Expression of Rod-Derived Cone Viability Factor: Dual Role of CRX in Regulating Promoter Activity and Cell-Type Specificity

Background RdCVF and RdCVF2, encoded by the nucleoredoxin-like genes NXNL1 and NXNL2, are trophic factors with therapeutic potential that are involved in cone photoreceptor survival. Studying how their expression is regulated in the retina has implications for understanding both their activity and the mechanisms determining cell-type specificity within the retina. Methodology/Principal Findings In order to define and characterize their promoters, a series of luciferase/GFP reporter constructs that contain various fragments of the 5′-upstream region of each gene, both murine and human, were tested in photoreceptor-like and non-photoreceptor cell lines and also in a biologically more relevant mouse retinal explant system. For NXNL1, 5′-deletion analysis identified the human −205/+57 bp and murine −351/+51 bp regions as having promoter activity. Moreover, in the retinal explants these constructs drove expression specifically to photoreceptor cells. For NXNL2, the human −393/+27 bp and murine −195/+70 bp regions were found to be sufficient for promoter activity. However, despite the fact that endogenous NXNL2 expression is photoreceptor-specific within the retina, neither of these DNA sequences nor larger upstream regions demonstrated photoreceptor-specific expression. Further analysis showed that a 79 bp NXNL2 positive regulatory sequence (−393 to 315 bp) combined with a 134 bp inactive minimal NXNL1 promoter fragment (−77 to +57 bp) was able to drive photoreceptor-specific expression, suggesting that the minimal NXNL1 fragment contains latent elements that encode cell-type specificity. Finally, based on bioinformatic analysis that suggested the importance of a CRX binding site within the minimal NXNL1 fragment, we found by mutation analysis that, depending on the context, the CRX site can play a dual role. Conclusions/Significance The regulation of the Nucleoredoxin-like genes involves a CRX responsive element that can act as both as a positive regulator of promoter activity and as a modulator of cell-type specificity.


Introduction
Retinitis pigmentosa (RP) is a genetically and clinically heterogeneous group of inherited retinal degenerative diseases that causes severe visual impairment in as many as 2 million patients worldwide [1,2]. Most identified forms of RP are caused by mutations in genes whose expression is enriched in or restricted to rod photoreceptor cells [3,4]. The loss of rods in RP is generally followed by a secondary degeneration of cone photoreceptors [5,6]. It is this secondary loss of cones that is most significant clinically because cones are required for high acuity, color, and most daylight vision. In the vast majority of cases, RP and related forms of retinal degeneration are untreatable, and all that can be offered to patients are supportive measures.
Given the clinical importance of cone loss in retinal degenerative disease, in our effort to develop broad-based therapies for RP that go beyond single gene replacement, we have been studying the mechanisms of cone cell loss. We and our colleagues have demonstrated that rods secrete factors that promote cone survival, reviewed in [7]. Furthermore, through expression and homology cloning approaches, we have identified a novel family of rodsecreted factors that are trophic for cones. These factors, Rod-Derived Cone Viability Factor (RdCVF) and RdCVF2, are encoded, respectively, by the nucleoredoxin-like genes NXNL1 and NXNL2 [8,9]. The cone-directed protective activity of retinal explant conditioned media is significantly reduced after immunodepletion of RdCVF, suggesting that a major part of retinal derived cone survival activity in vivo is due to RdCVF [8]. The degeneration of rods in the rd1 mouse model of RP results in the loss of expression of RdCVF by rods as well as by bipolar cells at the onset of cone degeneration [8,10]. The potential of RdCVF administration for promoting survival of the central vision attributable to cones in patients suffering from RP is well illustrated by the finding that injection of RdCVF protein in the P23H transgenic rat model of dominant RP both slows down secondary cone loss and, more importantly, preserves cone function as measured by electroretinography [11].
Both RdCVF and RdCVF2 are encoded by an unspliced transcript from their respective genes. NXNL1 and 2 also express a spliced mRNA that encodes the protein products RdCVFL and RdCVF2L, which unlike RdCVF and 2 contain an entire thioredioxin-fold [9]. This suggests that the mammalian nucleoredoxin-like genes are bi-functional, with one mRNA encoding a cone survival factor and a second encoding an active thioredoxin enzyme involved in the defense against oxidative stress, similar to the situation reported for the RdCVFL homolog in the arthropod Carcinoscorpius [12,13]. The search for proteins targeted by the putative redox-controlling activity of RdCVFL has led to the identification of a functional interaction between RdCVFL and the microtubule association protein TAU [14]. The hyperphosphorylation of TAU in the retina of the mouse carrying a deletion of the Nxnl1 gene likely plays a role in the degeneration of rods observed in that model and in also in their sensitivity to oxidative stress [15].
A key element in understanding nucleoredoxin-like signaling in the eye is the restricted expression of the Nxnl1 and 2 genes in the retina. Nxnl1 expression is fully restricted to the eye, where it is expressed prominently by photoreceptors but also by bipolar cells [8,10]. Although Nxnl2 is expressed in selected regions of the brain as well as in the eye [9], within the retina RdCVF2 and RdCVF2L are expressed exclusively by photoreceptor cells. In the rd1 mouse model of RP, the death of rods (which account for 97% of all photoreceptors) results in the loss of retinal expression of all the products of the nucleoredoxin-like genes, including expression of Nxnl1 from bipolar cells [10]. This highly regulated expression pattern of the Nxnl1 and 2 genes raised the interesting question as to what are the mechanisms that control its cellular specificity within the retina. Understanding the regulation of the RdCVFs is not only important for what it can teach us about fundamental questions of cell type-specific gene transcription and processing, but also because manipulating expression of members of the RdCVF family could have therapeutic potential. Work from a number of groups, using a variety of experimental approaches, has identified some of the transcription factors that mediate photoreceptor gene expression. Among the factors identified are CRX, NRL, NR2E3, RX/RAX, PAX6, and QRX [16][17][18][19][20][21][22][23][24]. Interestingly, many of these factors are homeodomain proteins, and it seems that combinatorial use of a relatively small number of proteins can account for a significant degree of photoreceptor specificity. Also of interest, mutation of or modulation of the expression of many of these proteins can cause abnormalities of photoreceptor development and/or retinal degeneration. For example, the homeoprotein CRX (cone rod homeobox), which itself is expressed predominantly in photoreceptors, binds to the consensus sequence C/TTAATCC/A and has functionally important binding sites present in the upstream region of many photoreceptor-specific genes, including rhodopsin, arrestin, IRBP, and b-phosphodiesterase-6 [18,19,24]. Knock-out of CRX in mice does not significantly alter photoreceptor lineage decisions, but it does markedly affect photoreceptor differentiation both morphologically and at the gene expression level [25,26]. In humans, CRX mutations induce a variety of phenotypes of retinal degeneration [27].
In a first step towards understanding the regulation of RdCVF expression, we recently used a combined promoter and transcriptomic analysis to identify some of the transcription factors that modulate Nxnl1 expression, and among the factors identified were the homeoproteins CHX10/VSX2, VSX1, and PAX4, and the zinc finger protein SP3 [10]. Here, we have extended that analysis and focused on the cis-acting 59-upstream regulatory sequences in both the murine and human Nxnl1 and 2 genes. Since a bioinformatic evolutionary comparison analysis (phylogenic footprinting) did not reveal significant sequence conservation between the 59-regions, we utilized an experimental approach consisting of transient transfection of luciferase reporter constructs in four distinct cell lines and ex vivo electroporation of mouse retinal explants with GFP reporter plasmids. The use of the retinal explants allowed analysis of cell type-specific elements. We mapped active proximal promoter regions in both species to small regions of the Nxnl1 and Nxnl2 genes. The expression pattern of the electroporated GFP-reporters showed that while the identified Nxnl1 promoter sequence specifies expression to photoreceptor cells, the identified Nxnl2 active promoter fragments are not sufficient to restrict reporter expression to photoreceptors. However, we did find that combining a minimal and inactive 134 base pair (bp) piece of the Nxnl1 promoter with an active 79 bp piece of the Nxnl2 upstream region could restore photoreceptor specificity. This suggested that the 134 bp Nxnl1 sequence contains sufficient information for photoreceptor specificity. Furthermore, by mutation analysis we showed that a CRX binding element located within the 134 bp Nxnl1 fragment is necessary for promoter activity, thereby implicating CRX in the cell type-specific regulation of RdCVF expression.

Results
Mapping the transcriptional start sites of the murine and human NXNL1 and NXNL2 genes In order to delineate the 59 upstream regions of the mouse and human NXNL1 and NXNL2 genes, we determined the position of their putative transcriptional start sites (TSSs) by rapid amplification of cDNA ends (RACE), using RNA purified from mouse and human retinas. As negative controls, we performed the reactions using RNA isolated from the retina of the Nxnl1 and Nxnl2 knockout mice since the strategy used to generate these animal models results in the elimination of their respective mRNAs [15] and Jaillard et al., (submitted). Additionally, as control for human retinal tissue, we used HeLa cells (ATCC CCL-2) since northern blot analysis showed that NXNL1 transcripts are not expressed by these cells in Audo et al., (manuscript in preparation). As expected, RACE-PCR products for both Nxnl1 and 2 were present with wild-type (wt) retina, while with Nxnl1-/-RNA only a product for Nxnl2 was generated, and with Nxnl2-/-RNA only a product for Nxnl1 was generated ( Figure 1A). In addition, we confirmed that both NXNL1 and NXNL2 are expressed in the human retina and that HeLa cells express NXNL2 but not NXNL1. The products of the RACE reactions were cloned, analyzed by gel electrophoresis and representative clones were sequenced ( Figure 1B). Three and four retinal putative TSSs were identified for the mouse Nxnl1 and Nxnl2 genes, respectively, three for human retinal NXNL1, and only one for human retinal NXNL2 ( Figure 1C). For nomenclature purposes, we have chosen to label the nucleotides in each gene based on the position of its major putative TSS (+1) (since Nxnl1 had two major TSSs of similar strength, one at 268 bp and one 252 bp, we arbitrarily chose to use the one at 252 bp for labeling). Interestingly, while as noted above NXNL2 is expressed by HeLa cells, its major putative TSS in HeLa cells is at 272 bp, as compared to 228 bp in the retina, suggesting a difference in the mechanism controlling its expression in photoreceptor versus HeLa cells.

Murine and human NXNL1 and NXNL2 promoter activity in retinal and non-retinal cell lines
In most instances, the major DNA elements regulating promoter activity are located in the 59-upstream region of a given gene [28][29][30]. In order to map the promoter regions of the murine and human NXNL1 and NXNL2 genes, we used PCR to construct a collection of reporter plasmids containing various sequences 59 to the respective gene's ATG site into the promoterless luciferase reporter vector pGL4.17. The human samples were generated from DNA that was derived from a single donor's white blood cells. DNA sequencing of the resultant constructs, as well as sequencing of additional independent PCR products, and comparison with the UCSC database revealed both previously identified and novel sequence variants within the upstream regions of both the NXNL1 and NXNL2 genes (since the term ''polymorphism'' is generally used to refer to sequence variants that occur at a frequency of at least 1% in a population, and we do not have any frequency data on the variants we identified, in the text below we will refer to these as ''sequence variants''). The identity and location of the detected sequence variants and their occurrence within the various reporter constructs studied are noted in the appropriate figure legends. For initial analysis, reporter plasmids were transfected into the Y79 retinoblastoma cell line, which based on the NCBI EST database expresses RdCVF, and into the non-retinal cell line HEK 293, which was derived from adenoviral DNA transformed human embryonic kidney cells [31]. The activity of the various promoter constructs was measured using a dual luciferase assay to normalize expression relative to that of a HSV-thymidine kinase promoter (pRL-TKrenilla), and the data are presented as fold induction compared to the empty pGL4.17 plasmid.
With the NXNL1 transfection series and Y79 retinoblastoma cells, we found similar levels of reporter activity with the 22072 to +57 bp, 21034 to +57 bp, and 2501 to +57 bp constructs ( Figure 2A). In HEK 293 cells we found similar levels of reporter activity with the 21034 to +57 bp and 2501 to +57 bp constructs, but observed significantly less activity with the 22072 to +57 bp construct, suggesting that a negative DNA element(s) active in 293 cells, but not in the retina, might be present in the 22072 to 21034 bp region. Using an analogous approach with a set of shorter deletion constructs, we localized much of the promoter activity to the immediate upstream sequence in Y79 cells, and probably also HEK293 cells, to the region between 2205 and 277 bp (Figure 2A). Here again, there is a suggestion of a negative regulatory element specific to HEK 293 cells, this time localized to the 2400 to 2205 bp region.
In contrast to these findings of significant promoter activity with the human NXNL1 upstream region, with murine Nxnl1 we identified only minimal promoter activity with the constructs tested ( Figure 2B). Similar results were obtained with two other cell lines, the mouse cone-like line 661W [32] cells and COS-1 cells, derived from African green monkey kidney cells ( Figure S1). As noted above, there is surprisingly little sequence conservation between the murine and human NXNL1 upstream regions. Thus one possible explanation for the mouse constructs lack of significant promoter activity with the monkey and human cells is that they may lack the transcription factors necessary to recognize the murine regulatory elements, and based on the results with mouse retinal explants (see below), this does seem likely to be at least part of the explanation ( Figure 2B). However, species difference does not explain the difference observed with the murine 661W cells.
In the case of the NXNL2 gene, contrary to that observed with NXNL1, relative promoter activity was higher in the nonphotoreceptor cell line HEK 293 than in Y79 cells ( Figure 2C). The major component of the observed positive regulatory activity, in both HEK 293 and Y79, cells was localized to the region between 2393 and 2315 bp ( Figure 2C). As with the human gene, the murine Nxnl2 promoter constructs were also considerably more active in HEK 293 than in Y79 cells ( Figure 2D). The step-wise reduction in activity observed with the shorter deletion constructs in the HEK 293 cells suggests the presence of multiple regulatory elements, each of moderate activity ( Figure 2D). Comparable results were obtained with 661W and COS-1 cells ( Figure S1).

Murine and human NXNL1 and NXNL2 promoter activity in electroporated mouse retinal explants
Based on the lack of clear retinal specificity in the above described cell line studies, which in fact was not surprising given the clear importance of cellular environment on transcriptional activity, we switched to electroporation of perinatal mouse retinal explants, an ex vivo approach that we thought would be more amenable to analysis of retinal cell-specific promoter elements. Prior studies by Matsuda and Cepko had shown that mouse retina can be transfected reasonably efficiently by electroporation and that in these experiments retinal promoters tend to retain their cell type specificity [33,34]. Using an adaptation of their methods, we performed electroporation with various green fluorescent protein (GFP) reporter constructs analogous to the previously described Nxnl1 and Nxnl2 luciferase constructs, and then quantified the activity of the promoter fragments by measuring the cellular fluorescence of the electroporated retinas. Mouse retinal explants were electroporated at post-natal day 0 (P0) and maintained in culture 5 days. Fluorescent signals were captured with a Charge Coupled Device (CCD) camera. Using this approach, although there was certainly significant variation from electroporation to electroporation, even with the same construct, we were able to observe fluorescent signals whose intensity was highly dependent upon the upstream sequence contained in the reporter construct. For example, with human NXNL1, we observed fairly similar expression levels with the 22072 to +57 bp, 2501 to +57 bp, and 2205 to +57 bp constructs, but when the upstream region was deleted to 277 to +57 bp there was essentially complete loss of activity ( Figure 3A). Although consistent in a general sense with the pattern of the Y79 and HEK 293 results, quantitatively the mouse retina result was significantly different in that with the cell lines the decrease in activity with deletion of the 2205 to 277 bp region resulted in only a 40% decrease in activity ( Figure 2A). Other differences between the cell line and retinal explant results were even more striking. Unlike with the cell lines, in the mouse explants the murine Nxnl1 promoter constructs are more active than the human NXNL1 (compare Figure 2A-B with Figure 3A-B). With the cell lines, the Nxnl1 construct with the maximal activity was the 277 to +51 bp fragment ( Figure 2B), whereas this construct had no detectable activity in the explant system ( Figure 3B). The consistently low activity of the 21749/+51 bp construct in the explant system compared to the 2924 to +51 bp construct most probably indicates the presence of a negative regulatory element located between 21749 and 2924 bp.
With the NXNL2 promoter constructs, the retinal explant results were fairly consistent with those obtained with the cell lines (compare Figure 2C and Figure 3C). As with the cell lines, the sequence from 2393 to 2315 bp was required for significant promoter activity. This finding delineates a region of 79 bp (2393 to 2315) that presumably contains a critical DNA element(s). The intensity of GFP fluorescence was lower with the longer NXNL2 constructs (22033/+27 and 2519/+27), suggesting the presence of a negative element in the more far upstream region. Surprisingly, the murine Nxnl2 constructs tested with the mouse explants showed minimal and inconsistent activity, although it was clear that the 2195 to +70 region did have some activity ( Figure 3D).
We have previously reported that the temporal pattern of expression of RdCVF and RdCVF2 mRNA in the retina of the mouse shows increasing expression during development that then plateaus when the retina reaches its mature state, a pattern that is fairly similar to that of rhodopsin except that the actual level of expression is considerably lower [8,9]. As part of our characterization of the various NXNL1 and NXNL2 promoter fragments, we performed a time course study with the retina explants, and also tested the bovine rhodopsin 2.2 kb promoter (22174/+70) as a comparative control [35]. A plasmid containing the 22174/+70 rhodopsin sequence upstream of red fluorescent protein (RFP) was co-electroporated with part of the deletion series of NXNL1 and NXNL2 GFP reporter constructs. Fluorescence was monitored daily over a period of 8 days and the intensity of the signal was quantified ( Figure 4). As expected, the intensity of RFP fluorescence increased over this time period, corresponding to the postnatal differentiation of the retina. The pattern of expression of the NXNL1 and NXNL2 promoters differed temporally from each other in that the NXNL1 constructs showed a continuous increase in activity with time whereas the NXNL2 constructs tended to show an early peak, usually at 1 day, and then show a decrease with time. The same regions that showed positive promoter activity in the single time-point studies, 2205 to 278 bp and 2351 to 278 bp for NXNL1 and Nxnl1, respectively, were found to be important for the observed temporal increase in promoter activity.

Murine and human NXNL1 promoter constructs drive expression of GFP specifically in photoreceptors
We next investigated the cell type specificity of the various promoter regions by determining the spatial distribution of GFP and RFP positive cells in slices of retinal explants 5 days after coelectroporation with the rhodopsin and NXNL1 and NXNL2 promoter constructs. With the human NXNL1 construct, 2205/ +57 GFP positive cells were observed exclusively in the outer nuclear layer, the part of the retina that corresponds to photoreceptor cell bodies ( Figure 5A). We also observed the colocalization of GFP and RFP in a substantial number of cells, indicating that the NXNL1 promoter construct is active in rods. The construct might also be active in cones as seen by the two GFP-positive/RFP-negative cells at the upper right corner of the panel, but this issue remains to be more definitively resolved. The results were quite similar with the mouse Nxnl1 construct 2351/ +51 ( Figure 5B), but by examining more extensively the labeled slices, we could identify some GFP positive cells in the inner part of the retina as represented in the merged portion of the panel. These non-photoreceptors cells where the Nxnl1 promoter is active are likely to be bipolar cells, because endogenous RdCVF mRNA is also expressed by bipolar cells [10]. These observations demonstrate that the 2205 to +57 bp human NXNL1 sequence and the 2351 to +51 bp murine Nxnl1 sequence are each sufficient, at least in retinal explants, to drive the expression of RdCVF specifically in photoreceptors.
The results were strikingly different when retinal explants were electroporated with GFP driven by the mouse and human NXNL2 promoter 59-upstream regions. With the human -393/+27 NXNL2 promoter construct, GFP-positive cells were observed mostly in the inner part of the retina ( Figure 5C) -the difference from the rhodopsin promoter can be easily appreciated by the absence of co-localization in the merged image. A similar situation was observed with the mouse Nxnl2 2382/+70 construct ( Figure 5D). The GFP pattern of expression resembles that of Muller glial cells. The absence of expression of these Nxnl2 reporters in photoreceptor cells cannot be attributed to the lack of transfection of the plasmids into photoreceptors since we could visualize RFP-positive rods in the same retinal explants. Similar results were obtained whatever the length of the upstream regions for both the human and mouse NXNL2 genes (data not shown), demonstrating that important photoreceptor-specific elements regulating the expression of Nnxl2 in photoreceptor cells are likely localized outside the 4.5 kb 59-upstream region that we tested.
The NXNL2 -393 to -315 bp region contains a positive transcriptional regulator and the NXNL1 277 to +57 bp region contains a photoreceptor-restricting element Our mapping analysis of the human NXNL2 promoter indicated the presence of a positive regulatory element located between 2393 and 2315 bp that is active in the retina but that is not sufficient to provide photoreceptor specificity ( Figure 3C and data not shown). To determine whether 79 bp sequence requires other elements of the NXNL2 promoter not included in the constructs for its activity, we separated it from the rest of the RdCVF2 promoter, by cloning it into pGL4.17 immediately upstream of the NXNL1 promoter sequence from 277 to +57 bp (79+134, Figure 6A). We chose the 277/+57 NXNL1 fragment as the scaffold for these experiments because this minimal promoter sequence contains needed core RNA polymerase II binding elements, since it is active in Y79 and HEK 293 cells (Figure 2A), but it is by itself insufficient for transcriptional activity in the more stringent and biologically relevant environment of retinal explants (Figures 3A and 4). When transfected into HEK 293 cells, the 79+134 fusion construct stimulated luciferase activity approximately 5-fold over the basic 277/+57 construct ( Figure 6B). When both constructs were tested using the retinal explant assay, the 277/+57 human NXNL1 construct was found to be inactive as observed earlier ( Figures. 3A  and 4). However, the 79+134 fusion construct was active, and perhaps surprisingly, its activity was found to be photoreceptorspecific ( Figure 6C-D). The merged image confirmed that the GFP-positive cells, which were only observed within the ONL, were also rhodopsin (RFP) expressing cells.
A CRX element controls the expression of RdCVF and is responsible for its photoreceptor-specific expression We took a bioinformatics approach, using Genomatix promoter analysis software, to try to identify elements within the NXNL1 277/+57 region that could account for its biological activity. The analysis identified four binding elements corresponding to transcription factors reported to be expressed specifically or preferentially in the eye (CRX, MAF, POU1F1, and POU3F4). A similar analysis was performed on the mouse Nxnl1 2352/+51 promoter sequence and two transcription factors specific to the eye were found to be common between the two analyses: a MAFresponse element half site and a putative CRX binding element. Although MAF-response element half sites can show activity, usually less than that observed with full sites [36], we chose to concentrate our efforts on the putative CRX sites, located at position 256 to 250 bp in the human gene and at 254 to 248 bp in the murine gene. The contribution of the identified CRX binding elements to NXNL1 promoter activity was assessed by mutating the TAA in the TAAT core recognition site to CCG in the human and mouse 2205/+57 and 2353/+51 promoter constructs, respectively. We choose the CCG mutation because this change has been previously reported to abolish CRX binding to DNA [37]. The wild-type and mutated human constructs were tested in Y79 cells, which express CRX as indicated by RT-PCR analysis (result not shown). As expected, the wild-type -205/+57 NXNL1 construct was active, but when the CRX site was mutated promoter activity dropped to that of the empty luciferase control vector ( Figure 7A). We then tested the effect of CRX mutation of the human and murine constructs in a time-course analysis in the retinal explant system ( Figure 7B). Mutation of the CRX site led to complete elimination of promoter activity for both the human and murine constructs. These results demonstrate that the proximal CRX binding element present in the human and mouse Nxnl1 upstream regions is necessary for their promoter activity.
We also tested the effect of the mutation of the proximal CRX binding element in the 79+134 NXNL2/NXNL1 fusion construct. In the Y79 assay, the NXNL1 CRX mutation within the fusion construct led to loss of promoter activity, just as observed with the NXNL1 -205/+57 construct ( Figure 7A). When tested in the ex vivo retinal explant assay, the CRX mutation in the context of the fusion construct did not lead to loss of promoter activity ( Figure 8A-B). However, the CRX mutation did lead to a major change, a loss of photoreceptor specificity, as shown by the presence of multiple GFP positive cells within the inner part of the electroporated retina ( Figure 8C).

Discussion
Our objective in this study was to begin to elucidate the ciselements and trans-factors that control the retinal expression of the trophic factors RdCVF and RdCVF2, which are encoded by NXNL1 and NXNL2, respectively. We initially performed transient transfection analysis of a deletion series of murine and human Nxnl1 and Nxnl2 59-uptream region reporter constructs using both retinal (Y79 and 661W) cell and non-retinal (HEK 293 and COS-1) cell lines. While promoter activity could be detected for a number of the constructs, the absence of evidence of cell-type specificity, i.e. absence of clear and consistent differences in promoter activity for the various constructs between the retinal and non-retinal cell lines, led us to utilize a more physiological assay for assessing retinal promoter activity. Although mouse transgenic studies are considered by many to be the gold standard for examining the cellular and temporal specificity of promoters, such studies are limited by position effects and the by the costs and time involved. As an intermediate approach that combines the practical advantages of cell line based approaches with the cellular specificity of transgenic studies, we adapted the electroporation of mouse retinal explants to provide rapid and reproducible quantitation of cell-specific promoter activity. Significantly, we found that regulatory regions identified, using the same promoter constructs, differed depending upon whether we performed the assays with the cell lines versus the retinal explants, with the explants giving stronger evidence of cell-type specificity. For example, with the Y79 cell assay the NXNL1 -77 to +57 bp construct retained considerable activity compared to the 2205 to +57 bp construct, while in the more stringent explant system the 59 deletion from 2205 to 277 bp led to loss of all detectable promoter activity, indicating that the deleted region contained a sequence with regulatory function that was essential in retinal cells but not in the cell lines tested. This suggests that the activity of the Nxnl1 promoter, like that of other genes that show highly cell-type restricted expression patterns, is regulated by transcription factors that are sensitive to the cellular context of the cell. Of even greater significance, as demonstrated with the NXNL1 and NXNL1/2 fusion gene studies, the explant system allows analysis of cell type  specificity within the retina. Immortalized cell lines are appropriate and provide a powerful system to reveal potent cis-acting elements with broad cellular spectrum, and also can be useful for gain of function studies in which non-expressed cell type-specific trans-acting factors are added to the cells, but are of limited use for revealing the promoter elements responsible for conferring highly cell-restricted patterns of expression.
The deletion series of the NXNL1 promoter constructs revealed a critical region proximal to the transcriptional start site of both the human (2205 to 277 bp) and mouse (2351 to 277 bp) genes ( Figure 3A and B). Within these regions, cis elements are able to confer promoter activity to the reporter constructs, and within the context of more downstream sequences to restrict expression to the naturally RdCVF expressing cells, photoreceptor cells and, to a lesser extent, bipolar cells ( Figure 5A and B). The promoter activity scored for these constructs (NXNL1 2205/+87 and Nxnl1 2351/+51) seems to depend on the state of differentiation of the retinal cells, and mostly of the differentiation state of photoreceptors since the activity gradually increases in concert with the activity of the rhodopsin promoter construct with time ( Figure 4). The bioinformatic analysis of the shortest active NXNL1 promoter constructs revealed the existence of a putative CRX binding element (256 to 250 bp in NXNL1 and 254 to 248 bp in Nxnl1). Because of the essential role of the homeodomain protein CRX in the maturation of photoreceptors and its implication in inherited photoreceptor degeneration [18,38,39,25], the role of this binding element was further analyzed by mutagenesis. The integrity of the CRX binding element was found to be essential for the activity of the Nxnl1 promoter in vitro and ex vivo (Figures 7A and B). While we cannot rule out the possibility that another promiscuous homeodomain protein regulates the expression of RdCVF through this ''CRX'' element, CRX it is the most likely candidate to be mediating the observed biological activity.
The deletion analysis of the human NXNL2 gene identified a 79 bp promoter element that showed promoter activity in both the cell line and retinal explant assay systems (Figures 2C and 3C). The activity of this element is not restricted to photoreceptor cells ( Figure 5C). Interestingly, this 79 bp sequence is active when fused to an inactive 134 bp NXNL1 minimal proximal promoter sequence ( Figure 6B and C), resulting in a reporter expression pattern that shows restriction to photoreceptors ( Figure 6D). These findings demonstrate that the NXNL2 79 bp sequence from 2393 to 2315 bp contains positive regulatory activity that can act independently of its own basal promoter and also strongly suggest that the NXNL1 277 to +57 bp proximal promoter region contains sufficient cell type-specific information, when supplemented with a less specific transcriptional activator, to restrict expression to photoreceptor cells. The identity of the biologically active regulatory elements within the 79 bp NXNL2 sequence, and the transcription factor(s) that bind to them, are presently unknown. For the cell type restricting activity of the 134 bp NXNL1 region, the putative CRX binding site discussed above appears to be important. Mutation of the CRX element abolished promoter activity of the NXNL2/NXNL1 fusion construct in Y79 cells ( Figure 7A). Although mutation of the CRX site did not significantly reduce promoter activity in the explant system, it did lead to loss of photoreceptor specificity (Figure 8). These results, taken together, suggest that the CRX site within the NXNL1 proximal promoter region plays a dual function depending on its context. Within the native NXNL1 promoter it appears to act as a positive transcriptional regulator whose activity is essential for expression. Without its activity there is no detectable expression so in this context its possible role in influencing cell type specificity cannot be assessed. In the context of the NXNL2/NXNL1 fusion construct, presumably because an element, or elements, within the 79 bp NXNL2 sequence is able to replace its transcriptional activating role, its function in restricting expression to photoreceptors is unmasked. Future studies will hopefully provide mechanistic insight into how CRX carries out these dual roles, and also provide information on the other regulatory sites and factors responsible for the photoreceptor expression of RdCVF and RdCVF2.

Animals
No animal experimentation was performed. The animals used in this study were obtained from Charles River France (L'Arbresle). Retinal tissues were obtained from newborn C57BL/6 mice. Animals were sacrificed the day of birth and eyes were dissected to recover neural retinal tissues (Authorization 75-1208 delivered on July 21 2008 by the Minister of Agriculture). The methods used to secure animals and tissues complied with the ARVO Statement fro the Use of Animals in Ophthalmic and Vision Research. No animal experimentation was performed. The animals used in this study were obtained from Charles River France (L'Arbresle).

Rapid Amplification of cDNA Ends (59 RACE)
The rapid amplification of cDNA ends (59 RACE) was carried out using total RNA purified from mouse retina, human retina and cultured HeLa cells by CsCl 2 ultracentrifugation. A First-Choice TM RLM-RACE Kit (Ambion) was used to identify transcription start sites according to manufacturer's instructions. Briefly, total RNA was treated with calf intestinal phosphatase to remove the 59-phosphate. The RNA was then treated with tobacco acid pyrophosphatase to remove the cap structure of the mRNA. A synthetic RNA adapter was then ligated to the RNA population to allow ligation to only RNA having been capped. After reverse transcription reaction, cDNA was obtained by nested-PCR using two reverse primers for each gene (human NXNL1: 59 CCGCCCGCAGTACATAGAACTCA 39 and 59 TCTGTGAGCCGCACGAAGAAGTC 39; mouse Nxnl1: 59 TC-AGCACGTCCCCACCAG GCTTAA 39 and 59 TCAGCACG-TCCCCACCAGGCTTAA 39; human NXNL2: 59 GCGCAT-GAAGTCCAGCATCTCC 39 and 59 TCGGCTGACACGAA-GACCACTTC 39; mouse Nxnl2: 59 CGTTCTGCTTGATGAC-CACCAGC 39 and 59 GGGGATGGCGGTGAT TTCGTA 39) and the two forward primers provided with the kit. PCR products were analyzed on 2% agarose gels and the reaction products were cloned into pGEM T Easy (Promega). A subset of recombinant clones were analyzed on agarose gel and the plasmid displaying different insert sizes were sequenced from both directions. Theses sequences were aligned to the genomic sequences.

Promoter reporter plasmids construction
The sequence upstream ATG of both murine and human Nxnl1 and Nxnl2 were cloned in the promotorless pGL4.17 (expressing a firefly luciferase, Promega) and peGFP (Clontech) plasmid using the gateway technology (Invitrogen). The vectors were converted into a gateway destination vector by introducing a cassette containing the attR sites flanking the ccdB gene into the multiple cloning site. To identify DNA sequences that regulate murine and human Nxnl1 and Nxnl2 transcription, a series of deletions from the 59 end of each gene was created. Primers were designed to amplify different lengths of the 59end regions of each gene, from approximately 100 base pairs (bp) to 4.5 kb and the attB sites were incorporated into the forward (attB site 59 GGG GAC AAG TTT GTA CAA AAA AGC AGG CTT A 39) and the reverse (attB site 59 GGG GAC CAC TTT GTA CAA GAA AGC TGG GT 39) primers. These primers introduced a TA at the 59-end of each construct except the CRX mutant and wild-type constructs (see below). The PCR reactions were performed using murine and human genomic DNA as the template and were subjected to a hot start of 5 min at 94uC prior to the addition of the Platinium Taq DNA Polymerase High Fidelity (Invitrogen) followed by 7 cycles amplification: denaturing for 25 sec at 94uC, annealing for 30 sec to 7 min at 66uC (depending on the size of the PCR product), 35 cycles amplification: denaturing for 25 sec at 94uC, annealing for 30 sec to 7 min at 60uC (depending on the size of the PCR product) and a final extension at 66uC for 10 min. The primers used for each construct were listed in Table S1. The PCR products were purified after gel agarose electrophoresis using the gel extraction kit (Qiagen). The BP reaction was performed using 300 ng of the PCR product, 300 ng of the pDONR vector (Invitrogen), 4 ml of the BP Clonase Reaction Buffer (Invitrogen) and 4 ml BP Clonase enzyme mix (Invitrogen) to obtain entry clones. The reaction was incubated one hour at room temperature. OneShot TOP10 chemically competent E.coli cells (Invitrogen) were transformed by using 5 ml of the BP reaction. Positive clones were identified by restriction enzyme analysis of purified plasmid and verified by DNA sequencing. The LR reaction was performed by using 300 ng of the entry clone, 300 ng of the destination vector (pGL4.17 or peGFP), 4 ml of LR Clonase Reaction Buffer (Invitrogen) and 4 ml of the LR Clonase enzyme (Invitrogen) to obtain the expression vectors. The reaction was incubated one hour at room temperature. OneShot TOP10 chemically competent E.coli cells (Invitrogen) were transformed by using 5 ml of the LR reaction. Positive clones were identified by restriction enzyme analysis.

Construction of mutated CRX binding site
Plasmids containing a mutated CRX binding site (the TAA in the TAAT core recognition site was modified to CCG) were generated by DNA synthesis (DNA2.0) and cloned in pDONR. The LR reaction was performed as previously described to subclone the sequences into pGL4.17 and peGFP. For the synthesis, the consensus sequence from GenBank (NCBI) was used. Because this sequence contained some sequence variants from the above described PCR generated constructs (due to both sequence variations in the donor DNA sample and to nucleotide differences in the PCR primer sequences) and because for the CRX mutation analysis we wanted the mutant constructs to differ from the wild-type constructs only in at the CRX binding site, we used the synthesized constructs as templates to generate wild-type constructs that were indeed identical except at the CRX site. A 'mutagenic' primer was used to convert the CCG sequence back to the wild type CRX binding site (TAA) by PCR-based mutagenesis. The resulting fragment was then used as a mega-primer to amplify the full-length construct in a second round of PCR amplification. The full-length wild-type constructs were then cloned back into pDONR on an AvaI -EcoRV fragment, and the resulting plasmids were sequenced. Reporter constructs were generated by Gateway recombination as described above.

RNA extraction and RT-PCR
Total RNA was extracted from Y79 cells using the RNeasy extraction kit (Qiagen). One microgram of total RNA was reverse transcribed into cDNA by incubation with 1 ml (500 mM final) of deoxyribonucleoside triphosphate solution (dNTP), 1 ml of oligodT (10 mM), 4 ml 56 reaction buffer, 2 ml of 0.1 M dithiotreitol (DTT), 1 ml of RNasin an 200 U of superscript II (Invitrogen) to a final volume of 20 ml. Samples were incubated 1 hour at 37uC and the enzyme was inactivated at 72uC during 10 min. The resulting cDNA were amplified in PCRs using Platinium Taq High Fidelity (Invitrogen) in 2 mM MgSO 4, 200 mM dNTP and 50 pmol of primers. Oligonucleotides used were: human CRX forward: 59  GGC TCT GAA GAT CAA TCT GCC TGA 39, reverse: 59  CCA TAG CTC TGG CCT GAT AGG GAG 39: human  G6PDH forward: 59 ATG TTCGTC ATG GGT GTG AAC C 39,  reverse: 59 AGG GAT GAT GTT CTT CTG GAG AGC C 39. Cycling protocol was: 94uC, 2 min; 35 cycles of 94uC, 30 sec; 60uC, 30 sec; 72uC, 2 min, followed by a final extension step at 72uC for 5 min. PCR products were run on a 1.5% agarose gel stained with ethidium bromide and visualized under UV illumination.

Transient transfection and reporter gene assay
Luciferase activity was determined using a luciferase assay system (Promega) according to the manufacturer's protocol. Briefly, Y79 and HEK 293 cells were plated for 24 h at respectively 5610 5 and 2610 5 cells/well in 24-well plates. Five hundred (for Y79) or 100 ng (for HEK 293) of promoter constructs or the empty vector (pGL4.17) was transiently transfected into the cells with Lipofectamine 2000 (Invitrogen). Cells were cotransfected with pRL-TK containing Renilla luciferase gene driven by the thymidine kinase promoter (Promega). One day after transfection, cell extract was lysed, and the luciferase activity of the extract was measured with the luminometer LB-96V (Berthold). Firefly luciferase activities were normalized by Renilla luciferase activities and relative luciferase activities were expressed as fold induction as compared to the empty vector (pGL4.17).

Electroporation of retinal explants
Mouse retinas from newborns were electroporated the day of birth as described in Matsuda & Cepko [33,40]. Briefly, newborn retinas were dissected in Hank's Balance Sodium Salts (HBSS) without calcium and magnesium (Invitrogen) and were cut into 3 pieces and were submerged in a 1 mg/ml DNA solution. They were subjected to 5 electric pulses at 25 V with 50 ms duration and 500 ms intervals using a BTX ECM830 square wave generator (BTX). Explants were then placed on polycarbonate filter discs (Dutscher) and cultured in DMEM/F12 medium (Invitrogen) with 10 mM HEPES pH 7.0 containing 5% FBS (Invitrogen). Cultures were maintained up to 9 days and media were changed every 2-3 days. The empty vector pCIG-eGFP used as control expresses the eGFP under the control of the chicken bactin promoter [40].

Image acquisition
The fluorescence of retinal explants was monitored after electroporation. Fluorescent signals were captured with a macroscope Z6 APO (Leica) equipped with a CCD CoolSnap camera (Roper Scientific). MetaMorph software (Universal Imaging Corporation) was used to quantify the intensity of the emitted fluorescence Immunofluorescence Retinal explants were fixed with 4% paraformaldehyde (PAF) in PBS for 10 min and incubated 30 min in 30% sucrose at room temperature. Retinal explants were frozen in PBS-10% sucrose-7.5% gelatin using a immersion in dry ice and isopentane. The samples were cut into 10 mm thick sections and mounted on glass slides. To reduce non-specific labeling, the sections were incubated for 2 h at room temperature in blocking solution (PBS, 0.1% Triton X-100 and 2% goat serum). The sections were incubated with the primary GFP-(Roche) and RFP-(Mbl) antibodies in blocking solution (respectively 1:400 and 1:2000) for 1 h at room temperature. Sections were washed three times in PBS and were then incubated with 4',6-diamidino-2-phenylindole (DAPI) and secondary antibodies (goat anti-rabbit AlexaFluor 594 and goat anti-mouse AlexaFluor 488, Invitrogen) diluted at 1:500 in PBS for 1 h at room temperature. After three washes in PBS, immunolabeled sections were mounted with Gel Mount (Biomedia) and were examined with a DM 5000 microscope (Leica) equipped with a CCD CoolSnap camera (Roper Scientific).

Statistical Analysis
All experiments were performed at least three times, each in triplicate. Data are expressed as the means 6 S.E.M. Comparison of means was performed using test. p,0.05 is considered significant.

Supporting Information
Table S1 Oligonucleotide PCR primers used to generate the various human and mouse NXNL1 and NXNL2 promoter constructs.