each other in the 5Ј sequence upstream of the adenine residue at position Ϫ158 (filled arrowhead in Fig. 2, A and B) but are identical beginning with this residue. In contrast to type I cDNA, the open reading frame of type II cDNA extends all the way to the 5Ј end of the available sequence (Fig. 2B). Type II cDNA has a potential acceptor splice site (CAG) (Cech, 1986) at position Ϫ161, just where the 5Ј sequence unique to this cDNA ends.
Structure of the Deduced Bovine Proteins-Conceptual translation of cDNAs, thus, generates two, or possibly three, potential isoforms of the bovine protein: type I with 234 and/or 252 amino acid residues, depending on which translation start codon is utilized, and type II with over 1085 amino acid residues. An approximately 150-amino acid long "cyclophilin" (PPIase/CsA-binding) domain, which shows sequence homology to other cyclophilins (Fig. 3), is found in the COOH-terminal region common to both types of protein (Figs. 1, E and F). These bovine proteins are characterized by extended NH 2 -terminal sequences not found in other cyclophilins reported to date (Fig.  1F). The NH 2 -terminal sequences are approximately 100 and over 900 residues long for type I and type II proteins, respectively. Type I cyclophilin, for which the NH 2 -terminal sequence has been fully characterized, lacks a cleavable signal peptide sequence, typically found at the amino termini of many cyclophilins. In addition, unlike other cyclophilins reported to date, the COOH terminus of these bovine cyclophilins ends with a potential prenylation motif, CXXX ( Fig. 2A) (Clarke, 1992).
The amino-terminal region of type I cyclophilin contains a putative amphipathic, ␣-helical, coiled coil structure that resembles the leucine zipper (Alber, 1992;O'Shea et al., 1991). A stretch of 30 -40 amino acids beginning with Phe-7 forms a heptad repeat, (abcdefg) n , with positions "a" and "d," corresponding to the same face of the helix, occupied mostly by hydrophobic residues, and some of the residues at position "d" are leucines. Type II protein has a very long NH 2 -terminal region (Ͼ900 amino acids), the full extent of which has not yet been determined. The region contains at least two sets of repeat domains: V1 and V2 and W1 and W2 (Fig. 1, E and F). V1 and V2 are about 150 amino acids each and display 55% sequence identity with each other (Fig. 4A), while W1 and W2 are about 100 a.a. each and show 34% identity (not shown). The much shorter NH 2 -terminal region of Type I protein contains only the COOH-terminal half of V2 (Fig. 1, E and F). Another region of interest is the cyclophilin-like domain (CLD) of about 150 amino acids found between V1 and W1 of type II protein (Fig. 1, E and F). Although the overall sequence identity of this domain with the cyclophilin domain is only about 24%, most of  B,BamHI;Bx,BstXI;H,HpaI;Ha,HaeIII;Hc,HincII;N,NcoI;P,PstI;R,RsaI;S,Sau3AI;St,StuI. C, cDNA clones used to assemble type I and type II cDNA sequences. C14 is the only type I clone. Clones C10, C13, and C47 are incomplete at the 5Ј end and contain only sequences that are common to both type I and type II cDNAs. All remaining clones are type II clones. Clone 17 contains a deleted region when compared to other clones. This is indicated by two pairs of slash marks interrupting the line. Clone L6 was isolated from the Nathans library. All the other clones labeled C were isolated from the libraries constructed in this work. D, restriction fragments of cDNA clones used to probe RNA blots (Fig. 5) and/or to synthesize riboprobes for tissue in situ hybridization (Fig. 6). Probes a-c correspond to different regions of 5Ј-non-coding sequence specific to type II cDNA. Probe h corresponds to the HpaI 5Ј end fragment of clone C14, containing the 5Ј-non-coding sequence specific to type I cDNA. E, schematic representations of bovine retinal cyclophilin cDNA structure. The thick and thin bars represent coding and noncoding regions, respectively. V1 and V2 (dark-shaded boxes) and W1 and W2 (dashed boxes) are two sets of repeat domains. CLD (light-shaded box) represents the cyclophilin-like domain. The black box represents the PPIase/CsA-binding (cyclophilin) domain found in the region shared by the two types of cDNA and is conserved in all known cyclophilins. Type II cDNA has three alternative poly(A) ϩ adenylation signals (small black boxes in the COOH-terminal region). The type II cDNA clones analyzed utilize either the first (IIA) or the last (IIB) signal. The only type I cDNA identified (C14) utilizes the first signal (I). F, structural comparison of the deduced bovine retinal cyclophilins with other cyclophilins. Types I and II are the bovine isoforms reported in this work. Type III is a recently reported protein of high molecular weight with a putative transmembrane domain (Anderson et al., 1993). Type IV has both a signal sequence and a COOH-terminal transmembrane domain and is represented by the Drosophila NinaA protein (Schneuwly et al., 1989;Shieh et al., 1989) and yeast-4 cyclophilin (Koser et al., 1990). Type V has a signal sequence but no transmembrane domain (e.g. yeast-2, rat-C, and Neurospora cyclophilins) (Friedman and Weissman, 1991;Koser et al., 1990;Tropschug et al., 1988). Type VI has neither a signal nor a transmembrane sequence (e.g. human A, yeast-1, and Drosophila cyp 1 cyclophilins) (Haendler et al., 1987(Haendler et al., , 1989Stamnes et al., 1991). SS, signal sequence; TM, transmembrane domain. Note that schematic representations of cyclophilin structures in E and F are not aligned with A-D or with each other. the structural amino acids (Gly, Pro, and Ala) present in the cyclophilin domain are conserved at corresponding positions in this domain (Fig. 4B).
Comparison between the Bovine Proteins and Other Proteins- Fig. 3 compares the sequence of the COOH-terminal cyclophilin domain of the deduced bovine retinal proteins with the corresponding regions of 11 other members of the cyclophilin family. Within a region of about 160 amino acids (amino acid 88 -252 in Fig. 2A), sequence identity between the bovine proteins and other cyclophilins ranges from 26% with E. coli cyclophilin (Kawamukai et al., 1989) and 45% with NinaA protein of Drosophila (Schneuwly et al., 1989;Shieh et al., 1989) to 67% with another Drosophila cyclophilin, cyclophilin-1 . The 160-amino acid region being compared corresponds to less than 15% of the type II cyclophilin but to essentially the entire protein for many of the other cyclophilins (Fig. 3).
About a dozen highly conserved amino acid residues in this region have been identified as being important in CsA binding (# in Fig. 3) or in both CsA binding and PPIase activity (ϩ in Fig. 3) from x-ray crystallographic studies of human cyclophilin A complexed with CsA or a model peptide substrate (Kallen and Walkinshaw, 1992;Ke et al., 1993;Pflugl et al., 1993). Three of these are nonconservatively substituted in the bovine retinal cyclophilins: Val-149 (*ϩ in Fig. 3), Arg-193 and Arg-209 (*# in Fig. 3), which correspond to Met-61, Ala-103, and Trp-121 of human cyclophilin A, respectively (Haendler et al., 1987). Two of these residues, Met-61 and Ala-103, are also nonconservatively substituted in the NinaA protein (Fig. 3).
Protein data base search also revealed that domains V1 and V2 are, respectively, 43 and 52% identical in sequence (Fig. 4A) to the Hypothetical Protein A encoded by the HTF-9A (HpaII Tiny Fragments) locus of the mouse (Bressan et al., 1991). The COOH-terminal two-thirds of V2 domain is 41% identical in sequence to the COOH-terminal region of NUP2, a novel yeast nucleoporin with functional overlap with other proteins of the nuclear pore complex (Loeb et al., 1993) (Fig. 4A). In addition, several local stretches of similarity to DNA-RNA-binding do-FIG. 2. Nucleotide and deduced amino acid sequences of type I and type II cDNAs. A, type I cDNA sequence and the deduced protein sequence. Two potential translation start sites are either circled or boxed in black. In the coding region, amino acids for the deduced protein are shown directly below the corresponding codons. The numbering of both nucleotide and amino acid residues starts with the first methionine start codon shown. The number to the right of each row corresponds to the last nucleotide or amino acid residue in the row. Type I cDNA differs from type II cDNA in the 5Ј end sequence up to residue A at position Ϫ158, marked with the filled arrowhead. Type I and type II cDNA sequences are identical from this residue on. The three potential polyadenylation signals are underlined. Type I cDNA uses the first polyadenylation signal, and the poly(A) ϩ tail is added after residue A at nucleotide position ϩ967 marked with an open arrowhead. The 3Ј-non-coding sequence following this residue was found in type II cDNA clones analyzed but shown here with type I cDNA sequence for convenience. B, type II cDNA sequence is shown up to nucleotide position Ϫ1 along with the deduced protein sequence. The 5Ј sequence up to residue A at position Ϫ158, marked with the filled arrowhead, differs from that of type I cDNA. Unlike in type I cDNA, the reading frame is open all the way to the 5Ј end of the available sequence. Two of the potential polyadenylation signals shown in A, the first and the third, were found utilized in cDNA clones of this class examined. mains or motifs are found embedded within V1 and V2 and in surrounding regions.
RNA Blot Analysis and Tissue in Situ Hybridization-RNA blot analyses were carried out using probes that are specific to type I (h in Fig. 1D) or type II (a-c in Fig. 1D) cDNA. Probes a and h recognize, respectively, 10 kb (Fig. 5A) and 3.1 kb (Fig.  5B) transcripts expressed specifically in the retina. These transcripts are very much larger than most other cyclophilin transcripts and are compatible with the deduced type I and type II proteins in size. Probe h detects, in addition to the retinaspecific 3.1-kb transcript, a fainter cerebellum-specific 4.6-kb transcript (Fig. 5B). Although probes b and c, like probe a, are also specific for type II cDNA (Fig. 1C), they detect, in addition to the 10-kb transcript, a 3.3-kb retina-specific transcript (Fig.   5C), suggesting the presence of at least another retina-specific isozyme.
To compare the RNA expression profiles of the new retinaspecific bovine cyclophilins with those of previously characterized cyclophilins, RNA blot analyses were carried out, under the same conditions, using as probes other classes of bovine cyclophilin cDNA clones isolated in this work ("Experimental Procedures"). One of them was identical in sequence to the previously characterized bovine thymus cDNA (Harding et al., 1986), and the other was very nearly identical in sequence to the human- (Price et al., 1991) and chicken- (Caroni et al., 1991) secreted cyclophilin cDNAs. As seen in Fig. 5D, the bovine thymus cyclophilin probe recognizes a 1.15-kb transcript expressed at least 100 -200 times more abundantly and in a wider FIG. 3. Amino acid sequence conservation between the bovine retinal cyclophilins and other cyclophilins. Amino acid sequences within the PPIase/CsA-binding domain are compared between the bovine retinal cyclophilins and 11 other cyclophilins. The bovine retinal cyclophilin sequence is shown at the top of the alignment. Dots are placed wherever the amino acids in other sequences are identical to the bovine retinal sequence, and dashes represent gaps introduced to obtain optimal alignment. Highly conserved amino acids in this domain that are thought to be important in both PPIase activity and CsA binding are marked with ϩ, and additional conserved amino acids implicated in CsA binding alone are marked with # (Kallen and Walkinshaw, 1992;Ke et al., 1993;Pflugl et al., 1993). Among these amino acids, those that are not conserved in the bovine retinal cyclophilins are indicated by * over ϩ or #. Two other nearby amino acids that are highly conserved among other cyclophilins but not in the bovine retinal cyclophilins are marked with filled arrowheads. Amino acid identity in this domain between the bovine retinal cyclophilins and each of the other cyclophilins is shown at the end of each sequence. Bov Ret, bovine retinal cyclophilin; Dro ninaA, Drosophila NinaA cyclophilin (Schneuwly et al., 1989;Shieh et al., 1989); Yeast-4, yeast cyclophilin 4 (Franco et al., 1991); Hum NK-TR, human natural killer cells cyclophilin (Anderson et al., 1993); Dro cyp-1, Drosophila cyclophilin 1 ; Hum-A, human cyclophilin A (Haendler et al., 1987); Hum-SCYLP, human-secreted cyclophilin (Spik et al., 1991); Rat-C, rat cyclophilin C (Friedman and Weissman, 1991); Yeast-1 and 2, yeast cyclophilins 1 and 2 (Koser et al., 1990;Haendler et al., 1989); Neurosp., Neurospora cyclophilin (Tropschug et al., 1988); E. coli, E. coli cyclophilin (Kawamukai et al., 1989).

FIG. 4. Sequence comparison between domains within bovine retinal cyclophilin II and other proteins.
A, amino acid sequences are compared among V1 and V2 domains of bovine retinal cyclophilin type II, the mouse Hypothetical Protein A, HTF-9A (Bressan et al., 1991), and the nucleoskeletal protein, NUP2 (Loeb et al., 1993). The two percentage figures following the sequences correspond to amino acid identities between HTF-9A and V1 and V2, respectively. The next figure is percent identity between V1 and V2, and the last figure corresponds to percent identity between V2 and NUP2. B, amino acid sequence comparison between the cyclophilin (PPIase/CsA-binding) domain and the cyclophilin-like domain (CLD) of type II bovine retinal cyclophilin. Conserved structural amino acids are shown in boldface type. Regions for which secondary structural features have been identified from x-ray crystallographic analysis of human cyclophilin A (Ke et al., 1991) are so indicated: B, ␤-sheet; T, turns; H, ␣-helices. Throughout the figure, the first and last amino acid positions are shown for each sequence being compared, dots are used to indicated amino acids that are identical to the ones immediately above them, dashes are used to represent gaps introduced to obtain optimal alignment, and percent amino acid identities are shown in boldface type following the sequences. range of tissues than the retina-specific cyclophilin transcripts. Similarly, the bovine homolog of human-and chicken-secreted cyclophilin cDNAs recognizes a 1.0-kb transcript expressed abundantly and ubiquitously (Fig. 5E).
Antisense RNA probes synthesized from cDNA fragments h and c, which are specific to type I and II cDNAs, respectively, (Fig. 1D) were used for in situ hybridizations to tissue sections of the bovine retina. With either probe, hybridization signals were found in all nuclear layers of the retina: the outer nuclear layer, consisting of photoreceptor nuclei, the inner nuclear layer, and the ganglion cell layer (Fig. 6). Sense RNA probes did not yield any specific hybridization signal.
Immunolocalization of Type I and II Proteins-Polyclonal antisera were raised in rabbits against two different protein fragments of the bovine retinal cyclophilins: Ab321 against the COOH-terminal half of V2 domain, which is present in both type I and type II cyclophilins, and Ab318 against the NH 2terminal end of V2, present only in type II protein (Fig. 1E). The two antisera showed similar immunostaining patterns in the retina. Shown in Fig. 7A is a retinal section stained with Ab321. Intense staining is found in the proximal margin of the photoreceptor layer, just distal to the external limiting membrane. Also seen are some intensely stained spots in the distal and proximal margins (arrows and arrowheads in Fig. 7A) of the inner nuclear layer and the ganglion cell layer. Staining is conspicuously absent in rod outer segment and outer (photoreceptor) nuclear layers. These results are consistent with those of tissue in situ hybridization showing hybridization of probes to all nuclear layers of the retina (Fig. 6).
Another view of photoreceptor staining shown in Fig. 7B begins to allow determination of the identity of photoreceptors being stained. The staining pattern suggests that both inner and outer segments of cones and possibly inner segments, but not outer segments, of some rods are stained. To better determine the identity of photoreceptors being immunostained, some retinal tissue sections were "mechanically teased" in an attempt to disperse the otherwise densely packed photoreceptor cells. Fig. 7, C-E, are examples of immunostaining obtained with such mechanically teased retinal sections shown at a higher magnification. Fig. 7C confirms that both inner and outer segments of cones and probably inner segments of some, but not all, rods are stained. Any rod outer segments that are isolated by mechanical teasing are completely devoid of staining (Fig. 7D), while cones that are similarly isolated show staining usually in both inner and outer segments (not shown). A small number of cones, however, exhibit staining only in the inner segment (Fig. 7E). To summarize the results obtained in the photoreceptor layer: 1) rod outer segments and photoreceptor nuclei are not stained; 2) inner segments of some rods may be stained; 3) most cones have staining in both inner and outer segments; 4) a small subset of cones display staining only in the inner segment.
PPIase and CsA Binding Activities of Bovine Proteins Expressed in Vitro-PPIase and CsA binding activities were tested in vitro using three different recombinant protein constructs, each containing the COOH-terminal cyclophilin/PPIase domain. They were generated from cDNA fragments corresponding to (i) amino acid no. Ϫ104 to ϩ252 (CYP 1K), (ii) 19 -252 (CYP BS), and (iii) 70 -252 (CYP HS) of type II cyclophilin (Fig. 2). The second construct, CYP BS, is equivalent to the native form of type I cyclophilin if the second Met codon is assumed to be utilized in translation initiation ( Fig. 2A) and consists of the cyclophilin domain plus the COOH-terminal half of V2 domain (Fig. 8A and Fig. 1, E and F). CYP HS, the shortest of the three constructs, consists almost entirely of the cyclophilin domain, while CYP 1K, the longest of the three, has, in addition to the cyclophilin domain, the entire V2 domain (Fig. 8A).
PPIase activities of these cyclophilin constructs were determined by evaluating their ability to catalyze cis-trans-isomerization of the proline amide in the tetrapeptide substrate, cis-Suc-Ala-Ala-Pro-Phe-p-nitroanilide (Fischer et al., 1989;Harrison and Stein, 1990) (Fig. 8B). The bovine thymus cyclophilin (Fischer et al., 1989;Harding et al., 1986) was used as a positive control. Observed rate constants, k (obs) , were 0.932, 1.487, and 1.387 min Ϫ1 for CYP 1K, CYP BS, and CYP HS, respectively, when tested at a concentration of 100 nM. By contrast, purified bovine thymus cyclophilin had a k (obs) value of 3.372 min Ϫ1 at a concentration of 1 nM. Thus, these constructs exhibit only a modest level of PPIase activity, even when tested at a 100-fold higher concentration than control, and the largest construct CYP 1K has about 40% lower activity than the other two.
The CsA binding capacity of the bovine retinal cyclophilin FIG. 5. RNA blot analysis of bovine tissues. A, RNA blot hybridized with probe a specific for type II cDNA (Fig. 1D). The probe recognizes a 10.0-kb retina-specific transcript. B, RNA blot hybridized with probe h specific for type I cDNA (Fig. 1D). The probe recognizes a 3.1-kb retina-specific transcript and a fainter 4.6-kb transcript specific to the cerebellum. C, RNA blot hybridized with probes b or c (Fig. 1D), also specific for type II cyclophilin. These probes detect, in addition to the 10-kb transcript, a retina-specific 3.3-kb transcript. D and E, these blots were probed with cDNAs from two of the other four groups of ninaA-hybridizing cDNA clones isolated in this work. The probe in D corresponds to bovine thymus cyclophilin (Harding et al., 1986), and the probe in E corresponds to a bovine homolog of human (Price et al., 1991) or chicken-secreted cyclophilin (Caroni et al., 1991). An equivalent amount (ϳ10 g) of poly(A) ϩ RNA was loaded in each lane. RT, retina; BR, brain; CB, cerebellum; RPE, retinal pigment epithelium; KD, kidney; SP, spleen; SM, facial skeletal muscle. constructs was determined by the Sephadex LH-20 column assay ("Experimental Procedures"). As summarized in Table I, all three constructs bound much less [ 3 H]CsA than bovine thymus cyclophilin. For example, the [ 3 H]CsA binding of Cyp BS and Cyp HS at concentrations of 1.9 and 2.1 M, respectively, was substantially lower than that of bovine thymus cyclophilin at a 10-fold lower concentration of 0.2 M. Binding by the three constructs, however, was specific for CsA since unlabeled CsA displaced the [ 3 H]CsA binding by the three constructs.
Taken together, it is concluded that both the PPIase and CsA binding activities of the three constructs are substantially reduced compared to those of bovine thymus cyclophilin.

Key Properties Shared by the Retina-specific Bovine
Cyclophilins and the NinaA Protein-One of the main objectives of the present work was to identify bovine cyclophilins with properties similar to those of the NinaA protein of Drosophila. Although the newly identified bovine retinal cyclophilins are not structurally very similar to the NinaA protein, they nevertheless do share several key properties with the NinaA protein.
The NinaA protein is unusual among known cyclophilins because of its relatively low abundance (data not shown) and highly tissue-restricted expression. The expression of the ni-naA gene is confined to the photoreceptors , and its protein product is utilized only in a subset of photoreceptors (Larrivee et al., 1981;Stamnes et al., 1991). Similarly, the bovine retinal cyclophilins occur in relatively low abundance and are expressed only in the retina (Fig. 5), in sharp contrast to other cyclophilins described to date. Morover, although they are expressed in other retinal neurons as well as photoreceptors (Figs. 6 and 7), their expression in photoreceptors is restricted predominantly to cones (Fig. 7, B-E). Thus, NinaA of Drosophila and the new bovine retinal cyclophilins are the only members of the cyclophilin family with retina-specific expression identified to date, and they are both apparently utilized only, or mainly, in distinct subsets of photoreceptors.
In addition, the bovine retinal cyclophilins and NinaA protein both exhibit nonconservative substitutions of certain highly conserved amino acids that are implicated in PPIase and/or CsA binding activity in human cyclophilin A (Fig. 3). Three such substitutions are found in the PPIase domain of the bovine retinal cyclophilins, and two of these amino acids are also altered in the NinaA protein: Met-61 and Ala-103 of human cyclophilin A substituted for Val and Arg, respectively, in the bovine cyclophilins (marked *ϩ and *#, respectively in Fig.  3). Crystallographic analyses of human cyclophilin A complexed with a model substrate or CsA have shown that the Met-61 residue binds to the Ala-Pro peptide substrate and that the Ala-103 residue makes backbone contact with CsA (Kallen and Walkinshaw, 1992;Ke et al., 1993;Pflugl et al., 1993). It thus appears likely that nonconservative substitutions of these residues would reduce both PPIase activity and CsA binding and may be responsible for the reduced PPIase and CsA binding activities of the bovine cyclophilin constructs observed in vitro (Fig. 8B, Table I). Since these residues are also nonconservatively substituted in the NinaA protein, the PPIase activity of this protein is also likely to be reduced.
Finally, like the NinaA protein, the bovine retinal cyclophilins may be associated with the membrane at their COOH terminus. The NinaA protein has a hydrophobic, potential membrane-spanning domain near its COOH terminus (Schneuwly et al., 1989), and biochemical evidence suggests that it is, indeed, an integral membrane protein . Although no such potential membrane-spanning domain is found in the COOH-terminal region of the bovine retinal cyclophilins, their COOH-terminal sequence ends in the motif, CXXX. Proteins ending in the COOH-terminal motif, CAAX, are known to associate with the membrane by prenylation of the motif by either farnesyl or geranylgeranyl isoprenoid (Clarke, 1992). It has been reported that the motif CXXX can also be subjected to prenylation (Glenn et al., 1992).
Novel Structural Features of the Bovine Cyclophilins-The bovine retinal cyclophilins identified in this work display a number of unusual structural features. The most striking structural characteristic of these bovine cyclophilins is the presence of extended NH 2 -terminal sequences linked to the COOH-terminal cyclophilin domain (Fig. 1, E and F), making type II protein one of the largest cyclophilins ever reported. Among the several domains identified within the extended NH 2 -terminal region of type II protein is a pair of repeat domains V1 and V2 (Fig. 1, E and F) that share significant sequence identity with the protein encoded by the mouse HTF-9A gene (Bressan et al., 1991) (Fig. 4A). The observed sequence identity between V1 and V2 domains and almost the entire HTF-9A protein suggests that this bovine cyclophilin gene arose by a process of exon amplification and shuffling involving other gene(s), thus generating a gene encoding a mosaic protein made up of multiple functional modules.
Functional significance of the extended NH 2 -terminal sequences of these bovine cyclophilins remains to be elucidated. However, structural and functional domains and motifs identified within the NH 2 -terminal regions provide clues to potential functions of these regions. For example, the observed sequence identity between V1 and V2 domains and the protein product of the mouse HTF-9A gene (Fig. 4A) took on added significance with the recent identification of the HTF-9A-encoded protein. Coutavas et al. (1993) recently identified the likely effector protein for the nuclear GTPase, Ran/TC4, named Ran-binding protein-1 (RanBP1), and showed that it is identi- FIG. 6. Tissue in situ hybridization. A, a bright field micrograph of a bovine retinal section probed with an antisense cRNA probe synthesized from cDNA fragment h (Fig. 1D), specific for type I bovine retinal cyclophilin cDNA. B and C are, respectively, dark field micrographs of sections probed with antisense riboprobes synthesized from cDNA fragments, h and c, specific for type I and II cyclophilin cDNAs, respectively (Fig. 1D). ROS, rod outer segment; ONL, outer nuclear layer; OPL, outer plexiform layer; INL, inner nuclear layer; IPL, inner plexiform layer; GCL, ganglion cell layer. Scale bar, 50 m.
cal in sequence to the HTF-9A-encoded protein. Ran proteins are abundant nuclear GTPases of the Ras superfamily implicated in the regulation of nuclear trafficking of mRNAs and/or proteins required for the onset of mitosis. In addition to the sequence similarity, RanBP1 was reported to contain both RNA binding and leucine zipper-like regions (Bressan et al., 1991;Coutavas et al., 1993), also found in V1/2 domains of type II bovine cyclophilin (see "Results"). These findings suggested FIG. 7. Immunocytochemical localization of bovine retinal cyclophilins. A, bovine retinal section immunostained with the polyclonal antiserum Ab Cy321 and viewed in Nomarski optics. Curved arrows point to stained bipolar and/or horizontal cells in the distal margin of the inner nuclear layer, and arrowheads point to stained amacrine cells in the proximal margin of the inner nuclear layer. ROS, rod outer segment; RIS, rod inner segment; ELM, external limiting membrane; ONL, outer nuclear layer; OPL, outer plexiform layer; INL, inner nuclear layer; IPL, inner plexiform layer; GCL, ganglion cell layer; NFL, nerve fiber layer. Scale bar, 70 m. B, photoreceptor layer of a retinal section immunostained with Ab Cy321 and viewed by Nomarski optics. Black arrowheads point to cone inner and outer segments, and an open arrowhead points to rod outer segments. ROS, rod outer segment; COS, cone outer segment; CIS, cone inner segment. Scale bar, 70 m. C, photoreceptor layer of a mechanically teased bovine retinal section immunostained with Ab Cy321 and viewed in Nomarski optics at a higher magnification. Scale bar, 30 m. D, rod outer segment isolated by mechanical teasing and immunostained with Ab Cy321. It is completely devoid of staining. Scale bar, 10 m. E, cone photoreceptor isolated by mechanical teasing and immunostained with Ab Cy321. This is an example of occasional cones with staining in the inner but not the outer segment. Scale bar, 10 m. that the V1 and V2 domains function as the effector domains for Ras-related retinal protein(s). Recent findings make this suggestion even more compelling. In a report that appeared after the present paper was submitted, Beddow et al. (1995) announced the cloning of two additional mammalian Ran-binding proteins and the identification of a highly conserved Ranbinding sequence motif. The motif contained in one of their clones, Ab2-RBD, is almost 100% identical to V1 domain of type II bovine cyclophilin. In fact, the V2 domain has also now been shown to bind the Ran protein in a GTP-dependent manner. 2 PPIase Activities of Bovine Cyclophilin Constructs Expressed in Vitro-Bovine cyclophilin constructs expressed in vitro exhibit only a modest level of PPIase activity compared to that of purified bovine thymus cyclophilin (Fig. 8B). Since these constructs were expressed in a bacterial expression system, it might be argued that the observed low activities of these con-structs may be due to their bacterial origin. However, several in vitro-expressed cyclophilins have been shown to be fully active under similar conditions (Bergsma et al., 1991;Liu et al., 1990;Price et al., 1991;Schonbrunner et al., 1991;Spik et al., 1991). Thus, the low PPIase activities of these proteins likely are real and may be related to the nonconservative substitution of amino acids presumed to be important for PPIase/CsA binding activity as discussed above.
The PPIase activity of the largest of the three constructs, CYP 1K, is about 40% lower than that of the other two constructs (Fig. 8B). Since the only difference between this construct and another tested, CYP BS, is the presence of the NH 2 -terminal half of the V2 domain in CYP 1K (Fig. 8A), this region may have an inhibitory influence over the cyclophilin/ PPIase domain.
Possible Biological Functions of the Bovine Retinal Cyclophilins-The NinaA protein is required for functional expression of a subset of rhodopsin (R1-6 and probably ocelli) in Drosophila (Larrivee et al., 1981;Stamnes et al., 1991). The findings that the ninaA gene encodes a cyclophilin (Schneuwly et al., 1989;Shieh et al., 1989) and that the enzyme peptidyl prolyl cis-trans-isomerase is a cyclophilin (Fischer et al., 1989;Takahashi et al., 1989) led to the suggestion that the primary role of the NinaA protein is to catalyze proper folding of the subset of opsin (Schneuwly et al., 1989;Shieh et al., 1989;Stamnes et al., 1991). Immunolocalization of the NinaA protein in the endoplasmic reticulum and inhibition of opsin transport out of the endoplasmic reticulum by ninaA mutations were taken as supporting the above hypothesis (Colley et al., 1991). Although it seemed likely that mammalian opsins would also require a cyclophilin in a similar role, no retina-specific candidate mammalian cyclophilin that might fulfill this role had been previously identified. In view of the several key properties that they share with the NinaA protein, including retina-specific expression, the retinal cyclophilins herein described are the prime candidates for cyclophilins with such a role in mammalian photoreceptors.
However, the idea that these retina-specific cyclophilins act solely as folding catalysts for opsins may require revision. For one thing, PPIase activities of the retina-specific bovine cyclophilins, expressed in vitro are about two orders of magnitude lower than that of purified bovine thymus cyclophilin (Fig. 8B). In fact, reduced PPIase activity may be a characteristic of both the bovine retinal cyclophilins and NinaA protein because of nonconservative substitutions of highly conserved amino acids likely to be important in PPIase/CsA binding activities (see "Results"). Moreover, molecular studies of D. ninaA mutants have shown that protein alterations caused by ninaA mutations need not be within the PPIase/CsA-binding domain to cause severe reductions in rhodopsin levels (Ondek et al., 1992;Schneuwly et al., 1989;Shieh et al., 1989), suggesting that PPIase activity and rhodopsin levels are not directly linked. These observations, taken together, suggest that the action of the retina-specific cyclophilins on opsins probably is not simply as PPIases. For example, these proteins could act on opsins as a chaperone, possibly in addition to their role as PPIases. Such a dual role of cyclophilin as both a chaperone and PPIase has been reported in the folding of carbonic anhydrase and ribonuclease T1 (Freskgard et al., 1992;Schonbrunner and Schmidt, 1992). Recently, Baker et al. (1994) demonstrated that the NinaA protein forms a specific stable complex with R1-6 rhodopsin in Drosophila in vivo and that the maturation of rhodopsin is quantitatively linked to NinaA. They suggest that the NinaA protein acts as a chaperone or, alternatively, both as a chaperone and PPIase for rhodopsin.
The bovine retinal cyclophilins reported here probably rep-2 P. A. Ferreira, unpublished results.
FIG. 8. PPIase activity of bovine retinal cyclophilin constructs. A, schematic diagram of three bovine retinal cyclophilin constructs. B, hydrolysis of the substrate tetrapeptide, cis-Suc-Ala-Ala-Pro-Phe-p-nitroanilide, by the cyclophilin constructs was monitored spectrophotometrically at 400 nm. The observed rate constants, k, in min Ϫ1 were 0.932, 1.487, and 1.387 for CYP 1K, CYP BS, and CYP HS, respectively, when tested at a concentration of 100 nM. In contrast, purified bovine thymus cyclophilin had a k value of 3.372 min Ϫ1 at a concentration of 1 nM. Retina-specific Bovine Cyclophilin resent specialized cyclophilins with multifunctional domains. We have suggested above that V1 and V2 domains of type II bovine cyclophilins may be the binding domains for Ras-related GTPase(s). It is thus conceivable that V1 and V2 domains, through their interaction with a Ras-related protein, such as Ran, provide the signal for proper targeting of rhodopsin while the cyclophilin domain of the same protein interacts with opsin/ rhodopsin directly to escort it through the secretory pathway in a properly folded state. Human photoreceptors appear to be very sensitive to mutations in rhodopsin and other retina-specific genes. For example, mutations in the rhodopsin gene (Berson, 1993) have been found to cause structural deterioration of photoreceptors leading to blindness. It thus seems possible that mutations in the gene encoding the human homolog(s) of the bovine retinal cyclophilins would lead to specific degeneration of cones. Specific degenerative changes in cones do occur in humans and are serious causes of visual impairment.