Results

Cloning of mouse mr-s
In order to identify novel mouse genes preferentially expressed in the developing retina, we screened the National Institute for Biotechnology Information (NCBI) database, UniGene, using Digital Differential Display (DDD) and found EST fragments which are frequently present in mouse retinal cDNA libraries. We found that one clone in these cDNAs encodes a protein containing a SAM domain related to that of polyhomeotic protein. A PCR fragment corresponding to this mouse clone was used to screen a mouse P0-P3 retinal cDNA library to obtain a full-length cDNA clone. Sequence analysis showed that this cDNA was a novel gene encoding a SAM domain-containing protein. We referred to this protein as mr-s (major retinal SAM domain protein). As shown in Fig. 1A, a translation initiation codon is present in the same open reading frame as the SAM domain. This initiation site shows similarity to the consensus sequence proposed by Kozak [32] including the presence of the highly conserved purine at position -3. The stop codon of the predicted mr-s protein is also indicated in Fig. 1A. The amino acid sequence of the SAM domain of mr-s protein (Fig. 1A, boxed sequence) exhibits homology with SAM domains of EphB2, EphA4, MPH1, TEL and Smaug (Fig. 1B). By phylogenetic analysis, the SAM domain of mr-s is most closely related to that of Mph1/Rae28, a mouse homolog of ph (Fig. 1C). Mouse mr-s protein is conserved in rat, human, chick and zebrafish, which display 91%, 70%, 36% and 26% identity with mouse mr-s, respectively (Fig. 1D). The SAM domains of rat, human, chick and zebrafish mr-s protein are highly conserved and display 96%, 90%, 76% and 72% identity with the SAM domain of mouse mr-s protein. The chromosomal localizations of mouse and human mr-s genes were determined by searching the mouse and human genome databases (NCBI), respectively. Mouse mr-s is mapped to chromosome 4E2, and human MR-S is mapped to chromosome 1p36.33. LCA is the most common cause of inherited childhood blindness. Human MR-S maps in the vicinity region of the LCA9, recently identified as a new locus for LCA [33].
Figure 1  mr-s nucleotide and amino acid sequences. (A) mr-s nucleotide and amino acids sequences. Boxed amino acids are the SAM domain sequence and the dashed box indicates a putative nuclear localization signal. The underline indicates a putative polyadenylation termination signal. (B) Alignment of SAM domain sequences for SAM domain-containing proteins. The five alpha helices are marked H1-H5. Conserved amino acid residues are shown with a dark shadow and functionally similar residues are shown with a light shadow. The sites that were targeted for mutagenesis are indicated by arrows. (C) Phylogenetic tree of SAM domain-containing proteins. Amino acid sequences were analyzed by the neighbor-joining method in MacVector 7.2. Branch lengths reflect the mean number of substitutions per site. (D) Schematic comparison of the amino acid sequences for mouse, rat, human, chick and zebrafish mr-s proteins. The percent similarity of the SAM domains and other regions to the corresponding regions of the mouse protein is shown. Overall sequence similarity with the mouse protein is shown on the right.

Expression of mr-s in the developing retina and the pineal gland
To investigate mr-s expression, we first performed whole mount in situ hybridization in mouse embryos. No hybridization signal was detected at E9.5, E10.5 and E11.5 with an mr-s probe (data not shown). We then investigated the expression of the mr-s gene in the developing retina by section in situ hybridization (Fig. 1A–G). No significant signal was detected in the developing retina at E13 (Fig. 2A). A weak signal was initially detected in the outer aspect of the neuroblastic layer (NBL), a presumptive photoreceptor layer at E18 (Fig. 2B, arrow). At P3, an mr-s transcript was clearly detected in the developing photoreceptor layer (Fig. 2E). At P6, mr-s showed peak expression in the photoreceptor layer (Fig. 2F). This pattern correlates with the rapid increase in cells expressing rhodopsin and other phototransduction genes between P6-P8 [34-37]. At P9, the intensity of the mr-s signal was significantly reduced but was localized to the outer nuclear layer (ONL) (data not shown). Faint expression of mr-s mRNA was observed in mature photoreceptors in the adult retina (Fig. 2G).
Figure 2  mr-s expression in developing mouse retina and pineal gland. (A-G) mr-s expression during development of the mouse retina. The in situ hybridization signal of mr-s was not detected at E13 (A). The signal (arrow) was first detected in the outer aspect of NBL at E18 (B). A strong mr-s signal was detected in outer layer of the retina at P3-P6, and then the signal decreased in the adult retina (E-G). Control with the sense probe in E18 retina is shown (C). Scale bar, 100 μm. (H) Northern blot analysis of mr-s expression in adult mouse organs. The arrow corresponds to 2.2kb mr-s transcript. (I) RT-PCR analysis of total RNAs extracted from E13 whole embryo, P0 whole body (except for the eye), P7 retina, P7 pineal gland, P7 brain, P7 liver, adult retina, adult pineal gland, adult brain and adult liver, respectively. RPE, retinal pigment epithelium; NR, neural retina; NBL, neuroblastic layer; GCL, ganglion cell layer; ONL, outer nuclear layer; INL, inner nuclear layer. To determine the tissue specificity of mr-s expression, the expression of the mr-s gene in various adult tissues was examined by Northern hybridization (Fig. 2H). As a control, P7 retinal RNA was used. Four bands corresponding to 7.2-kb, 4.0-kb, 2.5-kb and 2.2-kb were detected in P7 retina. The 2.2-kb band corresponds to the cDNA characterized in this study. The larger bands, possibly alternative spliced transcripts, have not yet been characterized. The mr-s probe did not detect a band in the adult tissues examined, indicating that these tissues do not express mr-s at a level comparable to that of the developing retina.
Previous reports revealed that many photoreceptor-specific genes are also expressed in the pineal gland [38]. We examined the expression of mr-s transcripts in the whole embryo, whole body, retina, pineal gland, brain, liver and other organs at various stages by RT-PCR (Fig. 2I and data not shown). We amplified PCR fragments of 292 bp and 452 bp with primer pairs for genes encoding mouse mr-s and G3PDH, respectively. In E13 whole embryo and P0 whole body (except for the eye), no mr-s signal was detected. As expected, we observed that mr-s is expressed in the P7 and adult pineal gland. In the P7 and adult brain, liver and several other organs, the RT-PCR amplified band of mr-s was not detected (Fig. 2I and data not shown). Our data showed that mr-s is predominantly expressed in developing photoreceptors and the pineal gland.

Regulation of mr-s transcription by Crx homeodomain protein
Transcription factors Otx2 and Crx are known to be key regulators of retinal photoreceptor development [3,8,9]. Although both Otx2 and Crx are expressed in developing photoreceptor cells in the embryonic retina, only Crx is highly expressed in the postnatal photoreceptors, suggesting that mr-s may be regulated by Crx. To test whether Crx regulates mr-s transcription, we performed in situ hybridization of mr-s mRNA on wild-type and Crx KO P5 retinas (Fig. 3A, B). In the Crx KO retina, the mr-s transcript was dramatically reduced (Fig. 3B). This indicates that Crx is essential for transactivation of mr-s. Moreover, to test whether the transcription of mr-s is also regulated by Crx in the pineal gland, RT-PCR analysis was used to compare transcriptional levels of mr-s in P28 wild-type and Crx KO pineal gland (Fig. 3C). The results revealed that mr-s transcription was significantly downregulated in the Crx KO pineal gland. Taken together, these data indicate that mr-s transcription is regulated by Crx in the developing photoreceptors and pineal gland.
Figure 3  The transcription of mr-s is regulated by Crx. (A, B) In situ hybridization using a probe for mouse mr-s was performed on the wild-type (A) and Crx KO retinas (B) at P5. mr-s expression was drastically reduced in the Crx KO retina (B). Scale bar, 100 μm. (C) RT-PCR analysis of total RNAs extracted from the pineal glands of P5 wild-type and Crx KO mouse. The upper and lower lanes show PCR products amplified by the primer pairs specific for mr-s and G3PDH cDNAs, respectively. Water was used for control RT-PCR reaction. (D-F) Crx transactivates mr-s transcription. Reporter plasmids for the luciferase assay are shown. Blue boxes represent Crx binding sites (D). Relative activity of the luciferase is indicated when Pro1.2k was co-transfected with Crx, Otx2, Nrl, and Crx+Nrl, respectively (E). Fold activation is indicated when Pro1.2k, mut1259, mut198, mut72 and mut all were co-transfected with the Crx expression vector (Crx+) or the empty vector (Crx-) into HEK293T cells (F). Error bars represent standard error of mean. To further examine whether Crx regulates mr-s transcription directly or not, we next performed a luciferase assay using the 1.2-kb proximal promoter region of mr-s fused to a luciferase gene as a luciferase reporter (Fig. 3D, Pro1.2k) and the Crx, Otx2, Nrl expression vectors, respectively. This 1.2-kb region of the mr-s upstream sequence contains three Crx binding consensus sequences. As shown in Fig. 3E, the luciferase activity was significantly upregulated when the Crx or Otx2 expression vector was co-introduced with Pro1.2k into HEK293T cells, while the luciferase activity was not altered when the Nrl expression vector was co-introduced. A previous report suggested that the transcriptional activity of Crx is augmented with Nrl when the rhodopsin promoter was used as a reporter gene [6]. On the other hand, our present data showed that the luciferase gene expression was not upregulated when both Crx and Nrl expression vectors were co-introduced with Pro1.2k compared to the activity when the Crx only expression vector was introduced. This may be due to cell type differences because a cell type of retinal/pineal origin was not used in our luciferase assay. In addition, our present data showed that Otx2, which is reported to have the same binding consensus as Crx, also transactivated mr-s expression. As shown in Fig. 2A–F, the expression pattern of mr-s correlates with that of Crx. In contrast, the transcripts of Otx2 are mainly detected in the photoreceptor layer at embryonic stages. Therefore, we concluded that mr-s transcription is directly regulated mainly by Crx.
We also constructed reporter vectors in which mutations were introduced at the three Crx binding sites (Fig. 3D, mut1259, mut198, mut72, mut all). Then the Crx expression vector was co-introduced with mut1259, mut198, mut72 and mut all, respectively (Fig. 3F). The transactivaton activity by Crx was clearly reduced when mut198 or mut72 was co-introduced. On the other hand, when mut1259 was co-transfected, the transactivation activity by Crx was not altered. These results suggest that the Crx binding sites 72-bp and 198-bp upstream from the transcription initiation site are crucial for the direct regulation of mr-s transcription by Crx.

Self-association of mr-s protein
SAM domains are known to function as protein-protein interaction modules [15-17]. Although SAM domains can bind to various non-SAM domain-containing proteins, many homo-SAM and hetero-SAM domain interactions have been reported. To investigate whether the SAM domain of the mr-s protein can also function as a protein-protein interaction module, we performed yeast two-hybrid screening using full-length mr-s protein as the bait. Using this bait, we screened the transcriptional activator fusion protein library in which mouse P0-P3 retinal cDNAs were fused to the GAL4 activation domain. The most frequent positive clones (5 out of 28) were cDNA fragments containing the SAM domain of mr-s (Fig. 4A). This result strongly suggests that mr-s protein self-associates through SAM domain-containing regions. We then directly tested this self-association of mr-s protein in yeast. We fused full-length or truncated portions of the mr-s protein to the DNA-binding domain of the yeast transcription factor GAL4 to make bait constructs. We fused full-length or truncated portions of the mr-s protein to the GAL4 transcriptional activation domain to make prey constructs (Fig. 4B). These constructs were transformed into yeast that contain a transgene with GAL4 binding sites upstream of the lacZ gene. We found that the full-length mr-s bait construct induced lacZ expression with the full-length mr-s prey construct (Fig. 4B, full × full). The N-terminus 400 amino acid (aa) stretch of mr-s, which does not contain a SAM domain, does weakly activate transcription of lacZ (Fig. 4B, full × N). The N-terminus 400 aa stretch of mr-s was able to induce transcription of lacZ weakly with the same N-terminus 400 aa stretch of mr-s (Fig. 4B, N × N). Although the N-terminus 400 aa mr-s protein weakly activates lacZ transcription with the same N-terminus portion, a much stronger activation of lacZ expression was observed with a C-terminus portion encoding the 391–542 aa stretch of mr-s (Fig. 4B, full × C, C × C). Our GAL4 assay indicated that the signal when the full-length mr-s was present in both the bait and prey contexts was weaker than when isolated SAM domains were used. This may simply reflect the tendency for the small fusion proteins to enter the yeast nucleus and occupy GAL4 binding sites. Alternatively, the SAM domain may be less accessible for interaction in the full-length protein context as previously reported [39].
Figure 4  Summary of yeast two-hybrid screening and GAL4 assay. (A) Full-length mr-s as a bait used in the yeast two-hybrid screening and positive clones are indicated. Note that all of five mr-s clones identified in the screening contain SAM domains. (B) Schematic diagram of the mr-s fusion proteins used in the yeast GAL4 assay. Black boxes represent the position of SAM domains. Relative levels of LacZ expression are shown on the right, respectively. Note; ++ indicates an intense blue signal visible after 12hr of incubation at 37°C, + indicates a blue signal visible after 24hr of incubation. BD, binding domain; AD, activation domain; full, full-length mr-s; N, N-terminal portion of mr-s (amino acids 1 to 400); C, C-terminal portion of mr-s (amino acids 391 to 542). To confirm self-association of the mr-s protein in mammalian cells, we next performed co-immunoprecipitation studies in HEK293T cells by co-transfection of HA-tagged full-length/truncated mr-s and Flag-tagged full-length/truncated mr-s (Fig. 5A). As a negative control, we constructed Flag-tagged Sonic hedgehog (Shh) (lane 2 and 7). In accordance with the result of the yeast two-hybrid GAL4 assay, HA-tagged full-length mr-s (full-HA) was co-immunoprecipitated with Flag-tagged full-length mr-s (Flag-mrs) and the Flag-tagged C-terminus portion containing the SAM domain (Flag-SAM), respectively (Fig 5B, lane3 and lane5). We also found a weak co-immunoprecipitation band in co-transfection of full-HA and Flag-tagged N-terminus portion of mr-s (Flag-ΔSAM, Fig. 5B, lane 4). When ΔSAM-HA and Flag-tagged deletion mutants were co-transfected, ΔSAM-HA was co-immunoprecipitated with Flag-mrs and Flag-ΔSAM (Fig. 5B, lane 8 and lane 9), while ΔSAM-HA was not co-immunoprecipitated with Flag-SAM (Fig. 5B, lane 10).
Figure 5  The mr-s protein can self-associate in mammalian cells. (A) Schematic drawing of the constructs used for immunoprecipitation assay. HA-tagged or Flag-tagged full-length (amino acids 1 to 542), ΔSAM (amino acids 1 to 400) and SAM (amino acids 401 to 542) regions were inserted into pcDNA3 vector, respectively. (B) The constructs indicated above were transfected into HEK293T cells. Each lane was co-immunoprecipitated by anti-Flag antibody and detected by anti-HA antibody. Input protein lysates are shown in the lower panels. (C) Flag-tagged two site-directed mr-s mutants, Flag-W404A and Flag-G453A were generated and co-transfected with full-HA. Each lane was co-immunoprecipitated by anti-HA antibody and detected by anti-Flag antibody. To investigate whether the mr-s protein self-associates mainly through the SAM domain, two site-directed mutations were generated in the SAM domain of mr-s (Fig. 1B, arrows). These mutations alter residues that are conserved in the SAM domain of ph and previous report indicates that these mutations of ph-SAM cause significant reduction in binding activity to the other SAM domain-containing protein, Sex comb on midleg (Scm) (41). Based on this result, we introduced two types of site-directed mutations, which correspond to the mutations introduced in ph protein, into Flag-tagged full-length mr-s (Flag-W404A and Flag-G453A). We found that Flag-W404A binding activity was significantly reduced and Flag-G453A binding activity was also slightly reduced compared to Flag-mrs (Fig. 5C). These results, together with yeast two-hybrid GAL4 assay, indicate that the mr-s protein self-associates strongly through its SAM domain and weakly through the N-terminus portion lacking SAM domain.

The subcellular localization of mr-s protein in mammalian cells
The putative nuclear localization signal at the N-terminus of the mr-s protein (Fig. 1A, dashed box) suggests that the mr-s protein localizes in the nucleus. To determine the subcellular localization of mr-s in mammalian cells, we introduced an HA-tagged full-length mr-s plasmid into HEK293T cells. Confocal microscopy showed that the mr-s protein localized mainly in the nucleus of HEK293T cells (Fig. 6A, B). To confirm the precise localization of the mr-s protein, full-length mr-s was co-immunostained with DAPI (Fig. 6C–E). These data showed that the mr-s protein localizes preferentially to the nucleus in mammalian cells, supporting the idea that the mr-s protein is involved in transcriptional regulation as are ph and/or TEL.
Figure 6  Subcellular localization of mr-s in mammalian cells. (A, B) HA-tagged full-length mr-s was transfected into HEK293T (B) and detected by anti-HA antibody (A). Scale bar, 20 μm. (C-E) HEK293T cells immunostained with anti-HA antibody (C), DAPI (D), and merged image (E). Scale bar, 50 μm.

The GAL4-mr-s fusion protein functions as transcriptional repressor
A member of PcG proteins, ph, does not contain an obvious sequence-specific DNA binding motif (16). Ph functions as a transcriptional repressor through its polymerization and protein-protein interaction with other sequence-specific transcriptional repressors, which can form a higher order chromatin structure. The mr-s protein also does not have an obvious DNA-binding domain. To characterize the biochemical activity of mr-s, we next performed a luciferase assay. We generated effector plasmids, which express various deletion constructs of mr-s fused to the GAL4 DNA-binding domain (Fig. 7A). We first confirmed that full-length mr-s fused to GAL4 DNA binding domain (DBD-mrs) had no effect on the pGL3 promoter plasmid lacking GAL4 binding sites (data not shown). When the 5xGAL4-pGL3 reporter plasmid was co-transfected, DBD-mrs repressed luciferase activity by about 90% in a dose-dependent manner (Fig. 7B). As a control, GAL4 DNA-binding domain (DBD) had no significant effect on the 5xGAL4-pGL3 reporter plasmid. In addition, we confirmed that full-length mr-s without GAL4 DBD had no effect on the same reporter plasmid (data not shown). We next analyzed deletion constructs in which the N-terminus 400 aa stretch of mr-s (amino acids 1 to 400) or the C-terminus portion (amino acids 391 to 542) were fused to the GAL4 DBD (Fig. 7A, DBD-N, DBD-C). While DBD-N had no repressive effect on this reporter activity, DBD-C repressed luciferase gene expression by about 65%. This result suggested to us the possibility that DBD-mrs exerts transcriptional repressive activity via self-association through its SAM domain. To investigate whether the homophilic association of mr-s is required for transcriptional repression, two site-directed mutants, DBD-W404A and DBD-G453A, either of which may reduce self-binding ability, were analyzed (Fig. 7C). Compared to DBD-mrs, DBD-W404A and DBD-G453A had repression activity of 72% and 87%, respectively. While the ability of mr-s self-association partially correlates with transcriptional repressive activity, these mutants do not compromise the ability to repress transcription critically. To determine the regions of mr-s involved in transcriptional repression more precisely, the C-terminus portion was divided into two regions and each was fused to the GAL4 DBD (Fig. 7A, DBD-tail, DBD-SAM). As a consequence, DBD-SAM (amino acids 384 to 462) did not have a repressive effect on the 5xGAL4-pGL3 reporter plasmid. On the other hand, luciferase activity was repressed by 55% when DBD-tail (amino acids 459 to 542) was co-transfected with this reporter plasmid (Fig. 7D). To assess the transcriptional repressive activity of mr-s in cells of retinal origin, we performed similar experiments using human Y79 retinoblastoma cells. The results indicated that DBD-mrs also reduced luciferase activity significantly in Y79 retinoblastoma cells (Fig. 7E). However, luciferase activity was repressed by about 30% in Y79 cells, while it was repressed by 90% in HEK293T cells. This might be due to the difference in transfection efficiency between these cell lines. Another possibility is that intracellular environment in Y79 cells, a retinoblastoma cell line, is insufficient for recapitulating developing photoreceptors. In this study, we did not address the question whether or not our in vitro data reflect native mr-s transcriptional repression in vivo. However, these in vitro experiments using HEK293T and Y79 cells strongly support our hypothesis that mr-s functions as a transcriptional repressor in developing photoreceptors. Our results suggested that the C-terminal region of mr-s (amino acids 463 to 542) is required for transcriptional repression of mr-s and the SAM domain appears to be dispensable for this repressive activity. This C-terminal region of mouse mr-s is highly conserved among species (Fig. 7F). The sequence identity of the region was 93%, 73%, 41% and 40% for rat, human, chick and zebrafish, respectively (Fig. 1D). This strongly suggests that the C-terminal region of mouse mr-s functions as a transcriptional repressive domain in photoreceptor development. However, this region does not contain characteristic amino acid motifs and the mechanism through which the region achieves and/or maintains gene repression remains to be clarified in the future.
Figure 7  mr-s fused to GAL4 DNA binding domain functions as a transcriptional repressor in HEK293T cells. (A) Schematic drawing of the constructs used for the luciferase assay. 5xGAL4-pGL3 reporter plasmid was co-transfected into HEK293T cells with effector plasmids expressing various deletion mutants fused to GAL4-DBD. (B) Various amounts of DBD, DBD-mrs, DBD-N or DBD-C plasmids were transfected with 0.1 μg of 5xGAL4-pGL3 reporter plasmid. The reporter activity in the presence of the pcDNA3 vector (pcDNA3) was designated as 1. Error bars represent standard error of mean. (C) DBD-W404A and DBD-G453A were co-transfected into HEK293T cells with 5xGAL4-pGL3 reporter plasmid. Fold repression was calculated as the fold decrease in luciferase activity compared with DBD-mrs. Error bars represent standard deviation. (D) Various amounts of DBD-tail or DBD-SAM were transfected with 5xGAL4-pGL3 reporter plasmid. Error bars represent standard error of mean. (E) pcDNA3 or DBD-mrs (5 μg) was co-transfected into Y79 retinoblastoma cells with 0.5 μg of 5xGAL4-pGL3 reporter plasmid. The reporter activity in the presence of pcDNA3 was designated as 1. Error bars represent standard deviation. Asterisk marks statistically significant difference (Student's t test: p < 0.03). (F) Alignment of the C-terminal regions for mouse, rat, human, chick and zebrafish mr-s proteins. Conserved amino acid residues are shown with a dark shadow. Taken together, our findings suggest that DBD-mrs functions as a transcriptional repressor and that the repression activity of mr-s is not due to a homophilic interaction through its SAM domain but to the C-terminal region (amino acids 463 to 542).