Regional modulation of a stochastically expressed factor determines photoreceptor subtypes in the Drosophila retina

Abstract

Stochastic mechanisms are sometimes utilized to diversify cell fates, especially in nervous systems. In the Drosophila retina, stochastic expression of the PAS-bHLH transcription factor Spineless (Ss) controls photoreceptor subtype choice. In one randomly distributed subset of R7 photoreceptors, Ss activates Rhodopsin4 (Rh4) and represses Rhodopsin3 (Rh3); counterparts lacking Ss express Rh3 and repress Rh4. In the dorsal third region of the retina, the Iroquois Complex transcription factors induce Rh3 in Rh4-expressing R7s. Here, we show that Ss levels are controlled in a binary On/Off manner throughout the retina, yet are attenuated in the dorsal third region to allow Rh3 co-expression with Rh4. Whereas the sensitivity of rh3 repression to differences in Ss levels generates stochastic and regionalized patterns, the robustness of rh4 activation ensures its stochastic expression throughout the retina. Our findings show how stochastic and regional inputs are integrated to control photoreceptor subtype specification in the Drosophila retina.

Introduction

The Drosophila eye provides an excellent paradigm to study how stochastic and regionalized regulatory inputs intersect to affect cell fate specification. Underlying its uniform morphology, the fly eye contains two randomly distributed subtypes of ommatidia (unit eyes) defined by the mutually exclusive expression of specific Rhodopsin (Rh) proteins in the inner photoreceptors (R7 and R8). In the pale (p) subtype, pR7s express Rhodopsin3 (Rh3) and pR8s express Rhodopsin5 (Rh5), whereas in the yellow (y) subtype, yR7s express Rhodopsin4 (Rh4) and yR8s express Rhodopsin6 (Rh6). Though the p and y subtypes are randomly distributed, they consistently occur in a p:y ratio of approximately 35:65 (Bell et al., 2007; Johnston Jr and Desplan, 2010) (Figure 1A–B, 1D, 1F). Throughout the majority of the retina, the mutually exclusive expression of Rhs defines the p and y ommatidial subtypes. However, in the dorsal third region of the retina, Rh3 is co-expressed with Rh4 in yR7s. Thus, the dorsal third region consists of p ommatidia containing pR7s that express Rh3 only and “dorsal third y” ommatidia containing yR7s that express both Rh4 and Rh3 (Figure 1C–E) (Mazzoni et al., 2008).

Figure 1. Rhodopsin expression in the fly eye.

For A-C, top left indicates rhabdomeres (membranous structures containing Rh proteins) within an ommatidium. Top right indicates cross-sections in the R7 (top) and R8 (bottom) layers. Gray indicates cell bodies and nuclei. White circles with black outlines indicate outer PR rhabdomeres. Central, colored rhabdomeres indicate R7 (top) or R8 (bottom). Below, the regulatory network controlling Rh expression in R7 (top) or R8 (bottom).

• A.
  pale: Rh3 (blue) is expressed in pR7s and Rh5 (purple) is expressed in pR8s. In the absence of Ss in pR7s, Sal represses Dve and acts with Otd to induce Rh3 expression. Rh4 expression is not activated. In the absence of Ss, a signal is de-repressed which upregulates expression of the growth regulator Melted (Melt) in R8s. Melt represses the tumor suppressor Warts (Wts) inducing expression of Rh5 and repression of Rh6.
• B.
  yellow: Rh4 (red) is expressed in yR7s and Rh6 (green) is expressed in yR8s. In yR7s, Ss activates Rh4 and Dve, which represses Rh3. Ss also represses the signal to R8. Melt is not expressed in the absence of the signal, allowing for Wts expression inducing expression of Rh6 and repression of Rh5.
• C.
  dorsal third yellow: Rh3 (blue) and Rh4 (red) are expressed in yR7s and Rh6 (green) is expressed in yR8s. In yR7s, Ss activates Rh4 and Dve. In the dorsal third region, Ss levels are reduced to allow for IroC-mediated activation of Rh3. Ss also represses the signal to R8. Melt is not expressed in the absence of the signal, allowing for Wts expression inducing expression of Rh6 and repression of Rh5.

For D-F, dorsal is up and ventral is down.

• D.
  In the main part of the retina, Rh3 and Rh4 are expressed in stochastic and exclusive subsets of R7s. Main part = below white dashed line; dorsal third region = above the white dashed line.
• E.
  Higher magnification view of dorsal third region. In the dorsal third region, Rh3 is expressed in all R7s including Rh4-expressing yR7s. Top (Rh3 and Rh4) and bottom (Rh3 alone).
• F.
  Rh5 and Rh6 are expressed in stochastic and exclusive subsets of R8s.

Stochastic ommatidial subtype specification is controlled by the PAS-bHLH transcription factor Spineless (Ss) (Wernet et al., 2006). Ss is expressed in a random subset of R7s where it determines y subtype fate. In yR7s, Ss has three main functions: (1) activate Rh4, (2) repress Rh3, and (3) repress a signal from R7 to R8, leading to the default yR8 fate (Rh6 expression) (Figure 1B). In the absence of Ss, pR7 fate (Rh3 expression) and pR8 fate (Rh5 expression) are induced (Figure 1 A).

yR7-specific expression of Rh4 appears to be simply activated by Ss. In contrast, pR7-specific expression of Rh3 is regulated by complex interlocked feedforward loops of transcription factors (Johnston et al., 2011). The Defective Proventriculus (Dve) homeodomain protein, a repressor that directly binds the rh3 promoter, is a critical node in this motif (Figure 1A–B). The Orthodenticle (Otd) homeodomain protein activates Dve expression in all PRs, whereas the Spalt zinc finger transcription factors (Salm and Salr, referred to collectively as “Sal”) repress Dve in R7s (Figure 1A–B). In pR7s, Sal and Otd together activate Rh3 in the absence of Dve (Figure 1A). In yR7s, Ss re-activates Dve which represses Rh3 despite the presence of Otd and Sal (Figure 1B) (Johnston et al., 2011; Sood et al.,2012).

The regionalized co-expression of Rh3 in Rh4-expressing yR7s in the dorsal third of the retina is activated by the transcription factors of the Iroquois Complex (IroC) Figure 1C–E) (Mazzoni et al., 2008). Whereas Ss provides a stochastic input, IroC supplies a regionalized input into the regulation of Rh expression.

Here, we show that, as in other biological contexts, Spineless acts with the ubiquitously expressed PAS-bHLH protein, Tango (Tgo) (Emmons et al., 1999), to regulate Rh expression. We show that the proper stochastic and regional control of Rh expression requires five mechanistic features: (1) Ss levels are high in yR7s in the main part of the retina to ensure repression of Rh3 and activation of Rh4, (2) Ss levels are reduced in dorsal third yR7s to allow Rh3 expression, (3) IroC activates Rh3 in dorsal third yR7s, (4) low Ss levels (as found in dorsal third yR7s) are sufficient to activate Rh4 expression, and (5) the absence of Ss expression produces pR7 fate including expression of Rh3 and absence of Rh4. The sensitivity of rh3 to regional inputs is likely due to the presence of multiple IroC (activating) and Dve (repressing) binding sites in the rh3 promoter, whereas the robustness of Rh4 activation appears to be due to the presence of a single Ss (activating) binding site. Our data demonstrate how stochastic and regionalized regulatory inputs are integrated to determine ommatidial subtype specification throughout the retina.

Results

Tgo is required for stochastic Rh expression

In most biological contexts, the Tgo PAS-bHLH transcription factor is required as a heterodimeric partner for Ss function. Ss is expressed in specific cell types where it binds ubiquitously-expressed Tgo in the cytoplasm. The Ss/Tgo heterodimer then localizes to the nucleus to regulate target genes (Emmons et al., 1999; Ward et al., 1998). Although it has been suggested that Ss works independently of Tgo in some contexts, these analyses were conducted with available tgo alleles, which were all hypomorphs (Kim et al., 2006).

Because we could not detect staining in the eye using the available Tgo antibody (data not shown), we generated a tgo transcriptional reporter (tgo^prom>nuGFP) that drove GFP expression in all cells in the retina including all R7 cells, consistent with previous reports that Tgo is ubiquitously expressed (Figure 2o).

Figure 2. Tgo is required for Ss-mediated regulation of Rhs.

• A.
  Schematic of the tgo gene locus. The blue triangle indicates the P-element containing the hobo transposon used to generate the tgo^del25 and tgo^del6molecular null alleles.
• B.
  Schematic of the Tgo protein with mutant allele annotation.
• C.
  Rh3 and Rh4 are expressed in random subsets of R7s in wild type animals.
• D.
  Rh3 is expressed in all R7s in ss null mutants.
• E.
  Rh3 is expressed in all R7s in tgo^del6 null mutants.
• F.
  Rh3 is expressed in all R7s when tgo is knocked down by RNAi.
• G.
  Rh5 and Rh6 are expressed in random subsets of R8s in wild type animals.
• H.
  Rh5 frequency increases (Rh6 decreases) in ss null mutants.
• I.
  Rh5 frequency increases (Rh6 decreases) in tgo^del6 null mutants.

For J-N, GFP- indicates tgo^del6 mutant tissue; GFP+ indicates wild type tissue. Left, stain including GFP; Right, stain without GFP.

• J.
  Rh3 is expressed in all R7s in tgo^del6 clones. Rh3 and Rh4 are expressed in random subsets of R7s in wild type tissue.
• K.
  Ss nuclear expression is lost in tgo^del6 clones. Ss is expressed in random subsets of R7s in wild type tissue. White circles indicate Ss+ R7s; gray circles indicate Ss- R7s.
• L.
  Dve expression is lost only in R7s in tgo^del6 clones. Dve is expressed in random subsets of R7s in wild type tissue. Dve expression occurs in outer PRs in tgo^del6and wild type tissue. White circles indicate Dve+ R7s; gray circles indicate Dve-R7s.
• M.
  Ectopic expression of Ss induces Rh4 in R7s and outer PRs in wild type tissue. Rh3 is expressed in all R7s in tgo^del6 clones with ectopic expression of Ss.
• N.
  Strong nuclear expression of Ss is observed upon ectopic expression in all PRs in wild type tissue. Weak nuclear expression of Ss is observed upon ectopic expression in all PRs in tgo^del6 clones.
• O.
  tgo^prom>GFP is expressed in all PRs of the retina including all R7s, R8s, and outer PRs. See also Figure S1.

To clearly ascertain the role of Tgo in stochastic Rh regulation, we generated two tgo null mutant alleles, tgo^del6 and tgo^del25, using the hobo transposable element system (see Materials and Methods). tgo^del6 removes the bHLH, PAS, and PAC domains required for dimerization and DNA binding, whereas tgo^del25 removes the entire tgo locus and part of the 3’ UTR region of the neighboring cg11986 gene (Figure 2A–B). Both tgo^del6 and tgo^del25 null mutant retinas displayed expression of Rh3 and loss of Rh4 in all R7s, similar to ss mutants (Figure 2D–E, Figure S1A). Retinas in which tgo was knocked down using RNAi as well as tgo null mutant clones displayed a similar phenotype (Figure 2F, 2J, S1B). tgo null mutants faithfully phenocopied ss null mutants for all features of Rh regulation including de-repression of the signal to R8s causing a dramatic increase in the frequency of Rh5-expressing R8s (Figure 2H–I), loss of Dve expression in yR7s (Figure 2L), and no effect on general cell fate markers (Figure S1C–G)

Consistent with the requirement of Ss/Tgo dimerization for efficient localization to the nucleus, Ss was not detectable in nuclei in tgo null mutant clones (Figure 2K). Further, although we could detect nuclear Ss in all photoreceptors when provided ectopically by a strong heterologous promoter (IGMR>Gal4), this localization was dramatically weaker in tgo mutant tissue compared to neighboring wild type tissue (Figure 2N).

Tgo is cell-autonomously required for regulation of Rh expression by Ss since ectopically-expressed Ss induced expression of Rh4 in all PRs and repression of Rh3 in all R7s in wild type clones, but failed to do so in tgo null mutant clones, leading to Rh3 expression and loss of Rh4 in all R7s (Figure 2M).

Ss is also required for the elaboration of dendrites in ‘dendrite arborization’ (da) sensory neurons. Although the hypomorphic tgo⁵ allele exhibited no dendritic arborization defects (Kim et al., 2006), tgo null mutant clones displayed decreases in the number of dendritic termini similar to ss null mutants, suggesting that Tgo is required for all Ss functions (Figure S1H–L).

Stochastic binary On/Off regulation and regional modulation of levels determines Ss expression

Rh3 and Rh4 are expressed in mutually exclusive R7 subtypes in the main part of the retina. However, in the dorsal third region, IroC activates Rh3 co-expression in Rh4-expressing yR7s. We hypothesized that Ss could also play a role in this co-expression phenomenon and thus, we assessed the levels of Ss protein to determine regional differences across the retina. We defined four regions of the retina based on ommatidium position relative to the equator: dorsal third (DT), dorsal equatorial (DE), ventral equatorial (VE), and ventral third (VT). Ss was expressed in an On/Off manner across the retina as indicated by the bimodal distributions for each region (Figure 3A–H). Interestingly, Ss levels were significantly lower for cells in the On state in the DT than in other regions of the eye (Figure 3A–B). Ss levels in IroC mutant clones were comparable to neighboring wild type tissue (Figure 3I–J) and ectopic expression of IroC in all R7s did not significantly reduce Ss levels (data not shown), showing that this reduction of Ss levels in the DT is IroC-independent.

Figure 3. Ss levels are regionally modulated.

For A-H, DT=dorsal third, DE=dorsal equatorial, VE=ventral equatorial, VT=ventral third. For B, D, F, and H, the bimodal distribution indicates the On/Off nature of Ss expression. The shift for the DT R7s (Figure 3B) shows the reduced levels of Ss expression in the On State in this region

1. Ss expression is lower in DT R7s.
2. Quantification of Ss expression in DT R7s.
3. Ss expression in DE R7s.
4. Quantification of Ss expression in DE R7s.
5. Ss expression in VE R7s.
6. Quantification of Ss expression in VE R7s.
7. Ss expression in VT R7s.
8. Quantification of Ss expression in VT R7s.
9. Ss levels are similar in IroC mutant and wild type clones. GFP- indicates IroC mutant tissue; GFP+ indicates wild type tissue. Top, Ss and GFP; Bottom, Ss alone. White circles indicate high Ss expression in yR7s; gray circles indicate no Ss expression in pR7s.
10. Quantification of Ss levels in IroC mutant and wild type clones. IroC mutant yR7s (gray) express Ss at similar levels to wild type yR7s (red).

Rh3 and Rh4 are differentially responsive to Ss/Tgo activity levels

We hypothesized that the reduced levels of Ss in the DT lower repression and allow IroC-mediated activation of Rh3 in yR7s in this region. rh4>ss generates a positive feedback loop to increase levels of Ss specifically in yR7s. These increased levels of Ss caused repression of Rh3 in yR7s in the DT (Figure 4A, 4G), showing that Ss must be maintained at low levels to allow for Rh3 expression.

Figure 4. Rh3 and Rh4 are differentially responsive to Ss/Tgo activity levels.

For A-G, DT=dorsal third, DE=dorsal equatorial, VE=ventral equatorial, VT=ventral third. For A-F, top, Rh3 and Rh4; middle, Rh3 alone; bottom, Rh4 alone. Green lines mark the regional boundary of Rh3 expression in yR7s. Dotted yellow lines mark the regional boundary of the normal frequency of Rh4 expression in yR7s.

1. Ectopic expression of Ss (rh4>ss) represses Rh3 in dorsal third yR7s. Rh3 and Rh4 are expressed in exclusive subsets of R7s in all regions.
2. In wild type animals, Rh3 and Rh4 are expressed in exclusive subsets of R7s in the DE, VE, and VT regions. The DT region is composed of R7s that express Rh3 alone or co-express Rh3 with Rh4.
3. In tgo⁶ mutants, Rh3 expression in yR7s expands to the DE region. Rh3 and Rh4 are expressed in exclusive subsets of R7s in the VE and VT regions. The DT and DE regions are composed of R7s that express Rh3 alone or co-express Rh3 with Rh4.
4. In tgo⁵ mutants, Rh3 expression in yR7s expands to the entire retina. These retinas are composed of R7s that express Rh3 alone or co-express Rh3 with Rh4.
5. In ss¹¹⁶⁵ mutants, Rh3 expression in yR7s expands to the entire retina. These retinas are composed of R7s that express Rh3 alone or co-express Rh3 with Rh4. The frequency of Rh4 expression is slightly reduced in the DT.
6. In tgo^del6 mutants, Rh4 is lost and Rh3 is expressed in all R7s throughout the retina.
7. Quantification of the series of ss and tgo alleles. Data is presented in order of decreasing Ss/Tgo activity (i.e. increasing phenotypic severity). The six main phenotypic classes are separated by dashed lines.

Therefore, Rh3 and Rh4 appear to be differentially responsive to modulation of Ss levels in regions of the retina: Whereas Rh3 repression is sensitive to the reduction of Ss levels in the DT, Rh4 activation is robust. We characterized these differences further by evaluating a series of ss and tgo mutant alleles that cause premature termination and truncation of activation domains, leading to reduction in activity levels (Figure 2B, Figure 5A). To determine the molecular lesions, we sequenced the ss^116.4, tgo⁶ and other ss alleles (Figure 2B, Figure 5A, Materials and Methods). Including the ectopic expression of Ss experiment, we observed six phenotypic classes:

1. Increased levels in yR7s (rh4>ss): Rh3 and Rh4 were expressed in stochastically distributed, complementary subsets of R7s throughout the retina, including the DT where Rh3 became excluded from yR7s (Figure 4A, 4G)

2. Wild type (tgo³): Rh3 and Rh4 were expressed in stochastically distributed, mutually exclusive subsets of R7s in the majority of the retina (i.e. DE, VE, and VT). Rh3 was expressed in Rh4-expressing yR7s in the DT (Figure 4B, 4G).

3. Weak loss-of-function (tgo⁶, tgo¹): De-repression of Rh3 occurred readily in yR7s of the DT and DE regions and sporadically in the VT and VE regions. Rh4 remained expressed in stochastically distributed yR7s (Figure 4C, 4G).

4. Medium loss-of-function (tgo⁵): De-repression of Rh3 occurred in nearly all R7s. Rh4 was still expressed in stochastically distributed yR7s (Figure 4D, 4G).

5. Strong loss-of-function (ss^d116.4/def, ss^d116.5/def): De-repression of Rh3 occurred in nearly all R7s. Rh4 was still expressed in stochastically distributed yR7s, but the frequency was subtly reduced in the DT region (Figure 4E, 4G).

6. Null (ss^d115.7, tgo^del6, tgo^del25): Rh3 was expressed and Rh4 was lost in all R7s (Figure 4F–G)

Figure 5. Ss/Tgo with reduced activity can activate Rh4 but not repress Rh3.

• A.
  Schematic of the Ss protein. The premature stop causing protein truncation in the ss^d116.4 allele is annotated.
• B.
  Table summarizing Ss protein domain analysis. −> Rh4 = Rh4 activation; −| Rh3 = Rh3 repression; Tgo nuclear local = Tgo localization. (−) indicates no effect. (+) indicates an effect. For the Tgo nuclear localization assay, (++) indicates strong nuclear localization whereas (+) indicates weak nuclear localization.

For C-H, examples of each type of phenotype are shown. For Rh3/Rh4 expression and Tgo nuclear localization, image data for all constructs are shown in Figures S2K–BB. For C-E, examples of effects on Rh3/Rh4 expression. Left, Rh3 and Rh4 expression; middle, Rh3 alone; right, Rh4 alone. Expression was assessed in the main part of the retina (excluding the dorsal third).

• C.
  Activation of Rh4 with repression of Rh3.
• D.
  Activation of Rh4 without repression of Rh3. Yellow circles indicate examples of R7s that co-express Rh3 and Rh4.
• E.
  No effect
• F.
  For F-H, examples of effects on Tgo nuclear localization.
• F.
  Strong nuclear localization of Tgo.
• G.
  Weak nuclear localization of Tgo.
• H.
  No effect.

See also Figure S2.

These data show that Rh3 and Rh4 respond differently to Ss/Tgo activity levels. Lowering Ss/Tgo activity allowed for de-repression of Rh3 without affecting activation of Rh4 in yR7s. Rh3 was susceptible to an underlying dorsal/ventral gradient of regulation since the region in which Rh3 co-expresses with Rh4 in yR7s expanded ventrally as Ss/Tgo activity decreased. Whereas Rh3 was very sensitive to Ss/Tgo activity levels, Rh4 was robust, with only subtle changes in the frequency of Rh4 expression observed in the DT (where Ss levels are reduced) in the strongest loss-of-function alleles. Our data suggest that Ss/Tgo activity occurs at specific levels to induce repression of Rh3 in yR7s throughout the majority of the retina while allowing co-expression with Rh4 in yR7s of the DT region.

Rh regulation is sensitive to the activation capacity of Ss

The ss/tgo allelic series suggested that the C-terminal activation domains are important for the regulation of Rh expression in R7s. For the tgo alleles, the degree of activation domain truncation correlated with the loss of activation capacity and phenotypic severity (Figure 2B, Figure 4C–D, Figure 4F–G) (Sonnenfeld et al., 2005).

To further characterize the differential response of Rh3 and Rh4 to Ss/Tgo activity levels, we used gain of function assays with Ss proteins with deletions of functional domains. The Ss bHLH domain binds DNA sequences in target genes upon dimerization with Tgo. The PAS domains mediate Ss/Tgo dimerization whereas the PAC motif contributes to folding of the PAS domain. The C-terminal region functions as an activation domain (Figure 5A) (Crews, 1998; Crews and Fan, 1999; Ponting and Aravind, 1997; Zhulin et al., 1997).

Considering our allelic series analysis, we predicted that Ss protein whose C-terminal activation domain had been truncated should activate Rh4 but not repress Rh3. We generated wild type and modified UAS>Ss (Ss^modified) constructs that lacked one or more of the distinct functional domains (Ss^ΔAS/PAC Ss^ΔbHLH, Ss^ΔPAS-N, Ss^ΔPAS-C Ss^ΔPAC) or had truncations of the C-terminal activation region (Ss^ΔC1, Ss^ΔC2, Ss^ΔC3) (Figure 5A–B). Upon ectopic Ss^modified expression (panR7>Gal4; UAS>Ss^modified), we observed three classes of phenotypes that corroborated our ss/tgo allelic series:

1. Ss^{wild type}, Ss^{ΔPAS -C}, Ss^ΔPAC, and Ss^ΔC1 induced expression of Rh4 and repression of Rh3 in all R7s (Figure 5B–C, Figure S2J–K, S20-Q), suggesting that these constructs were fully functional. It was surprising that the second PAS domain and the PAC domain were not required to regulate Rh expression since they have important roles for PAS-bHLH protein function.

2. Ss^ΔC2 and Ss^ΔC3 induced Rh4 but did not repress Rh3 in pR7s (Figure 5B, 5D, Figure S2J, S2R-S), consistent with the phenotype observed for ss and tgo mutants with truncated C-terminal domains.

3. Ss^ΔbHLH, Ss^ΔPAS-^N, and Ss^ΔPAS/PAC caused no change in Rh expression (Figure 5B, 5E, Figure S2J, S2L-N), consistent with critical roles for the bHLH and N-terminal PAS domains. We confirmed that these non-functional transgenes were indeed expressed using Ss antibody staining (Figure S2E–I).

Truncation of the C-terminal region reduces the transcriptional activity of the Ss protein. However, these changes in Ss could also impair heterodimerization with Tgo, prevent nuclear localization, or destabilize the protein in a non-specific way. We therefore tested the Ss^modified proteins for their capacity to localize Tgo to the nucleus. Ectopic expression of Ss driven by the engrailed promoter (en>Gal4) caused Tgo nuclear localization in the ectodermal en stripes in the fly embryo (Figure 5B, 5F) (Emmons et al., 1999). Another PAS-bHLH partner of Tgo, Trachealess, localizes Tgo to the nucleus in tracheal tubules and salivary primordia, serving as an internal control for Tgo antibody staining (Figure S2T–BB) (Ward et al., 1998).

Ss^ΔC1, Ss^ΔC2, Ss^ΔC3 induced Tgo nuclear localization as well as Ss^{wild type}, suggesting that truncation of the C-terminal activation region does not impair heterodimerization, localization, or stability (Figure 5B, 5F, Figure S2T, S2Z–BB). Ss^ΔPAC was also sufficient to induce Tgo nuclear localization, consistent with the dispensability of the PAC domain in the retina (Figure 5B, 5F, Figure S2Y). Interestingly, Ss^{ΔPAS -C} was only able to induce lower levels of Tgo localization (Figure 5B, 5G, Figure S2X), suggesting that the C-terminal PAS domain increases the efficiency for dimerization and nuclear localization but is not absolutely required. As expected, Ss^ΔbHLH, Ss^ΔPAS-N, and Ss^ΔPAS/PAC were not able to induce Tgo nuclear localization (Figure 5B, 5H, Figure S2U–W).

This structure/function analysis of Ss is consistent with the ss/tgo allelic series. Ss and Tgo proteins with truncations of the C-terminal activation domains causing reduced activity levels were able to induce Rh4 expression, but not Rh3 repression.

Stochastic and regional expression of Rh3 is controlled by repressing and activating inputs

Ss/Tgo does not regulate rh3 directly, but rather activates Dve to repress rh3 in yR7s (Figure 1A–B). However, Dve is expressed at levels that allow IroC to overcome repression and activate Rh3 in yR7s of the DT region (Johnston et al., 2011). We wondered whether Dve expression levels were affected in hypomorphic mutant situations that displayed Rh3 de-repression. In tgo⁵ mutant clones, Dve expression was decreased but not lost (Figure 6A–B). Thus, the dramatic changes in Rh3 expression observed upon modulation of Ss/Tgo activity appear to be mediated by changes in Dve levels.

Figure 6. Regional rh3 expression is sensitive to activating and repressing inputs.

• A.
  Dve levels in yR7s are decreased in tgo⁵ mutant clones. GFP- indicates tgo⁵mutant tissue; GFP+ indicates wild type tissue. Left, Dve and GFP; Right, Dve alone. White circles indicate high Dve expression in wild type yR7s; light gray circles indicate low Dve expression in tgo⁵ mutant yR7s; dark gray circles indicate no Dve expression in wild type and tgo⁵ mutant pR7s
• B.
  Quantification of Dve levels in tgo⁵ mutant clones and wild type tissue. Wild type R7s (green) express Dve at higher levels than tgo⁵ mutant R7s (black).
• C.
  Schematic of rh3 promoter that recapitulates Rh3 protein expression. Orange = K₅₀, Purple = IroC site, Green = RCSI, Yellow = TATA box. The RCSI is a conserved element found in all rh promoters that is required for expression.

For D, gray circles indicate DT yR7s that have lost expression of rh3^prom>GFP. For E-H, white circles indicate the ventral-most yR7s that express rh3^prom>GFP. For D-H, when IroC sites are mutated, expression of rh3^prom>GFP is lost in DT yR7s, similar to the loss of expression observed when Ss levels are ectopically high. As K₅₀sites are mutated, de-repression of rh3^prom>GFP expands ventrally similar to derepression of Rh3 protein in ss/tgo hypomorphic alleles. For all, rh3^prom>GFP is expressed in pR7s in all regions of the retina. DT=dorsal third, DE=dorsal equatorial, VE=ventral equatorial, VT=ventral third. Dashed yellow lines mark regions where rh3^prom>GFP is expressed in yR7s.

• D.
  rh3^{prom IroC mut134}>GFP is expressed in pR7s only. Expression in DT yR7s is lost.
• E.
  rh3^{prom wild type}>GFP is expressed in yR7s in the DT.
• F.
  rh3^{prom K50 mut1}>GFP is expressed in yR7s in the DT and DE.
• G.
  rh3^{prom K50 mut2}>GFP is expressed in yR7s in the DT, DE, and VE.
• H.
  rh3^{prom K50 mut12}>GFP is expressed in yR7s throughout the retina (including DT, DE, VE, and VT).

rh3 regulation is controlled by repressing (Dve) and activating (Otd, Sal, IroC) inputs. We next tested whether the sensitivity of rh3 regulation extends to its promoter. The rh3 promoter contains three canonical binding sites (TAATCC) for the K₅₀ homeodomain transcription factors, Dve and Otd (Figure 6C) (Johnston et al., 2011; Tahayato et al., 2003). In yR7s, these sites appear to mediate repression by Dve. A 194 bp rh3 promoter (rh3prom>GFP) induced expression in only pR7s in the main part of the retina and in both pR7s and yR7s in the DT region, similar to the Rh3 protein (Figure 6E). Mutation (K₅₀ mut1) of the distal K₅₀ site (K₅₀–1) caused de-repression in yR7s of the DE region (Figure 6F) whereas mutation (K₅₀ mut2) of the proximal site (K₅₀–2) led to de-repression in yR7s of the DE and VE regions (Figure 6G). Mutation of both sites (K₅₀ mut12) caused de-repression in yR7s throughout the retina (Figure 6H).

We next tested the roles of the four putative IroC binding sites in the rh3 promoter (Figure 6C)(Bilioni et al., 2005). Mutation of 3 of 4 sites (IroC mut134) or all 4 sites caused a complete loss of expression in DT yR7s (Figure 6D and data not shown). Both sets of mutations caused some loss of expression in pR7s, suggesting that these sites may also play a limited role in basal activation of rh3. These data suggest that IroC directly binds the rh3 promoter to upregulate expression in the DT.

Thus, Dve and IroC binding sites are critical for the sensitivity of rh3 to this regulatory network, dictating regional activation or repression outcomes in yR7s. The dorsal/ventral effects of these promoter mutations closely mirror the ss/tgo allelic series (Figure 4A–F, Figure 6D–H).

In summary, removing IroC activity or IroC binding sites causes loss of Rh3 expression in DT yR7s whereas ectopic expression of IroC at high levels induces Rh3 in the main region (Figure 6D)(Mazzoni et al., 2008). Removing Ss/Tgo or Dve activity, or Dve binding sites causes de-repression in the main region whereas increasing levels of Ss/Tgo or Dve causes repression of Rh3 in DT yR7s (Figure 4A–G, Figure 6A–B, Figure 6F–H)(Johnston et al., 2011). We conclude that Rh3 expression in DT yR7s is controlled by both activation by IroC and a release of repression by Dve (and indirectly Ss) on the rh3 promoter.

Rh4 activation is robust to modulation of Ss/Tgo activity

The Ss/Tgo heterodimer is a transcriptional activator that could act directly on the rh4 promoter to induce expression. Ss/Tgo binds Xenobiotic Response Elements (XREs; core sequence: CACGC) to activate target genes (Emmons et al., 1999). A 455 bp rh4 promoter (rh4prom>GFP) induced expression that recapitulates endogenous Rh4 expression in yR7s (Figure 7A–B) (Fortini and Rubin, 1990). We identified one core XRE site that is conserved in all 12 sequenced Drosophila species examined (Figure 5D). The XRE is part of a larger element previously defined as RUS4A (TTTGCGGGCACGCAA) that is required for rh4 reporter expression (Fortini and Rubin, 1990). A single point mutation in the XRE (to CAAGC) led to nearly complete abrogation of reporter expression (<2% R7s expressed GFP)(Figure 7B–C). Thus, the XRE sequence is required for rh4 expression, strongly suggesting that Ss/Tgo directly binds the rh4 promoter to induce expression.

Figure 7. Robust Rh4 activation requires a canonical Ss binding site.

For A-D, colors indicate known cis-regulatory regions. Blue = RUS4A, Red subset of RUS4A = XRE core site (Ss/Tgo binding site), Green = RCSI (inverted in D. vir, D. moj, D. gri), Yellow = TATA box.

1. Schematic of rh4 promoter that recapitulates Rh4 protein expression. Sequence shows known critical cis-regulatory elements.
2. rh4^prom>GFP with wild type XRE recapitulates Rh4 protein expression.
3. Expression is lost with a point mutation in the XRE.
4. Sequence alignment of the rh4 promoter for 12 Drosophila species highlighting the known cis-regulatory elements. The XRE core sequence is perfectly conserved in all 12 species. Sequence alignment was from the UCSC Genome Browser (http://genome.ucsc.edu/) (Fujita et al., 2011; Kent, 2002; Kent et al., 2002).
5. rh4^prom>GFP is expressed at similar levels in the DT.
6. rh4^prom>GFP is expressed at similar levels in the VT.
7. rh4^prom>GFP is expressed at similar levels in tgo⁵ and wild type yR7s. Panel 1: GFP, Rh4, and Rh3; Panel 2: GFP alone; Panel 3: Rh4 alone and Panel 4: Rh3 alone. The white circle indicates a tgo⁵ mutant yR7 that expresses rh4^prom>GFP with both Rh3 and Rh4. The solid gray circle indicates a wild type yR7 that expresses rh4^prom>GFP with Rh4 alone. The dotted gray circle indicates a pR7 that expresses Rh3 alone.
8. Quantification of rh4^prom>GFP levels in tgo⁵ and wild type yR7s. tgo⁵ mutant yR7s (magenta) express rh4^prom>GFP at similar levels to wild type yR7s (red).

Since Ss levels were lower in the DT, we wondered if Rh4 transcription levels were also reduced. Rh4 protein appears to be lower in the DT, although this could be due to Rh4 competing with Rh3 protein for rhabdomeric space. We assessed rh4 transcription with the rh4prom>GFP reporter gene, which showed no difference in levels between the DT (where Ss levels are low) and VT (where Ss levels are high)(Figure 7E–F), suggesting that reduction in Ss levels in the DT does not affect Rh4 expression. We also evaluated rh4prom>GFP in mutant clones that have reduced Ss/Tgo activity. rh4prom>GFP was expressed at similar levels in tgo⁵ and neighboring wild type clones (Figure 7G–H), suggesting that activation of rh4 is robust to perturbations of Ss/Tgo activity. Since Ss is expressed at lower levels in the DT, the responsiveness of the rh4 promoter ensures that Rh4 is still activated there. In contrast, rh3 expression is sensitive to levels of Ss/Tgo activity to induce expression in DT yR7s and repression in main region yR7s.

Discussion

The complex expression pattern of Rhs in R7s requires the integration of stochastic and regional regulatory information. In the main part of the retina, high levels of Ss in yR7s ensure repression of Rh3 and activation of Rh4. In the DT, reduced Ss levels allow IroC-mediated activation of Rh3, yet are sufficient to activate Rh4.

IroC activates Rh3 in the DT where Ss levels are lower (Figure 8H). In IroC mutants, even the low levels of Ss in the DT are sufficient to induce repression of Rh3 (indirectly through Dve) (Figure 8G). Reciprocally, high levels of IroC are sufficient to induce Rh3 in the main part of the retina despite the presence of normal levels of Ss (Figure 8I)

Figure 8. Model of region-specific regulation of Rh3 and Rh4 in yR7s by Ss and IroC.

Green indicates Ss/Tgo activity levels; magenta indicates IroC activity levels; blue indicate Rh3 expression frequency in yR7s; and Red indicates Rh4 expression frequency in yR7s.

As Ss/Tgo activity decreases (C–F) or IroC activity increases (I), Rh3 expression expands in yR7s. As Ss/Tgo activity increases (A) or IroC activity decreases (G), Rh3 expression is lost in yR7s. Despite changes in Ss/Tgo or IroC activity, the frequency of Rh4 expression is robust (A-D, G-l) with only subtle changes observed in the DT where Ss levels are low in strong ss lof (E).

lof = loss-of-function.

DT=dorsal third, DE=dorsal equatorial, VE=ventral equatorial, VT=ventral third.

The regional regulation of rh3 also requires reduction of Ss (and Dve) levels in the DT. In wild type animals, Ss levels are high in yR7s in the main region to induce repression of Rh3 and low in the DT to allow Rh3 expression (Figure 8B). Increasing the levels of Ss in the DT causes Rh3 repression in yR7s (Figure 8A). As the activity of Ss/Tgo is lowered in mutant conditions, Rh3 expression expands ventrally in the DE (Figure 8C) and then throughout the retina (Figure 8D–F). This intermediate expansion into the DE may be explained by the dynamic nature of IroC expression. IroC is initially expressed in all cells of the dorsal half (DT and DE) of the retina and then becomes restricted to the DT (Mazzoni et al., 2008; Sato and Tomlinson, 2007; Tomlinson, 2003). Perhaps, residual levels of IroC and/or chromatin changes induced at the rh3 promoter increase the activation capacity of rh3 in the DE (Figure 8B–C). At low levels of Ss activity or with complete ablation, the general activators Sal and Otd induce expression of Rh3 in all R7s, including those in the ventral half (VE and VT)(Figure 8D–F). These observations support the requirement for the modulation of Ss levels to ensure proper regional Rh3 regulation. The presence of multiple K₅₀ (repressing via Dve and activating via Otd) and IroC (activating) binding sites in the rh3 promoter is consistent with the nature of this regulatory mechanism.

In contrast, the control of Rh4 appears to be much simpler: Ss/Tgo, even at low levels, induces Rh4 expression (Figure 8B), likely by directly binding the lone canonical XRE (Ss/Tgo binding site) in the rh4 promoter. Perturbations of Ss/Tgo that yield derepression of Rh3, do not affect Rh4 expression (Figure 8C–D). A subtle decrease in the frequency of Rh4 expression occurs in the DT where Ss levels are reduced only when Ss/Tgo activity is strongly impaired, likely because Ss/Tgo activity levels are near the threshold for activation (Figure 8E). Therefore, rh4 is highly responsive to Ss ensuring that it is expressed in yR7s in the DT where Ss levels are low.

Ss expression is controlled by two main inputs: stochastic On/Off regulation and regional modulation of levels. The random, binary input determines pR7 (Ss Off) vs. yR7 (Ss On) fate whereas the regional input determines main region (Ss high) vs. DT (Ss low) yR7 fate. Though IroC is considered the critical factor determining dorsal identity in the retina, it does not appear to control regional modulation of Ss levels. Rather, another mechanism must work in parallel with IroC to control aspects of dorsal identity.

Despite dramatic changes in the activity of Ss and Tgo, Rh4 remains expressed in ∼65% of R7s. If the mechanism controlling stochastic subtype specification was dependent on Ss levels (e.g. via a feedback mechanism), a decrease in Ss/Tgo activity levels should cause a decrease in the frequency of Rh4-expressing yR7s with a concomitant increase in the number of Rh3-expressing pR7s. Here, we have shown that this is not the case: Ss levels do not play a role in determining the frequency of stochastic expression but rather are modulated to allow Rh3 expression in Rh4-expressing yR7s in the DT. Thus, the stochastic mechanism controlling Ss expression requires regulation of the ss promoter independent of feedback.

Proper Rhodopsin expression requires tight regulation of the levels of stochastically-expressed Ss. If Ss levels were highly variable, we would expect to see de-repression of Rh3 in other R7s throughout the retina. Instead, we only observe expression of Rh3 in DT yR7s consistent with our findings that Ss levels are specifically lower there.

Our ongoing promoter dissection reveals that, not surprisingly, ss is controlled by a complex cis-regulatory logic (Johnston and Desplan, submitted). It will be interesting to see how the ss gene integrates these two dramatically different types of inputs to produce its complex expression pattern.

Experimental Procedures

Drosophila strains and crosses

Flies were raised on standard corn-meal-molasses-agar medium and grown at 25°C. Below, we list all shortened genotypes, complete genotypes, figures in which they are examined, and original source for each reagent:

Short Genotype	Complete Genotype	Figures	Source
wild type	yw; +; + or yw; sp/CyO; + or yw; sp/CyO; TM2/TM6b	1D–F, 2C, 2G, 3A–H
ss null mutant	ey>flp;+; FRT82b SSd115.7 / FRT82b GMR>hid, cL	2D, 2H, 4G	ey>flp (Newsome et al., 2000) FRT82b GMR>hid (Stowers and Schwarz, 1999) FRT82b ssd115.7 (Duncan et al., 1998)
tgode16 null mutant	ey>flp;+; FRT82b tgode16/ FRT82b GMR>hid, cL	2A–B, 2E, 21, 4F–G
tgode125 null mutant	ey>flp;+; FRT82b tgode125 / FRT82b GMR>hid, cL	2A–B, S1A, 4G
tgo RNAi	UAS>Dicer2; ey>Gal4, GMR>gal4/+; UAS>tgo RNAi/+	2F	UAS>Dicer2, ey>Gal4, GMR>gal4, and UAS>tgo RNAi [VDRC stock center]
RNAi control	UAS>Dicer2; ey>Gal4, GMR>gal4/+; +	S1B
tgode16 mutant clones	ey>flp;+; FRT82b tgode16 / FRT82b ubi>GFP	2J–L, S1C–G	FRT82b ubi>GFP [Bloomington stock center]
GMR>ss; tgode16 mutant clones	ey>flp; IGMR>Gal4, UAS>ss/ +; FRT82b tgode16 / FRT82b ubi>GFP	2M–N	IGMR>Gal4 (Wernet et al., 2003) UAS>ss (Duncan et al., 1998)
tgoprom>nuGFP	yw; +;tgoprom>GFP /+	20
wild type single cell clones	hs>flp, elav>Gal4, UAS>mcd8GFP /+; +; FRT82b tubpGal80 / FRT82b +	S1H, S1L
ss single cell mutant clones	hs>flp, elav>Gal4, UAS>mcd8GFP /+; +; FRT82b tubpGal80 / FRT82b ssd115.7	S1I.S1L
tgode16 single mutant clones	hs>flp, elav>Gal4, UAS>mcd8GFP /+; +; FRT82b tubpGal80 /FRT82b tgodel6	S1J, S1L
tgode125 mutant clones	hs>flp, elav>Gal4, UAS>mcd8GFP /+; +; FRT82b tubpGal80 / FRT82b tgodel25	S1K, S1L
IroC mutant clones	ey>flp;+; FRT80 Df(3L)iroDFM3 / FRT80ubi>GFP	3I–J
rh4>ss	yw; rh4>Gal4 / UAS>ssmodified; +/+	4A, 4G
wild type whole eye clone	ey>flp;+; FRT82b + / FRT82b GMR>hid, cL	4B, 4G
tgo3 mutant	ey>flp;+; FRT82b tgo3 / FRT82b GMR>hid, cL	4G	FRT82b tgo3 (Sonnenfeld et al., 1997; Sonnenfeld et al., 2005)
tgo6 mutant	ey>flp;+; FRT82b tgo6 / FRT82b GMR>hid, cL	4C, 4G	FRT82b tgo6 (Emmons et al., 1999)
tgo1 mutant	ey>flp;+; FRT82b tgo1 / FRT82b GMR>hid, cL	4G	FRT82b tgo1 (Sonnenfeld et al., 1997)
tgo5 mutant	ey>flp;+; FRT82b tgo5 / FRT82b GMR>hid, cL	4D, 4G	FRT82b tgo5 (Emmons et al., 1999; Tajiri et al., 2007)
ss116.4 /def mutant	yw; +;ss116.4 / Df(3R)Exel7330	4G	ss116.4 (Duncan et al., 1998) Df(3R)Exel7330 [Bloomington stock center]
ss116.5 /def mutant	yw; +;ss116.5 / Df(3R)Exel7330	4E, 4G	ss116.5 (Duncan et al., 1998)
panR7>ssmodified; h4prom>GFP, subsequently panR7>ssmodified	yw; panR7>Gal 4/ UAS>ssmodified; rh4prom>GFP / +	5B–E, S2A–D, S2J–S
IGMR>ssmodified; rh4prom>GFP, subsequently IGMR>ssmodified	yw; IGMR>Gal4 / UAS>ssmodified; rh4prom>GFP / +	S2A–I
en>ssmodified	yw; en>Gal4 / UAS>ssmodified; +	5B,5F–H, S2T–BB	en>Gal4 [Bloomington stock center]
tgo5 mutant clones	ey>flp;+; FRT82b tgo5 / FRT82b ubi>GFP	6A–B
rh3prom IroCmut134>GFP	prom IroCmut134	6D
rh3prom wild type>GFP	yw; +; rh3prom wild type>GFP	6E
rh3prom K50 mut1>GFP	yw; +; rh3prom K50 mut1>GFP	6F
rh3prom K50 mut2>GFP	yw; +; rh3prom K50 mut2>GFP	6G
rh3prom K50 mut12>GFP	yw; +; rh3prom K50 mut12>GFP	6H
rh4prom>GFP	yw; +; rh4prom>GFP / +	7A–B, 7E–F
rh4prom>GFP (Point mutation)	yw; +; rh4prom>GFP / +	7C
tgo5 mutant clones; rh4prom>GFP	ey>flf; rh4prom>GFP / +; FRT82b tgo5 / FRT82b +	7G–H

Note: For UAS>Ss^modified genotypes, these P-element transgenes were on either chromosome two or chromosome three. We list only chromosome two for simplicity.

Antibodies

Antibodies and dilutions used were as follows: mouse anti-Rh3 (1:10) (gift from S. Britt, University of Colorado), rabbit anti-Rh4 (1:100)(gift from C. Zuker, Columbia University), mouse anti-Rh5 (1:200) (Chou et al., 1996), rabbit anti-Rh6 (1:2000) (Mikeladze-Dvali et al., 2005; Tahayato et al., 2003), guinea pig anti-Ss 2.21 (1:200)(gift from Y.N. Jan, University of California, San Franscisco)(Kim et al., 2006), rabbit anti-Dve (1:500) (Nakagoshi et al., 1998), mouse anti-prospero (1:10)(DSHB), ms anti-Tgo (1:1, concentrated 10X)(DSHB), rat anti-ElaV (1:50) (DSHB), rabbit anti-Sens (1:100)(Xie et al., 2007), rat anti-Sal (1:100)(Barrio et al., 1999), guinea pig anti-Otd (1:750)(Vandendries et al., 1996), sheep anti-GFP(1:500) and Alexa488 Phalloidin (1:80) (Invitrogen). All secondary antibodies were Alexa-conjugated (1:400) (Molecular Probes).

Antibody staining (pupal and adult retinas)

Adult or staged pupal retinas were dissected and fixed for 15 minutes with 4% formaldehyde at room temperature. Retinas were rinsed two times in PBX and washed in PBX for >2 hours. Retinas were incubated with primary antibodies diluted in PBX overnight at room temperature and then rinsed two times in PBX and washed in PBX for >4 hours. Retinas were incubated with secondary antibodies diluted in PBX overnight at room temperature and then rinsed two times in PBX and washed in PBX for >2 hours. Retinas were mounted in Slofade (adult retinas) or Vectashield (pupal retinas). Images were acquired using an SP5 Leica confocal microscope (Hsiao et al., 2012; Rister et al., 2013).

Antibody staining (embryos)

Embryos were collected, dechorionated, fixed and devitellinated. Embryos were washed three times and then stored in methanol. For staining, methanol was removed and embryos were rinsed and washed in PBX. Embryos were incubated with primary antibodies overnight at 4C and then rinsed two times in PBX and washed in PBX for >4 hours. Embryos were incubated with secondary antibodies diluted in PBX overnight at 4C and then rinsed two times in PBX and washed in PBX for >4 hours. Embryos were mounted in Aquapolymount (Polysciences, Inc.) and images were acquired using an SP5 Leica confocal microscope.

Initial screening of UAS>Ss^modified lines with water immersion microscopy

We used the panR7>Gal4 (expressed in all R7s) and IGMR>Gal4 (expressed in all PRs) drivers to induce UAS>Ss^modified expression in the eye and examined rh4^prom>GFP expression. Flies were adhered to a Petri dish using nail polish and immersed in water. The retina was observed for GFP expression using a compound fluorescence microscope (40x lens)(Pichaud and Desplan, 2001).We tested a minimum of two independent lines and found consistent results among lines (Figure S2A–D). We selected a single line for each UAS>Ss^modified transgene for further analysis.

Generating tgo null alleles

The P(wHy)tgo^DG08708 transposable element in tgo was used to generate null mutant alleles (Huet et al., 2002). yw; +; P(wHy)tgo^DG08708 /TM3 flies were crossed with yw; CyO, P(hsHT-2)/ ln(2LR)Gla, wg^Gla−1; +. Parents were flipped after one day of egg laying. Progeny were heat shocked 30 minutes each day for four consecutive days. yw; cyo,P(HSHT-2)/+; P(wHy)tgo^DG08708 /+ progeny were crossed with yw; +; + and their progeny were screened for w⁻y⁺. w⁻y⁺ deletion stocks were established.

tgodelprimer1 (5’-GAAGCTACGACCATGGGAGG-3’) and tgodelprimer2 (5’-ACGCAAAACACCGTATTGATTCGG-3’) were used to amplify the tgo locus and determine the size of the deletion. Amplicons were sequenced to precisely determine the deleted sequence.

Generating da neuron clones

For MARCM, FRT +, FRTss^d1157, FRT82b tgo^del6 or FRT82b tgo^del25 flies were mated to elav-GAL4, UAS-mCD8::GFP, hs-FLP; FRT82 tubP-GAL80 flies (Lee and Luo, 2001). Embryos were collected for a 2 hr period and aged for 3 hr at 25°C. Embryos were then heat-shocked at 39°C for 50 minutes, allo wed to recover for 30 minutes at 25°C, then heat-shocked again at 39°C for 45 min. A nimals were reared at 18°C until the wandering larval stage, when GFP-positive clones were imaged. Morphology was analyzed in larval fillet preparations (Ye et al., 2004) immunostained with 1:350 Alexa Fluor 488 rabbit anti-GFP (Invitrogen), mounted in 70% glycerol, and imaged on a Leica SP5 confocal microscope using a 40x/1.25 NA oil objective. The total number of terminal branches was quantified in projections of individual ddaC neurons from the second through fifth abdominal segment as previously described (Lee et al., 2003). In Figure S1L, all error bars are ± one standard deviation around the mean.

Transgenes

tgo^prom>GFP

The tgo promoter was PCR amplified from wild type flies (tgopromprimer1: 5’-CTGCAGCATGTGCATGTGCTACGACTG-3’ and tgopromprimer2: 5’-GGATCCGCTTGGAATGCGTAATTAGAAAACG-3’) and cloned into the P-GEMT Easy vector. The tgo promoter was subcloned into an attB vector containing nuclearGFP using Pstl and BamHI. Constructs were inserted into the J36 attP site (gift of K. Basler) on the third chromosome using phi-C31 mediated transgenesis (Bischof et al., 2007).

UAS>ss^modified

Fragments of ss were amplified from ss cDNA (Duncan et al., 1998), and subcloned into PCR-TOPOII vectors. Constructs were sequenced. ss fragments were subcloned into a modified pUAST vector (containing w⁺) using EcoRI, Bglll, and Xbal. Constructs were injected into yw flies and independent transgenic (w⁺) lines were established.

Below is the list of primers used to generate UAS>ss^modified constructs:

	Fragment 1	Fragment 2
Wild type	For:5’-GAATTCATGAGCCAGCTGGGCACCGTC-3’ Rev:5’-TCTAGAGCGGTGGCCGTGGTGCAGGTG-3’	N/A
PAS/PAC del	For:5’-GAATTCATGAGCCAGCTGGGCACCGTC-3’ Rev:5’-AGATCTCAGGCCGTGCTCGAAGGC-3’	For:5’-AGATCTGAGGGCCACGATCTGCTGGGC-3’ Rev:5’-TCTAGAGCGGTGGCCGTGGTGCAGGTG-3’
bHLH del	For:5’-GAATTCATGAAGGATAAGGAGGATAACGGAGTC-3’ Rev:5’-TCTAGAGCGGTGGCCGTGGTGCAGGTG-3’	N/A
PAS-N del	For:5’-GAATTCATGAGCCAGCTGGGCACCGTC-3’ Rev:5’-AGATCTCAGGCCGTGCTCGAAGGC-3’	For:5’-AGATCTCTGCCCGCCGACATGTCC-3’ Rev:5’-TCTAGAGCGGTGGCCGTGGTGCAGGTG-3’
PAS-C del	For:5’-GAATTCATGAGCCAGCTGGGCACCGTC-3’ Rev:5’-AGATCTCGGCCCGAAGGGCGTGCAGTAGGC-3’	For:5’-AGATCTACGGGTGCCTCTGGCATGATC-3’ Rev:5’-TCTAGAGCGGTGGCCGTGGTGCAGGTG-3’
PAC-del	For:5’-GAATTCATGAGCCAGCTGGGCACCGTC-3’ Rev:5’-AGATCTGCCAGAGGCACCCGTCTTCAAAAG-3’	For:5’-AGATCTGAGGGCCACGATCTGCTGGGC-3’ Rev:5’-TCTAGAGCGGTGGCCGTGGTGCAGGTG-3’
C-trun1	For:5’-GAATTCATGAGCCAGCTGGGCACCGTC-3’ Rev:5’-TCTAGACTCGTCCATCAGCTGGCGGTG-3’	N/A
C-trun2	For:5’-GAATTCATGAGCCAGCTGGGCACCGTC-3’ Rev:5’-TCTAGAAGCAGGGGATCCCACCTGGCC-3’	N/A
C-trun3	For:5’-GAATTCATGAGCCAGCTGGGCACCGTC-3’ Rev:5’-TCTAGACGAGGAGCAAGATATCACCTC-3’	N/A

rh3^prom#x0003E;GFP

The rh3 promoter was PCR amplified from yw flies and flanked by Bglll and Notl sites (rh3promprimer1: 5’-AGATCTGCACTAACCTTCAGATGAGC-3’ and rh3promprimer2: 5’-GCGGCCGCGTCTGCGGGCCAAGACGAAATCA-3’) and cloned into the pGEM-T Easy vector.

rh3^{prom lroC134}>GFP

IroC motifs 1 and 4 were mutated by amplifying the wild type promoter using modified oligos with mutations at the 5’ and 3’ end (JR3FWIroc1mut 5’-AGATCTGCACTAACCTTCAGATGAGCTGCACTTAGCC-3’ and JR3REVIroc4mut: 5’-GCGGCCGCGTCTGCGGGCCAAGACGAAATCTAGGC-3’). The QuikChange mutagenesis kit (Stratagene) was used to mutate the IroC motif 3 (JR3lroCmut3AFor: 5’-GGTAATCCCGCTGCGTAGATGCTAATCCAATTC-3’ and JR3lroCmut3ARev: 5’-GAATTGGATTAGCATCTACGCAGCGGGATTACC −3’) to generate the rh3^promlroC134>GFP construct.

rh3^{prom IroC1234}>GFP

The QuikChange mutagenesis kit was used to mutate the IroC motif 2 in rh3^{prom lroC134}>GFP (JR3lroCmut2HthAFor: 5’- CAGTGCCAGCGAAAAATCCAGCAAGGGATTAGG −3’ and JR3lroCmut2HthARev: 5’-CCTAATCCCTTGCTGGATTTTTCGCTGGCACTG −3’) to generate the rh3^promlroC1234>GFPconstruct.

rh3^{prom K5Omut1}>GFP, rh3^{prom K5Omut2}>GFP, rh3^{prom K5Omut12}>GFP

The QuikChange mutagenesis kit was used to mutate the K₅₀−1 (rh3mutprimer1For: 5’-GCCAGCGAAAATGTCAGCAAGGGGCGAGGCCAATCCCAAACGGGTAATC-3’and rh3mutprimer1Rev: 5’- GATTACCCGTTTGGGATTGGCCTCGCCCCTTGCTGACATTTTCGCTGGC −3’) and/or K₅₀−2 (rh3mutprimer2For: 5’- AGGCCAATCCCAAACGGGTCGCCCCGCTGCGACAATGCTA-3’ and rh3mutprimer2Rev: 5’- TAGCATTGTCGCAGCGGGGCGACCCGTTTGGGATTGGCCT-3’).

The rh3 promoter and the mutated rh3 promoters were subcloned into an attB vector containing GFP and miniwhite as selectable marker using Bglll and Notl. Constructs were inserted into the J36 attP site on the third chromosome using phi-C31 mediated transgenesis (Bischof et al., 2007).

rh4^prom>GFP and point mutation

The rh4 promoter was PCR amplified from transgenic flies (rh4promprimer1: 5’-CTTTGGAGTACGAAATGCGTC-3’ and rh4promprimer2: 5’-GTCCAGCTCGACCAGGATGGG-3’) and cloned into the P-GEMT Easy vector. The rh4 promoter was subcloned into pBS using BamHI and EcoRI. rh4mutprimerFor (5’- CAATTAGACTTTGTGGTTGCTTGCCCGCAAAGACGATTTTC-3’) and rh4mutprimerRev (5’- GAAAATCGTCTTTGCGGGCAAGCAACCACAAAGTCTAATTG-3’) were used to induce the point mutation. The rh4 promoter and the rh4 promoter (point mutation) were subcloned into an attB vector containing GFP using BamHI and EcoRI. Constructs were inserted into the J36 attP site on the third chromosome using phi-C31 mediated transgenesis (Bischof et al., 2007).

Quantification of expression

Using cell specific markers, antibody staining frequency was assessed for 5 or more retinas. >25 cells were scored per retina/region.

We used the LAS-AF software to quantified Ss levels in wild type animals (Figure 3A–H), Ss levels in yR7s (“On cells”) in wild type and IroC mutant clones (Figure 3I–J), Dve levels in tgo⁵ mutants (Figure 6A–B), GFP levels in wild type DT and VT (Figure 7E–F), and GFP levels in wild type and tgo⁵ mutant clones (Figure 7G–H). A 1.3 urn in diameter circular “region of interest” was manually placed at the center of each R7 to avoid signal from neighboring PRs. LAS-AF software assessed the Pixel intensity for each region of interest for each R7.

All error bars in figures are ± one standard deviation around the mean.

Identification of molecular lesion in ss and tgo alleles

ss^d116.4

We PCR amplified and sequenced the coding regions of the ss gene. We identified a deletion causing missense mutations and early termination (Figure 5A).

tgo⁶

We PCR amplified and sequenced the coding regions of the tgo gene. We identified a missense mutation causing early termination (Figure 2B).

Dp(3;2)P10

We used next-gen whole genome DNA sequencing to identify the breakpoint of this duplication in the ss locus (3R: 12,201,754).

T(1;3)ss^D1143

We used next-gen whole genome DNA sequencing to identify the breakpoints of this reciprocal translocation in the ss locus (3R: ∼12,237,800) and X chromosome (Xhet: ∼192,100).

SS^GSG2553exc8o

We generated imprecise P-element excision lines from the P(Switch2)GSG2553 P-element. We used next-gen whole genome DNA sequencing to determine that this allele contains an inversion (breakpoint in ss at 3R: ∼12,230,000) and a deletion (3R: ∼12,230,000–12,235,000).

SS^RS6279exc44

We generated imprecise P-element excision lines from the P(RS3)CB-6279–3 P-element. We used next-gen whole genome DNA sequencing to identify the endpoints of the deficiency that lie in the region upstream of the ss coding region (3R: 12,237,765) and in the second intron of the Pak3 gene (3R: 12,274,176).

Highlights.

• Spineless (Ss) requires its dimeric partner Tango to control Rhodopsin expression

• Ss is expressed in a stochastic On/Off pattern throughout the retina

• In the main region, Ss at high levels ensures exclusive Rhodopsin expression

• In the dorsal third region, Ss at low levels allows for Rhodopsin co-expression