|Home | About | Journals | Submit | Contact Us | Français|
Orthodenticle (Otd)-related transcription factors are essential for anterior patterning and brain morphogenesis from Cnidaria to Mammals, and genetically underlie several human retinal pathologies. Despite their key developmental functions, relatively little is known regarding the molecular basis of how these factors regulate downstream effectors in a cell- or tissue-specific manner. Many invertebrate and vertebrate species encode two to three Otd proteins, whereas Drosophila encodes a single Otd protein. In the fly retina, Otd controls rhabdomere morphogenesis of all photoreceptors and regulates distinct Rhodopsin-encoding genes in a photoreceptor subtype-specific manner. Here, we performed a structure-function analysis of Otd during Drosophila eye development using in vivo rescue experiments and in vitro transcriptional regulatory assays. Our findings indicate that Otd requires at least three distinct transcriptional regulatory domains to control photoreceptor-specific rhodopsin gene expression and photoreceptor morphogenesis. Our results also uncover a previously unknown role for Otd in preventing co-expression of sensory receptors in blue vs. green-sensitive R8 photoreceptors. Sequence analysis indicates that many of the transcriptional regulatory domains identified here are conserved in multiple Diptera Otd-related proteins. Thus, these studies provide a basis for identifying shared molecular pathways involved in a wide range of developmental processes.
Otd/Otx-related homeodomain transcription factors are essential for rostral head and forebrain patterning throughout the animal kingdom (Acampora et al., 2005; Finkelstein and Boncinelli, 1994; Plouhinec et al., 2003; Reichert, 2005; Simeone and Acampora, 2001; Tallafuss and Bally-Cuif, 2002). In Drosophila, Otd is required for early head and brain segmentation, midline axon guidance, and the development of all known fly visual systems: the larval Bolwig organ, and the adult ocelli, Hofbauer-Buchner Eyelet, and compound eye (Cohen and Jurgens, 1990; Finkelstein and Perrimon, 1990; Finkelstein et al., 1990; Hirth et al., 1995; Leuzinger et al., 1998; Ranade et al., 2008; Royet and Finkelstein, 1995; Sprecher et al., 2007; Tahayato et al., 2003; Vandendries et al., 1996; Wieschaus et al., 1992). The molecular control of Otd-dependent functions during early embryonic patterning remains unclear, but at least three direct Otd targets are known within the compound eye (Tahayato et al., 2003), making the eye a useful system for dissecting Otd’s transcriptional regulatory properties.
The adult Drosophila eye is comprised of ~700 individual eye units, called ommatidia. Within each ommatidium, eight photoreceptors differentiate into two functionally and anatomically distinct neuronal types: outer and inner photoreceptor (OPRs and IPRs, respectively) (Hardie, 1985; Wolff and Ready, 1993). OPRs consist of six neurons (R1-R6) that function much like vertebrate rod photoreceptors for functioning in dim light and motion detection, while the two IPRs, the R7 and R8 neurons, similarly to vertebrate cone photoreceptors, discriminate color (Cook and Desplan, 2001; Hardie, 1985; Morante et al., 2007; Wernet and Desplan, 2004; Yamaguchi et al., 2008). OPRs develop light-gathering apical surfaces (rhabdomeres) that span the entire depth of the retina (~100 μm) and express the same broad-spectrum rhodopsin protein, Rh1. In contrast, IPRs have centralized, smaller, and stacked rhabdomeres, with the R7 positioned distally to the R8, and express four different rhodopsins, Rh3-Rh6, in a coordinated manner. The rhodopsins expressed in paired R7 / R8 cells within an ommatidium defines two major ommatidial subtypes: pale [p] ommatidia express UV-sensitive Rh3 in R7s and blue-sensitive Rh5 in R8s, while yellow [y] ommatidia express UV-sensitive Rh4 in R7s, and green-sensitive Rh6 in R8s (Chou et al., 1996; Fortini and Rubin, 1990; Mazzoni et al., 2008; Papatsenko et al., 2001; Zuker et al., 1985). Pale and yellow ommatidia are randomly distributed in the eye in a 30 :70 p :y ratio, and relies on an interplay between the R7 and R8. First, the decision to become a pale or yellow ommatidia occurs in the R7 cell, with the stochastic activation of the transcription factor Spineless (Ss) in 70% of R7s prior to opsin expression (Wernet et al., 2006). Ss is genetically required for Rh4 expression, and R7s that do not express Ss express Rh3. This "pR7" then transmits an unknown signal to the underlying R8 cell, where it activates a bistable loop between the signaling molecules Melted and Warts (Wts, a.k.a. Lats) (Mikeladze-Dvali et al., 2005). Melt, a pleckstrin homology domain-containing protein, is activated in pale R8 cells (pR8s), represses Wts, and is necessary for Rh5 expression, whereas Wts, a serine-threonine kinase, is expressed in yellow R8s (yR8s), represses Melt, and is necessary for Rh6 expression. How these signaling proteins affect gene expression, however, remains unclear.
Otd is expressed in all OPRs and IPRs, beginning shortly after neuronal cell specification in the late 3rd instar larval stage and persists throughout PR differentiation (Tahayato et al., 2003; Vandendries et al., 1996) (this study). While Otd does not appear to affect early photoreceptor specification, many aspects of PR terminal differentiation are Otd-dependent. For instance, during mid-pupation, Otd is necessary for proper rhabdomere morphogenesis in all PRs, and later during pupation and in adulthood, Otd activates Rh3 and Rh5 in pale IPRs, and represses Rh6 in OPRs (Mishra et al., 2010; Tahayato et al., 2003; Vandendries et al., 1996).
How does a single transcription factor such as Otd achieve such diverse cell-specific functions? Here, we developed an in vivo rescue paradigm that allowed us to perform a structure-function analysis of Otd during fly ocular development. Combining this approach with in vitro transcriptional regulatory assays, we identify separable portions of the otd coding region that are critical for regulating different aspects of photoreceptor differentiation, including a previously unrecognized role in preventing co-expression of Rh5 and Rh6 in R8 cells.
The homeodomain of Otd-related factors is over 85% identical in species ranging from Cnidaria to humans. Outside of this region, however, remarkably little sequence homology exists among Otd-related factors across species, confounding attempts to identify conserved functional domains. With the recent genomic sequencing of multiple insect species, we sought to reexamine potential conservation among Otd-related proteins. Using a 550 amino acid (aa) coding sequence derived from Otd’s longer alternatively spliced mRNA (Vandendries et al., 1996), we performed a multi-species alignment with other predicted Diptera Otd/Otx proteins. This alignment revealed that the entire coding region of Otd is largely conserved through at least ten species of Drosophilidae, and shares restricted conservation through more distant groups (Fig 1B).
Similar to the Drosophila melanogaster (D. mel) Otd protein, other Drosophila Otd orthologs contain polyhomomeric stretches of glutamines (N), alanines (A), glycines (G), and serines (S) (located at aa 129–297 in D. mel Otd), and a glycine-valine (GV) repeat sequence (aa 345-367 in D. mel Otd). Interestingly, such repetitive sequences are commonly found in transcriptional regulators and their expansion has been associated with a number of neurodegenerative diseases (Caburet et al., 2004; Faux et al., 2005; Hancock and Simon, 2005). In addition to global conservation among closely related Drosophilidae species, portions of the N-terminus and a region encompassing the C-terminus are conserved in mosquitos (Culicidae: Culex, Aedes, Anopholes, Fig 1B), and more weakly conserved in the red flour beetle (Tribolium castaneum, Coleoptera) and honey bee (Apis mellifera, Hymenoptera) (data not shown). This data suggests strong selective pressure for maintaining these domains. Also noteworthy is that the LDY sequence, previously proposed to represent a degenerate “Otx tail” in Otd (Acampora et al., 2005; Freund et al., 1998), is conserved through Culicidae.
Currently, only three direct Otd target genes have been identified: the rhodopsins Rh3, Rh5, and Rh6. These genes are expressed in a photoreceptor subtype-specific manner (Tahayato et al., 2003), and their expression patterns can be recapitulated using <500 bp regulatory sequence located directly upstream of the TATA box (Fortini and Rubin, 1990; Papatsenko et al., 2001; Tahayato et al., 2003; Zuker et al., 1985). Recent studies have demonstrated that many of the same factors that are necessary for controlling Rh3-Rh6 gene expression in vivo also properly regulate Rh promoter activity in cultured Drosophila S2 cells (Xie et al., 2007; Mishra et al., 2010). Otd, for instance, is critical for Rh3 and Rh5 expression in vivo, and activates the Rh3 and Rh5 promoters in vitro (Fig 2) (Xie et al., 2007). Interestingly, although Otd represses Rh6 in OPRs in vivo, Otd weakly activates the Rh6 promoter alone and synergistically activates Rh6 with Senseless (Sens) in vitro (Mishra et al., 2010; Tahayato et al., 2003; Xie et al., 2007). Since Sens is restricted to R8 photoreceptors and is required for Rh6 expression, but does not bind directly to the Rh6 promoter sequence, these in vitro experiments provide evidence that Sens is involved in activating Rh6 expression by interacting with Otd, and reveal that Otd may activate or repress Rh6 in different PR subtypes. Thus, testing Rh promoter regulation in this culture system is useful for defining and refining transcriptional regulatory processes.
Taking advantage of this reporter system, we tested whether the conserved domains in Otd described in Fig 1 contribute to its ability to regulate the Rh3, Rh5, or Rh6 promoters in vitro. For this, we created deletions of the N- and/or C-terminus, as well as deletions of the middle regions of the protein (Fig 2A). All constructs tested retain Otd’s DNA-binding domain, a domain that is required for its ability to activate of all three promoters (Supp Fig 1). In addition, each construct is expressed at similar levels (Supp Fig 2A), localizes to the nucleus (data not shown), and retains at least some of Otd’s functions in vivo (Fig 6).
We first focused on Otd-dependent activation of the Rh3 and Rh5 promoters. As shown in Figure 2B and 2C, a full-length Otd cDNA (OtdFL) activates Rh3 ~18-fold and Rh5 ~130-fold over basal activity. Removing Otd’s N terminus (OtdDelta;N) reduces this activity by ~60% on either promoter (Fig 2B, C), while removing the last 120 aa of Otd (OtdΔC) reduces Otd-dependent activity by 79% and 95% on Rh3 and Rh5 respectively. Removing just the last 20 aa, the region carrying the LDY motif (OtdΔC”), however, does not show a significant reduction in reporter activity compared to OtdFL, suggesting that the activation function of the C-terminus is not mediated by this “remnant” Otx tail.
We next tested internal deletions that remove different portions of Otd between its homeodomain and the C-terminal 430-550 activation domain. Removing the A region (aa 137-215) from Otd shows a significant increase in reporter activity on the Rh5 promoter (2.4 and 2.9-fold more than OtdFL with OtdΔA and OtdΔAB, respectively), while removing the B region alone (aa 215-430, OtdΔB) shows no significant change compared to Otd FL. These domains exhibit a similar trend on the Rh3 promoter, but are not statistically significant (Fig 2B, C). We also tested a construct that retains just the homeodomain and the AB domain (OtdΔNC), and this shows weak, but detectable promoter activity on both rhodopsin promoters (13% and 2% of OtdFL activity on the Rh3 and Rh5 promoters, respectively) (Fig 2B, C). Together, these data suggest that the N- and C-terminus of Otd combinatorially function as activation domains on both the Rh3 and Rh5 promoters, while the A domain suppresses Otd-dependent activation of Rh5.
Since Otd represses Rh6 in OPRs, yet weakly activates Rh6 alone and synergistically activates this promoter in the presence of Senseless in vitro, we next examined how the different Otd deletions affect Rh6 promoter activity with and without Senseless. Similar to the Rh3 and Rh5 promoters, Otd’s N- and C-termini both contribute significantly to Otd-dependent Rh6 activation in S2 cells (OtdΔN, OtdΔC) (Fig 2D), as removing either provides only ~25% of the activation observed with OtdFL, and removing both (OtdΔNC) leads to ~14% of OtdFL activation. In addition, removing Otd’s AB region leads to ~10-fold more activation of the Rh6 promoter than OtdFL, while individually removing the A or B region shows no statistical differences from OtdFL. Interestingly, although removing the last 20 aa of Otd shows a slight, but insignificant reduction in reporter activity on the Rh3 and Rh5 promoters, OtdΔC” shows a similar reduction in Rh6 promoter activity as OtdΔN and OtdΔC, suggesting this domain may rely on promoter context.
We next tested whether different regions of Otd contribute to its synergistic activation with Sens on the Rh6 promoter. As shown in Figure 2E, Sens causes a 10-fold increase in OtdFL’s ability to activate Rh6, and this same increase is maintained with OtdΔAB. OtdΔC, on the other hand, fails to activate Rh6 in the presence or absence of Sens. Surprisingly, though, OtdΔN shows even higher Sens-dependent activation (>75-fold) compared with OtdFL. These data suggest that the N-terminus suppresses Sens-dependent synergism with Otd, while the C-terminus is critical for Otd- and Otd/Sens-dependent activation of Rh6.
Combined, these experiments indicate that the C-terminus functions as an essential activation domain on all three Otd Rhodopsin targets, while other domains appear to function in a context-dependent manner. For instance, removing the N-terminus from Otd leads to reduced activation on all three promoters, but in the presence of Senseless, even higher synergistic activation of the Rh6 promoter is observed. We also find that while removing the A region enhances Rh5 activation, removing the entire AB region is required to cause a significant increase in Rh6 activation. Finally, the 20 aa “tail” domain of Otd only weakly contributes to Rh3 and Rh5 activity, yet functions similarly to the entire 120 aa C-terminus activation domain on Rh6 promoter activity.
We next examined whether the domains that, when removed from Otd, affect Rh target gene expression in vitro, are also sufficient to regulate transcription of a heterologous promoter. For this, we fused various portions of Otd (Fig 2F) to the DNA-binding domain (DBD) of the yeast transcription factor GAL4 and measured the ability of these fusion proteins to regulate a UAS-luciferase reporter. Western blot analysis indicates that these constructs are expressed at similar levels (Supp Fig 2B).
The GAL4 DBD lacks its own transcriptional activation domain, and only minimal basal luciferase activity is observed with this control (Fig 2G, DBD). In contrast, fusing the region that is downstream of Otd’s homeodomain (aa 129-550) to the DBD causes approximately 500-fold activation of the reporter (Fig 2G, “ABC”). Constructs containing aa 215-550 (Fig 2G, “BC”) or 304-550 (data not shown) also activate the report strongly, by ~250-fold, while the C-terminal domain (aa 430-550) that is necessary for Otd to activate rhodopsin gene expression, activates reporter expression by ~150-fold (Fig 2G, “C”). Further reducing the C-terminus to include only the last 90 aa (Fig 2G, “C’”) shows weaker, but significant, activation of reporter gene expression (~20X), while fusing the last 20 aa of Otd to the DBD shows no significant difference in activity compare to the DBD alone (inset in Fig 2G, “C””). Like the C-terminus, the N-terminus alone (aa 1-70, N) is able to activate heterologous gene expression, but only by ~40-fold. In contrast, the A, B, and AB regions do not cause a significant increase in basal DBD activity, but rather, the entire AB region is sufficient to mediate weak repression (~30%, inset in Figure 2G). Together, these data support our rhodopsin reporter assays that a potent transcriptional activation domain resides between residues 430-530, that a weaker activation domain is present within the N-terminus, and that the AB region may function as a repression domain. However, since the ABC region shows higher activation than the C region alone, the AB region may participate in activation and/or repression based on protein and/or promoter context.
To test whether the transcriptional regulatory domains identified in vitro contribute to Otd’s functions during photoreceptor differentiation in vivo, we developed an eye-specific Otd rescue paradigm. Previous studies have shown that transient heat shock-inducible expression of a full-length Otd cDNA is sufficient to rescue otd mutants (Leuzinger et al., 1998; Nagao et al., 1998; Tahayato et al., 2003; Vandendries et al., 1996). However, this approach causes mosaic misexpression of Otd throughout the organism, making cell- and tissue-specific functions difficult to dissect. Thus, we developed a GAL4/UAS-based system that takes advantage of a 1.6 kb enhancer present within otd’s third intron. This region is deleted in the eye-specific otduvi mutant and is sufficient to drive reporter gene expression in all photoreceptors (Vandendries et al., 1996). As shown in Fig 3A and B, this enhancer cloned upstream of the yeast GAL4 transcription factor (otd1.6-GAL4) activates a UAS reporter similar to endogenous Otd expression: its expression initiates soon after the completion of photoreceptor specification in late third instar eye imaginal discs (Fig 3A-A”), and is maintained throughout photoreceptor development (Fig 3B and data not shown). We also observe limited reporter expression in a subset of neurons within the medulla neuropil of the optic lobe with this driver (data not shown, and (Morante and Desplan, 2008). Importantly, otd1.6-GAL4 remains active in otduvi mutants (Fig 3C), making it useful for expressing various UAS-transgenes in an otduvi background.
To verify that otd1.6-GAL4 driving a full-length Otd cDNA (otd1.6 >OtdFL) is sufficient to rescue otduvi mutant eye phenotypes, we compared ommatidia morphology and Rhodopsin gene expression patterns between control, otduvi, and otduvi flies carrying the otd1.6 -GAL4 and UAS>OtdFL transgenes (otduvi; otd1.6>OtdFL, referred to here as Otd rescues). Morphologically, control outer photoreceptor (OPR) rhabdomeric membranes span the entire depth of the retina (~100 μm) (Fig 3D), and their actin-rich surfaces, recognized by fluorescently-labeled phalloidin, form highly regular, trapezoidal arrangement surrounding the smaller rhabdomere of the R7 (or R8) (Fig 3E). In contrast, OPR rhabdomeres in otduvi flies fail to extend more than 1/3 the depth of the retina (Fig 3H), and these rhabdomeres fail to form properly (Mishra et al., 2010; Tahayato et al., 2003; Vandendries et al., 1996), making individual rhabdomeres difficult to visualize with phalloidin (Fig 3I). Similar to control flies, OtdFL rescues develop fully-elongated and organized rhabdomeres (Fig 3L, M).
Molecularly, Rh3 and Rh5 expression are expressed in subsets of R7 and R8 cells, respectively, in control flies (Fig 3F,G) but are absent in otduvi flies (Fig 3J,K). In addition, Rh6 expression is restricted to a subset of R8 photoreceptors at the base of the retina in control flies (Fig 3G), but is derepressed into OPRs in otduvi flies (Fig 3K and Fig 5A, otduvi) (Tahayato et al., 2003). In OtdFL rescues, Rh3 and Rh5 are restored to subsets of R7 and R8 cells, respectively (Fig 3N,O), and Rh6 is no longer expanded into OPRs, restricted to a subset of R8 cells (Fig 3O). Thus, re-expressing the full-length Otd protein is sufficient to rescue otduvi mutants. We do note, however, that the number of Rh5-expressing cells is reduced in OtdFL rescue flies compared to yw control or wild-type eyes: only ~15% of all R8 cells express Rh5 in otd1.6 >OtdFL rescue flies (Fig 3O), while ~30% of R8s express Rh5 in yw or wild-type control flies (Fig 3G and data not shown). Since the correct ratio of Rh3-expressing R7 cells is achieved in OtdFL rescues, this results in ommatidia that have miscoupling between Rh3-expressing R7s and Rh6-expressing R8s (data not shown). Identical results were observed with two individual OtdFL lines or by expressing both insertions of OtdFL together. Importantly, in otduvi heterozygotes, we also observe miscoupled Rh3/Rh6 ommatidia and reduced numbers of Rh5-expressing cells, indicating that Otd levels are important for properly establishing the p:y ratio in R8 photoreceptors (Fig 3P). Consistent with this, misexpressing OtdFL in otduvi heterozygous flies restores correct Rh3/Rh5 coupling (Fig 3Q). Thus, we conclude that OtdFL is able to rescue Rh5 expression, but does not reach levels equivalent to wild-type Otd. To further verify that the reduced number of Rh5-expressing cells in OtdFL rescues is not due to an inherent inability of the OtdFL transgene to activate Rh5, we co-expressed OtdFL with Melt, a factor that transforms R8 cells into Rh5-expressing pR8s (Mikeladze-Dvali et al., 2005). Consistent with a requirement for Otd to activate Rh5, no Rh5 expression is detected in otduvi mutants that misexpress Melt alone (Fig 5I); however, similar to control eyes misexpressing Melt (Fig 5E), otduvi mutants co-expressing OtdFL and Melt express Rh5 in almost all R8 cells (Fig 5O). Thus, for the remainder of the studies described here, we report qualitative, not quantitative differences in rhodopsin expression relative to OtdFL rescues. We also verified each construct for their ability to activate Rh5 in the presence of Melt, data which is included in Supplemental Figure S3.
Using the rescue paradigm described above, we next tested the same deletion constructs in vivo as we assayed in vitro. In addition, we created two additional constructs, OtdΔABC and OtdHD, to test the contribution of the N-terminus alone to various functions. Below, we describe our findings as they relate to individual aspects of otduvi phenotypes (Fig 4 & 5) and summarize the results in Figure 6.
All constructs except OtdΔNC, OtdΔAB, and OtdHD rescue Rh3 expression. However, empty R7 cells are more frequently observed with OtdΔC, OtdΔA, and OtdΔABC (circles, Fig 4A) than OtdFL, OtdΔN, and OtdΔB, suggesting these are less effective in Rh3 activation. We also note that all Rh3-expressing cells in OtdΔABC rescues co-express Rh4 (arrows, Fig 4A), and are restricted to the dorsal region of the eye, suggesting that OtdΔABC is only able to rescue Rh3 in a recently-described specialized subset of dorsal ommatidia that weakly co-express Rh3 with Rh4 (Mazzoni et al., 2008). Interestingly, although OtdΔABC is able to restore at least some Rh3 expression, no Rh3 is detected in rescues with OtdΔAB. This is not likely due to an inability of OtdΔAB to recognize and regulate gene targets since this construct activates transcription more strongly than OtdFL in S2 cells, and is the strongest activator of Rh5 expression in R8 cells (see below). Other possibilities for this finding are described in the Discussion. Nevertheless, based on the findings that OtdΔN, OtdΔC, and OtdΔABC maintain at least one activation domain identified in vitro and each can activate Rh3, while a construct lacking both the N- and C-terminus (OtdΔNC) shows no Rh3 rescue, these data support the possibility that both the N- and C-terminus contribute to Rh3 activation in vivo, and that the N-terminus is sufficient to partially activate Rh3.
Rh5 is re-expressed in otduvi mutants rescued with Otd factors that lack the A, B, or AB domains, but is absent when rescued with constructs lacking either the N- or C-terminus. However, distinct phenotypes are observed in rescues with OtdΔN and OtdΔC. First, in OtdΔN rescues, all R8 cells express Rh6 and very few empty R8 cells are detected, whereas in OtdΔC rescues, Rh6 is restricted to a subset of R8 cells, similar to wild-type animals, and cells expressing neither Rh5 or Rh6 are frequently observed (circles, Fig 4B). These data suggest that in OtdΔC rescues, the pale subset of ommatidia is largely maintained, whereas in OtdΔN rescues, the yellow subset is expanded. Second, co-expressing OtdΔN with Melt shows very weak, but detectable Rh5, whereas no Rh5 is observed with constructs lacking the C-terminus (Supp Fig 3), indicating that OtdΔN maintains some ability to activate Rh5, while OtdΔC does not. Interestingly, in flies rescued with OtdΔA, ΔB, and ΔAB flies, we note a unique and consistent phenotype: Rh5 and Rh6 are co-expressed in a subpopulation of R8 cells. In OtdΔA flies, all cells expressing Rh5 also express Rh6, whereas in OtdΔB and OtdΔAB flies, a subset of R8 cells express only Rh5, a subset expresses both Rh5 and Rh6, and a subset expresses only Rh6. Together, these data suggest that the C-terminus is essential for Rh5 expression, that the N-terminus is important to prevent Rh6 expression in pR8s and can weakly activate Rh5, and that the A and B regions are involved in preventing co-expression of R8 opsins.
As previously mentioned, Otd is necessary for preventing Rh6 expression from being inappropriately expressed in outer photoreceptors (OPRs). Surprisingly, we find that this opsin is not localized to the actin-rich rhabdomeric membrane like most Rhodopsin proteins in control or otduvi mutants, but instead is present in OPR cytoplasm (Fig 5A). Proper repression of Rh6 in OPRs is observed when rescued with Otd proteins lacking the N, A, B, or AB regions (Fig 5A and data not shown), but Rh6 expression in OPR cytoplasm is maintained in flies rescued with constructs lacking the C-terminus, including OtdDelta;C, OtdDelta;NC, OtdDelta;ABC, and OtdHD (Fig 5A). Thus, the C-terminus is necessary to repress Rh6 in OPRs.
Each deletion construct previously tested in vitro maintains at least some ability to restore proper photoreceptor morphogenesis (Fig 5B). However, only OtdΔN rescues similarly to OtdFL, whereas rescues with OtdΔC, OtdΔNC, OtdΔA, OtdΔB, OtdΔAB, and OtdΔABC only partially rescue trapezoid formation (best seen in sections at the R7 layer) and rhabdomere elongation (best seen in sections at the R8 layer). We note, though, that constructs lacking the C-terminus show fewer rhabdomeres reaching the R8 layer compared to constructs that retain this domain, suggesting that the C-terminus is particularly important for rhabdomere elongation. Since the smallest construct, OtdΔABC, is still able to partially rescue morphogenesis, and since the only common region in all of the deletions is the homeodomain itself, we also tested the ability of the homeodomain of Otd alone (OtdHD) to rescue morphogenesis. As shown in Fig 5B HD, distinct rhabdomeres are present in the R7 layer in OtdHD rescues that are more defined than in otduvi mutants, and more rhabdomeres have extended into the R8 layer than otduvi mutants, suggesting that the homeodomain can, albeit weakly, rescue some aspects of photoreceptor development. However, since no other constructs except OtdΔN rescue rhabdomeres to the same extent as OtdFL, it is likely that Otd utilizes multiple domains for regulating photoreceptor morphogenesis.
The results presented here yield interesting mechanistic insights into how Otd regulates gene expression during eye development. First, the activation of the two pale ommatidia-specific rhodopsins, Rh3 and Rh5, can be genetically separated; second, Rh5 and Rh6 co-expression in IPRs and Rh1 and Rh6 co-expression in OPRs is prevented using distinct mechanisms; and third, Otd likely functions at many levels to control the appropriate ratio of blue- and green-sensitive photoreceptors in the adult eye. Below, we describe the significance of these findings as they relate to different regulatory domains within Otd.
Rh3 and Rh5 expression is coupled between pR7 and pR8 cells, respectively, suggesting that a “pale-, IPR-specific” factor could similarly contribute to Otd’s ability to activate both genes. However, increasing evidence suggests that Rh3 and Rh5 expression may rely on different regulatory processes. First, the onset of Rh3 vs. Rh5 expression is different during pupation (Earl and Britt, 2006); second, several experiments now suggest that Rh3 is the “default” IPR opsin, whereas Rh5 expression requires an inductive signal from R7 cells (Chou et al., 1999; Mikeladze-Dvali et al., 2005; Papatsenko et al., 2001; Wernet et al., 2006; Xie et al., 2007); and third, we find that Otd’s C-terminus is essential for Rh5 activation in pR8s, whereas the N-terminus is sufficient to activate Rh3 expression (OtdΔABC, Fig 4A), demonstrating that the activation of these two opsins are functionally separable. N-terminus-mediated activation of Rh3, however, is only observed in the dorsal third of the eye, a region recently shown to have weaker but more widespread Rh3 expression than in the ventral part of the eye (Mazzoni et al., 2008). Since this dorsal region appears more “permissive” for Rh3 expression, it is likely that the N-terminus provides weak activation potential while the C-terminus may be important for “boosting” this expression in the remainder of the eye. One finding that is inconsistent with this model for Rh3 activation is that OtdΔAB, which maintains both the N- and C-terminus activation domains, is unable to activate Rh3. One possibility for this is that removing the AB domain creates a highly misfolded protein that prevents Rh3 activation. However, since OtdΔAB is able to activate Rh3 similar to OtdFL in vitro and is the most efficient construct tested in vivo for activating Rh5, this explanation seems unlikely. Another possibility is that Otd normally activates a repressor of Rh3 that is normally involved in preventing Rh3 expression in yR7 cells, and that the increased transcriptional activation by OtdΔAB inappropriately activates this repressor in pR7s. Indeed, such an Otd-dependent yR7-restricted Rh3 repressor has been recently identified (Robert Johnston, Claude Desplan, personal communication). While future experiments will be required to formally test this possibility, this postulate underscores that much remains unknown related to how Rh3 and Rh4 are properly regulated in R7 cells, and whether Otd plays direct and/or indirect roles in this process.
In terms of Rh5 activation, our results suggest that the C-terminus is essential for Rh5 expression, whereas the N-terminus is instead involved in promoting Rh5 activation and Rh6 repression in pR8s. For instance, in constructs that contain the N-terminus, but lack the C-terminus (e.g. OtdΔC and OtdΔABC), Rh5 is absent and a subset of R8 cells (presumably pR8s) are empty, whereas flies rescued with OtdΔN show no Rh5 expression but Rh6 is expanded into all R8s. One possible mechanism for how the N-terminus may function emerges from our in vitro studies: alone, Otd activates the Rh5 promoter stronger than Rh3 or Rh6 (~130-fold vs <20-fold). In the presence of Sens, however, Otd increases its ability to activate Rh6 to >100-fold (Fig 2E), while Rh5 promoter activity is only marginally affected (Xie et al., 2007). Interestingly, removing Otd’s N-terminus shows a dramatic increase in Otd/Sens synergism on Rh6 but decreases Otd-dependent activation of the Rh5 promoter in vitro. Similarly, Rh6 is expanded into pR8s while Rh5 is undetectable in flies rescued with OtdΔN in vivo. Thus, the N-terminus may be important for controlling the decision between whether Otd activates Rh5 in pR8s by itself or that Otd/Sens synergize to activate Rh6 in yR8s. Another possibility, not mutually exclusive from the first, is that Otd itself is important for regulating the Melt/Wts pathway, and that the N-terminus is important for this. Evidence for this model comes from our findings that otduvi heterozygotes show miscoupled Rh3/Rh6 ommatidia, indicating that Otd levels specifically affect R8 p:y ratio decisions. Unfortunately, due to the severe disruption in photoreceptor morphology, the derepression of Rh6 into OPR cytoplasm, expansion of Rh6 into some R7 cells, and the failure of R8 cells to be positioned to the proximal portion of the retina, phenotypes all associated with otduvi homozygous mutants (data shown here and (Tahayato et al., 2003; Vandendries et al., 1996), it is currently difficult to accurately assess Otd’s contribution to Rh5:Rh6-expressing R8 cells. Thus, future experiments that make use of cell-specific loss-of-function studies with Otd should help address these questions in more depth, and provide a better understanding of the potential role for Otd and Sens in activating Rh6 in yR7 cells.
Besides likely distinct roles for the N- and C-terminus in activating Rh3 and Rh5 in pale ommatidia, we also found in the current study that regions between Otd’s homeodomain and C-terminal activation domain are critical for preventing Rh5 and Rh6 co-expression in R8 cells. For instance, unlike rescues with OtdFL, OtdΔN, and OtdΔC, in which Rh5 and Rh6 are mutually exclusive, rescues with OtdΔAB and to a lesser degree, OtdΔA and OtdΔB, show co-expression of Rh5 and Rh6 in a subset of R8s. Currently, due to the disruption in morphology with many of our rescues, as well as the inability to fully rescue pR8s even with OtdFL, it is difficult to determine whether this co-expression is specific to pR8s or yR8s. However, these findings emphasize that cells must actively express one opsin and simultaneously repress the other, and suggest that Otd is important in this process in R8 cells. Since our in vitro studies indicate that removing the AB region from Otd causes a significant increase in Rh5 and Rh6 promoter activity, and transferring this domain to a heterologous DNA-binding domain is sufficient to weakly repress transcription, we speculate that the AB region is important for preventing Otd-dependent activation of Rh5 and/or Rh6 in different R8 subsets. Because co-expressing OtdΔAB with Melt or Wts leads to the same expansion of Rh5 and Rh6, respectively, as OtdFL or control flies (Supp Fig 3), it is likely that this function of Otd lies upstream of Melt/Wts. On the other hand, since Otd is also essential for Rh5 activation, Otd almost certainly functions downstream of Melt as well. Thus, elucidating the role of Otd in the Melt/Wts pathway will be a challenge, but nevertheless is an exciting avenue of future research.
Besides preventing Rh5 and Rh6 co-expression in R8 cells, Otd is also critical for preventing the IPR Rh6 from being co-expressed with the default OPR Rh1 (Tahayato et al., 2003). While the AB domain appears to be involved in the mutual exclusive expression of Rh5 and Rh6, the C-terminus is critical to prevent Rh1 and Rh6 co-expression in OPRs (Fig 5A). Currently, it is not clear how Otd prevents Rh6 expression in OPRs, but since the C-terminus behaves as a potent transcriptional activation domain both in vivo and in vitro, one possibility is that Otd indirectly represses Rh6 by activating an Rh6 repressor in OPRs. Since the Rh6 promoter requires a K50 binding site for OPR repression, such a repressor would either need to be a K50 homeodomain itself, or cooperate with Otd to mediate this function. Consistent with the former possibility, a K50-encoding, Otd-activated target gene, Dve, has recently been shown to be expressed in OPRs and thus may represent such a candidate (R.Johnston and C.Desplan, personal communication and ref). Thus, experiments aimed at testing whether Otd activates Dve via its C-terminus should aid in identifying the mechanism by which Otd regulates Rh6 expression in OPRs.
Another finding from this work is that the localization of Rh6 when it is de-repressed in OPRs is cytoplasmic rather than rhabdomeric. This may be attributed to the fact that Rh1 is still expressed in these cells, and thus the cells are not able to incorporate more opsin protein into the rhabdomeres. Alternatively, the defects in morphogenesis that occur in otduvi mutants may not allow proper insertion of the opsin into rhabdomere. Consistent with this possibility, all rescue constructs that fail to repress Rh6 also fail to fully restore photoreceptor morphogenesis. However, it remains possible that Otd’s C-terminus is also important for regulating additional targets involved in protein trafficking. Indeed, recent microarray studies of otduvi mutants have uncovered several such candidates (Mishra et al., 2010; Ranade et al., 2008). Thus, further analysis of other Otd target genes with the different Otd deletions may be informative for better understanding the likely complex role of Otd in regulating photoreceptor morphogenesis.
Like Otd, Otd-related factors are important for opsin regulation and photoreceptor morphogenesis in a number of developmental models, suggesting that these factors may share common regulatory properties. Unfortunately, outside of the homeodomain, remarkably little sequence homology has been reported among Otd-related factors, limiting approaches such as sequence comparisons for identifying such functional domains. Two notable exceptions are the “WSP” and “Otx tail” domains (Acampora et al., 2005). These domains were originally identified in the middle region and C-terminus, respectively, of the vertebrate family of Otd-related factors (Otx1, Otx2, and Crx), and have been subsequently found in other Otd-related factors (Browne et al., 2006; Freund et al., 1997a; Furukawa et al., 1997; Williams and Holland, 1998). Surprisingly, neither domain has been identified within Drosophila Otd. However, the tri-peptide sequence LDY in Otd’s C-terminus could represent a reduced “Otx tail” (Acampora et al., 2005) (Fig 1A). To test this possibility, we removed the LDY sequence from Otd (OtdDelta;C”) and find all in vivo and in vitro functions we measured replicate OtdFL (Fig 2 and data not shown), with the exception that this domain appears to participate in Rh6 activation in cultured cells (Fig 2D). We also tested whether this domain is sufficient to activate a heterologous promoter like other Otx tails (Chatelain et al., 2006; Chau et al., 2000), but at least in S2 cells, this is not the case (Fig 3C”). Therefore, we postulate that this domain is not functionally relevant in Otd, at least during eye development. With regards to the WSP motif, we note from our sequence analysis that Otd contains seven copies of a di-amino acid repeat, SP, throughout its B and C regions (italicized in Fig 1A) which are similar in position with WSP domains in other Otd proteins. Thus, it cannot be ruled out that a WSP-like motif is functionally represented in Drosophila Otds. Nevertheless, our studies reveal that Otd’s N- and C-terminal domains, which contribute to several of its functions in vivo, are conserved beyond the genus Drosophila. Moreover, amino acid homopolymers downstream of Otd’s homeodomain, included in regions that participate in rhodopsin sensory receptor exclusion, are represented in all Drosophila Otds. Hence, future rescue experiments using Otd proteins from other species should be useful for further understanding how Otd-related factors participate in common processes across phyla.
The construction of the pUAST vectors carrying OtdFL, OtdDelta;N, OtdDelta;B, OtdDelta;C, and OtdDelta;C” (aka OtdDelta;tail) were described in Reischl (2002). Additional Otd deletion constructs were generated by PCR amplification using the OtdFL cDNA as a template, and verified by sequencing. Otd cDNAs were subcloned into pUAST (Brand and Perrimon, 1993) for in vivo rescue experiments, and into pAc5.1-His (Invitrogen) for S2 luciferase assays.
Gal4DBDmyc-Otd fusion constructs were generated as follows: the 3x Myc tag and Gal4DBD sequences were PCR amplified from pCMV-3tag-2C (Invitrogen) and pAsΔCyh2 (gift from Isabelle Brun, NYU) respectively, and subcloned into pAc 5.1 HisA (Invitrogen) with Gal4DBD upstream and in-frame with the 3x Myc tag. PCR-amplified fragments of Otd were then cloned downstream of the Myc tag. The Gal4 responsive luciferase reporter construct, UAS-act250-Luc was previously described (Gebelein et al., 2004).
The otd1.6-GAL4 driver was created as follows: pGAL4-pCHAB was generated by cloning the BamHI/SpeI GAL4-encoding fragment from pGaTB (Brand and Perrimon, 1993) into BamHI/XbaI-digested pCHABΔSal (Wimmer et al., 1997). The MCS and hs43 minimal promoter from pCaSpeR hs43LacZ (kindly provided by C. Thummel, U. Utah) was PCR amplified as an EcoR1/BglII fragment and cloned upstream of.GAL4 in pGAL4-pCHAB to create hs43GAL4-pCHAB. Finally, the 1.6kb BamHI/KpnI fragment from Otd’s third intron, previously identified as a minimal enhancer (Vandendries et al., 1996), was PCR amplified from yw67 genomic DNA and cloned upstream of hs43GAL4-pCHAB.
The rhodopsin promoter-luciferase reporters Rh3-Luc, Rh5-Luc, and Rh6-Luc, and the pAc-LacZ construct used for transfection normalization, were previously described (Xie et al., 2007). Drosophila S2 cells (Invitrogen) were maintained in HyQ SFX-Insect media (Hyclone) at room temperature (RT, 20-22 C). 1×106 cells in 1 ml HyQ SFX were plated in 12-well tissue culture dishes (Corning) 24 hours prior to transfection with 1.5 μL Fugene HD (Roche). For Rh-Luc assays, 150 ng pGL3 reporter, 150 ng pAc-LacZ and 150 ng pAc-Otd deletion constructs were transfected in triplicate, and for the GAL4-based reporter assays, 200 ng pAc-lacZ, 200 ng UAS-37tk-Luc (gift from Albert Courey) or UAS-250act-Luc, and 200 ng pAc 5.1 HisA 2C-Gal4DBDmyc-Otd fusion proteins were transfected in triplicate. Either reporter gave similar results. 48 hours post-transfection, luciferase assays were performed as previously described (Xie et al., 2007). Results were obtained from at least three independent experiments, and average, normalized luciferase values were pooled to calculate standard error and to generate the graphs presented. Statistical analysis was performed using a t-test assuming equal variance.
For western blot analysis, the remaining luciferase assay lysate from each condition (35 μL) was mixed with 15 μL 20mM MgCl2, 2U DNase (New England Biolabs), and 4X SDS loading buffer. 10 μL of the resultant lysate was separated on a 12% SDS-PAGE gel and transferred to Immobilon-Psq PVDF membrane (Millipore) using standard procedures. Membranes were blocked with 5% nonfat milk, stained with primary antibody for 2 hours at RT, then stained with secondary HRP antibodies for 1 hour at RT. Blots were developed with the ECL Plus western blotting detection system (GE Healthcare) and imaged using a Storm 860 scanner (Molecular Dynamics). Antibody dilutions used were: guinea pig anti-Otd (1:750, Xie et al., 2007), rabbit anti-β-galactosidase (1:1000, Rockland Immunochemicals), mouse anti-FLAG (1:1000, Sigma-Aldrich), donkey anti-guinea pig HRP (1:5000, Jackson Immunoresearch), donkey anti-rabbit HRP (1:5000, Jackson Immunoresearch), and donkey anti-mouse HRP (1:5000, Jackson Immunoresearch).
pUAST and otd1.6-GAL4 P-element vectors were injected into yw67 flies using standard transgenic methods (Terry Blackman, NYU, and Rainbow Transgenics). Several GAL4 lines were tested, and O2-GAL4 was retained for further analysis because of its ability to best rescue otduvi phenotypes with a single copy of UAS-OtdFL. Two insertions of each UAS-driven Otd deletion were tested, both individually and simultaneously, and gave similar results. pWIZ, an RNAi line against the white gene (Lee and Carthew, 2003), was recombined onto the Otd-GAL4 line to prevent autofluorescence created by eye pigmentation. Thus, a typical cross was: otduvi; otd-GAL4, pWIZ8; TM2/TM6B x otduvi; Sp/CyO; UAS-Otd. Cryosections, whole mount retinal dissections, and antibody stainings of adult retinas were performed as previously described (Xie et al., 2007). In addition, some analysis was performed using an agarose-embedding method prior to cryosection (ref). Briefly, adult heads were bisected and fixed for 1 hour in 3.2% paraformaldehyde and 1mM dithiobis(succinimidyl) propionate (Pierce) in 0.1M phosphate buffer (pH7.0). After fixation, the heads were washed in phosphate buffer, embedded in 2% agarose/phosphate buffer, infiltrated with 10% sucrose dissolved in phosphate buffer for 30 minutes, and incubated with 25% sucrose solution overnight at 4°C. The tissue was snap frozen in melting isopentane, and cryosectioned at 8 microns. The tissue sections were rehydrated with 0.01% Tween-20, reacted with 50 mM NH4Cl for 10 min, washed with PBS, blocked for 30 minutes in 1% normal goat serum/0.8% bovine serum albumin/0.5% Triton-X in PBS for 30 minutes, and incubated with the following primary antibodies overnight at 4°C: mouse-anti Rh3, 1:100; rabbit-anti Rh4 1:100, mouse-anti Rh5 1:50, rabbit-anti Rh6 1:100; phalloidin, 1:100 (Invitrogen). The tissue was washed in PBS, incubated in secondary antibody for 1 hour at room temperature, and after a final PBS wash, were mounted with Prolong antifade reagent (Invitrogen). Samples were imaged using a Zeiss ApoTome Imager Z1 microscope. Monoclonal antibodies against Rh3 and Rh5 were kindly provided by Steve Britt (Univ Colorado), and polyclonal antibodies against Rh4 and Rh6 were provided by Charles Zuker (Univ California San Diego) and Claude Desplan (NYU), respectively.
Figure S1 : Otd homeodomain (HD) is required for its ability to activate its target promoters. Luciferase reporter assays of the Rh3, Rh5, and Rh6 minimal promoters show that a full-length Otd protein with just the homeodomain deleted (OtdDelta;HD) loses its ability to regulate all three promoters.
Figure S2: Western blot analysis of S2 lysates transfected with Otd deletion constructs used for Rhodopsin reporter assays (A) or with myc-DBD-fused Otd deletions used for UAS-Luciferase assays (B). Westerns were immunoblotted with α-Otd (A) or α-Myc (B) antibody, and subsequently reblotted with α-β-galactosidase that was co-transfected as an internal control.
Figure S3: R8 Rhodopsins Rh5 (A) and Rh6 (B) expression in otdUVI flies rescued with Otd FL or Otd deletion constructs in the absence (rescue) or presence of UAS-melt (+ Melt) or UAS-wts (+ Wts). Compared to control rescues, Melt co-expression results in increased numbers of Rh5-expressing R8 cells with all constructs tested except OtdΔC and OtdΔNC, and only very weak Rh5 expression is observed with OtdΔN. UAS-Melt also decreases the number of Rh6– expressing cells compared with control rescues (N.B. UAS-Melt frequently does not fully convert all R8 cells to pR8s when co-expressed with another UAS-transgene such as the Otd deletions). UAS-Wts co-expression with Otd rescues leads to loss of Rh5 and increased numbers of Rh6-expressing cells with all constructs (N.B. few Rh6-positive rhabdomeres are observed with Otd^NC + Wts due to high levels of cytoplasmic Rh6 with this mutant).
The authors wish to thank Nadean Brown, Terry Blackman, Steve Britt, Isabelle Brun, Arzu Celik, Albert Courey, Claude Desplan, Markus Friedrich, Kerstin Meier, Carl Thummel, Charles Zuker, past and present members of the Cook and Gebelein lab, and the Development Studies Hybridoma Bank for reagents and/or input. This work was supported by Ohio Preventing Blindness and NIH T32 HD046387 (ECM), the Boehringer Ingelheim Funds (JR), the German Research Foundation (DFG Wi 1797/2-2 to EAW), Cincinnati Children's Research Foundation, Research Preventing Blindness, the E. Matilda Ziegler Foundation for the Blind, and NIH R01-EY017907 (TAC).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.